Walk the walk (bipedalism)

[Andrew Mckee]

Hi all!

Just wanted to start a thread to tease out the fundamentals required for bipedal locomotion in the hopes that over time we can build up a useful resource for anyone embarking on the challenging task of making a large/lifesize humanoid walk.

I must first highlight, I'm under no disillusionment with regards to the difficulties in achieving this goal, the number of post grad engineers (mechanical, electronics, software, et al), the funding budgets such teams have to achieve this, the manufacturing capabilities universities have under one roof (or at least at that disposal). However, despite all this we all know at least CHARLI-L (by RoMaLe team) was built on a budget of $20,000, so we can deduce that there are some moderately advanced full-scale bipedal solutions almost within reach.

I hope we can collectively contribute & discuss ideas and concepts that we possibly may never have even thought about.

If this thread picks up enough momentum possibly Gael could organise discussion into the main sub-categories (ideal actuators for bipedal locomotion, biomechanic discussion, energy efficiency of bipedal locomotion, materials science -best biped materials to use, considerations of the upper body in bipedal balance, Important Bipedal Terminologies&Concepts-ZMP/Inverted Pendulum/Passive Dynamic locomotion/etc) to better organise/disseminate critical info & especially so anybody, with any interest whatever their background can come in, start reading from the basics, until they are up to speed and could very well contribute previously unthought of solution in one of the more complex areas.

That's the pitch, I hope this interests many, I'm certain there are many like myself who like the prospect of exploring solutions to what appears an insurmountable challenge, it wouldn't be stimulating otherwise.

Ps, I did a bit of digging around affordable actuators that 'may', with a bit of modification be potential candidates for this task. I'd like to leave this post up for a bit first to see if it garners enough interest (esp from Gael himself) and then I'll throw them out there for debate (also, to include them now would make an already long post messy, confused and/or unreadable).

Thanks for reading & I look forward to your contributions!

Regards,
Andy

[Kevin Watters]

Hi Andrew,

  It's great to meet you.  Thank you for posting this.  I think I speak for many that would love to see the InMoov walk.  I believe that as a community, we can make that a reality.  I see this as two main projects.  The first is the physical design of the legs and the actuators.  The second is developing the software to power it. 

  From a design perspective, the legs need to be strong, reasonably fast, but more than anything they will need to be accurate and properly calibrated.  This brings me to the software.  Once the dimensions of the legs are known, and the linkages are known, we should be able to derive the D-H parameters for the leg to come up with a forward and an inverse kinematic model.  I have already started much of that work in MyRobotLab.  By specifying the robot arms & legs in terms of the D-H parameter we can solve for the position of the hand/foot.  We can also compute a trajectory to move the hand/foot to any point in space  (that is within the operational space for the arm/leg)

  Next is on to balancing of the robot.  It seems like an obvious choice to add a gyroscope module to the inmoov to measure roll / pitch and yaw of the torso.  By using a feedback control system we can/should be able to keep the robot balanced.  I'm particularly interested in the mathmatics related to an Inverted Pendulum.  It seems to me that we should be able to model a control system for balance in the same manor as the inverted pendulum.

  I have listed below some links to my work in MyRobotLab.  I'd love to collaborate with more people on that work. I also included a nice video from MIT open courseware.  Thanks for starting this thread!  I look forward to seeing what we come up with! 

http://myrobotlab.org/content/joystick-control-inverse-kinematics-service

https://www.youtube.com/watch?v=D3bblng-Kcc

https://en.wikipedia.org/wiki/Inverted_pendulum

[Mats Önnerby]

Hi.

Thaks for starting this thread.  

I also really want to see InMoov walk. It is an extremley difficult task, but I think that looking at how humans are constructed and how humans walk can be key to making progress. To be honest I think we even should consider how to make InMoov run. You may think that I'm insane, and that i'ts totally impossible. My point is that if you set the ulimate goal, then you can work in the right direction, and skip steps that go in the  wrong direction. What I have seen with other robots is that they have wide, heavy and clumsy feet, to make balancing easier. To me that is a step in the wrong direction. It's just a way to ignore the balancing problem, not solve it. If you had the goal to make the robot run, would you even consider clumsy feet ? I don't think so. I think light weight feet are key to make a robot walk well. If you were to carry two bags of something heavy, would you prefer to carry it in you hands or on your feet ? The obvoius answer is in my hands. Nobody would carry heavy things on their feet. 

/Mats 

[Andrew Mckee]

Hi Kevin & Mats,

It's a pleasure to meet you both! If this thread can attract enough talented individuals, with positive attitudes & enthusiasm, to tackle this monumental task, I'm sure great strides can be made in this area, and what an achievement that would be for us all!

I couldn't agree more Kevin with regards to the mechanical leg design, the actuators & the software required. I think possibly the greatest difficulties in large bipeds is that each of these key categories branching out into further subcategories in their own right, and each design choice invariably has a significant impact in another category altogether. Despite that I'm sure good communication between categories will minimise this impact.

Funnily enough the 3rd link you attached (inverted pendulum wiki) I'd recently read, it's an interesting & classic engineering problem. It did lead me to wondering, for the purposes of bipedal locomotion, should the body be considered as an inverted pendulum (torso), balancing atop a regular, albeit jointed pendulum (swinging legs)?

Another aspect worth considering (I don't believe I've seen discussed) is the tendency to always the twist torso into/lead with the shoulder on same side of the trailing foot, with the torso/shoulders performing a rhythmic twist during gait.

This pdf breaking down important factors of walk is quite a good read (and not OTT in length or content).

http://courses.washington.edu/anatomy/KinesiologySyllabus/GaitPhasesKineticsKinematics.pdf

I'll start to organise the servos & dc gear head motors I've come across so far. I'll attach a link for each, price, potential max stall torque, additional gear ratio required to achieve max torque, seconds per revolution on output shaft after gear stages. Whatever actuator seems most suitable it's worth noting some work/modifications will be required to get necessary feedback loop control beyond position feedback (ie, speed, voltage, current, temp etc).

One aspect regards to the actuators I wouldn't know where to even begin on thinking about is exactly which communication protocol would be most suitable for comm's with controller, but I suppose we're a ways off that decision.

Mats, I'm very much of the same view regarding the biomechanical design influence on the legs. I'm trying to find the energy efficient/mechanical advantage methods the human body employs, to reduce stress on actuators, minimise required power & subsequently maximising duration.

The idea crossed my mind today if it would be worthwhile pursuing the DarwinOP design, however upsized 200-300%, employing our own modified actuators. Seeing as the mechanical design is proven & sound, it would possibly allow us to break the challenge down into more bitesize, manageable parts. Ie, the initial task would mainly be to produce a cheap actuator, with the necessary torque to manage the increased mass of a larger platform, and incorporating all feedback features ordinarily reserved for extremely expensive smart servos.

Just a thought, but my thinking is the full mechanical design is already there, it's simply a matter of agreeing the % upscale of DarwinOP, then work can begin on actuator solutions. At least if we were to fail on that it would be a quicker journey to that realisation of failure. As opposed to the potential (lengthy & expensive) 'fools errand' of beginning every key category from scratch. If we can make an upscaled DarwinOP work, with all the invaluable lessons learned along the way, & ever emboldened with success we can attack a scratch-built mechanical design tailored more specifically as an InMoov platform.

Do you's think there's any merit or method in the madness to upscaling existing platform, can you's see any school-boy errors in my rationale? Please, don't hesitate if you's see any pitfalls in this approach.

Thanks again for reading, I'll try to keep it shorter in future post 😏

Regards,
Andy

[Kevin Watters]

Hi Andrew,

   I think we have a communication protocol that should be able to do what we want with MyRobotLab,  (It's called MRLComm).  It's currently being updated to support more types of motor controllers.  I have been doing some experiments with it where we've attached a DC motor to an Arduino and a potentiometer attached to one of the analog pins.  This allows us to choose any motor/motor controller that we want.  Effectively, these are the building blocks for a servo.  I have found some 12V DC gear motors for cheap price that seem to have a considerable amount of torque.  It's possible that a magnetic or optical shaft encoder might be more precise to get angular feedback, but those start getting more expensive.

  There are a lot of existing humanoid robots out there, it definitely makes sense to look at them as references to learn from.  As for the inverted pendulum, yes, I think the torso should be modeled as an inverted pendulum balanced on a platform.  The platform roll,pitch and yaw would be controlled by the kinematics of the two legs that it's attached to.  So, from my perspective, if we build legs hooked up to a platform, and if we can control the orientation of that platform, then we should be able to provide the feedback to maintain balance, perhaps as simply as using some PID based controllers to make the system "stable" ...  (we also have PID based control in MyRobotLab, currently, it's used for object tracking,  I've seen Arduino projects that use a PID to balance an inverted pendulum.)

  Best,

   -Kevin

[Andrew Mckee]

Hi Gents,

Was just running some numbers out of curiosity regards upsizing 2x different platforms, the DarwinOP & the Kondo KHR-3HV. My thinking was a DarwinOP scaled to factor of 3, and/or a KHR-3HV scaled to a factor of 4, bringing both closer to more humanlike proportions.

Whilst doing this and thinking about the required torque to animate the upscaled platforms I initially thought (foolishly), if for example we were looking at the DarwinOP platform with original height of 455mm with a view to increasing to 1365mm (3x times), scaling up by 300% will not increase the height 3x times, as we're increasing the volume 3 dimensionally.

For example, if we imagined a 10mm cube, and we wanted to scale it up so we had a 30mm cube, if we had a number of 10mm cubes to build the larger 30mm cube, we'd start by arranging 3x 10mm cubes side-by-side (width), then adding 2x more rows in the vertical axis (height), then repeating last 2x steps two more times (depth). Once we'd completed our 30mm cube we'd have used 27x of our smaller 10mm cubes, essentially scaling up by 2700%.

It's very important fundamentals like these are understood, a simple oversight on my behalf could have led me to think "oh, a standard DarwinOP weighs 2.8kg, we want to upscale by factor of 3, right so 2.8kg times 3 equals 8.4kg", or "the standard DarwinOP servos have a torque of 2.5Nm, so we're going to need actuators capable of 7.5Nm". Obviously a seemingly innocent mistake as this will lead to major headaches & expense thereafter.

Is there any errors in my thinking here or is my line of thought correct in this aspect??? If this logics correct I'll swiftly write a follow-up post on estimates on the DarwinOP & Kondo KHR-3HV respectively.

Thanks again,
Andy

[Andrew Mckee]

Hi Kevin,

Sorry, on the motor control, have you discovered any cheap motor controllers that could manage a magnetic encoder, current sensor, temp sensor & manage to crunch numbers from magnetic encoder to have speed control in addition to position control?.. Also, should the controller be at least 32bit?

Would it be wise to work on a couple of variants of the actuators, differing only in their communication protocol? This would mean we'd effectively be casting a wider net in terms of teasing out the pros & cons of each protocol.

If we could make a Half duplex asynchronous serial communication & an ICS3.5 protocol variant of actuators (in addition to the MRLcomm version) it would allow us to easily experiment with the DarwinOP's CM-730 management controller and/or the Kondo KHR-3HV's RCB-4HV controller board. I'm not entirely sure if the Kondo code is open although the DarwinOP most definitely is.

If there were more options regarding actuator protocols, it would open up many more investigations/experimentations to ensure as many avenues as possible are explored.

Have you come across the OpenServo project? Website hasn't been updated in a while, looked very promising at the time, not sure if the project just died or creators went commercial? I did see a guy appear on The Poppy Project forums on a thread (titled something like "Alternative servo's") that had claimed to be releasing an affordable smart servo in the coming months, it sounded like it was related to OpenServoproject, however that was nearly a year ago and hasn't been any follow-up on that servo?

Thanks,
Andy

[peter heim]

I have fitted magnetic encoders to Inmoov gearboxes they seem to work well (still need to do more testing)  and tune the PID loop

[Andrew Mckee]

Hi Peter,

That's what I'm talkin about!.. Awesome work! What type of resolution are you getting from your mag encoder?

Andy

[revwarguy7]

Very interested in walking/running.

I think the CHARLI paper shown here on the equations of motion and its kinematics looks pretty complete, at least to start - see any issues there?  I note they used the Dynamixel networked servos that have extensive power, temp, position, etc. information made available, and that they are $500 a pop times a number of needed servos.

One thing I would suggest, and would be happy to help with, is making sure that whatever source that is considered for the motors that they not be some surplus source with a limited supply or distribution, but something that is mass produced and therefore easily available at a low price.  Motors like the ones they use in car windows, seat adjusters, winshield wipers, etc. come to mind.  Encoders are getting much cheaper than just a few years ago (like this one)

Adding a small SBM like a teensy or nano to control, read the encorders, and network them would also probably save some money while increasing our options.  In otherwords, a more powerful DC motor version of where OpenServo left off.

[peter heim]

The encoder resolution 12 bits or about 0.1 degrees not bad for a encoder that me AU$15 to make

[Kevin Watters]

Very cool that CHARLI paper lists out the D-H parameter table.  It also uses the jacobian for the forward/inverse kinematics. Seems to be basically the same approach that I've started working on in MRL.  So, good news is, so long as we get the description of the leg as a table of DH parameters, we can compute the forward/inverse kinematics.  Generation of a GAIT after that is reasonably easy for simple GAITs.  The next and probably more important part is the dynamics of the system.  (Euler La Grange & ZMP control loops.) 

Here's a link to my implementation.  I'd love any code review/comments that people have about it!

https://github.com/MyRobotLab/myrobotlab/blob/develop/src/org/myrobotlab/kinematics/DHRobotArm.java#L191

As for the encoder:  Magnetic encoders (or even optical encoders) seem like they might be more accurate and stable.  After being at a few Maker Faires, there is often a tesla coil that generates a lot of electronic noise that could interfere with the operation of the encoders.  I like how the current InMoov arm puts the encoders on the joint, rather than on the shaft of the servo. I believe this is referred to as "load encoding".  One of the benefits is that you don't need to account for the gearing ratio because you're actually measuring the angle between the linkages.  (not a big deal, but it's a design point).  An analog potentiometer is cheap (1-2 $ for a decent quality one).  The only thing that dictates their resolution is the A-D converter that you use to sample it.  (Current Arduino is only 10 bit AD resolution, or 1024 steps or about 0.35 degrees per step.  The Arduino Due and Zero 12bit AD resolution equating to 4096/360 or 0.09 degrees per step.)  

[revwarguy7]

Hi Peter,

I looked at your "Magnetic Encoder" entry on your log yesterday, but it doesn't say too much about them.  Could you provide a link that has more info about them or for their source?

TIA,

[revwarguy7]

Yep, it sure is more of a 12 bit world than a 10 bit one.  There are some sheilds that offer better A/Ds, but that usually kills the smaller form factor.  I am sure we can find something suitable, though.  Haven't had much luck with pots being very repeatable, myself, but they are inexpensive.

[peter heim]

Its based on this http://www.thingiverse.com/thing:5609

but uses AS5045 encoder I made some changes so it fits the Inmoov gearbox

I changed the gearbox as well to include 2 bearings to remove as much freeplay as possible

I will release the complete project on thingiverse when I'm finished the rest of the gearbox mods

Peter

[Andrew Mckee]

Peter,

0.1 degree's "not bad"? It's incredible!!! So you'll have 3600 positions per rev, for $15AUD!!!

Is it possible to add more feedback control on your board (current sensor, temp sensor etc)?

Have you any idea how servo manufacturers build velocity/speed sensing?

My understanding is positional feedback sensors will let the motor know its 'current position' & 'target' position, read the variance between these 2 positions as an 'error' value, and then rapidly reduce this 'error' value until this values 'zero' reaching its target position.

I'm sure speed control can be done with just the magnetic encoder (likely in code side), I've just never seen it done before???

Thanks for sharing your progress btw!

Andy

[peter heim]

Hi Andy

The motor controller that I building will have velocity and position control with some form of current sense 

and RS485 interface instead of I2C that I use at the moment. the AS5045 encoder has  quadrature output

which I have converted to RPS to control the velocity by clipping the output of the position PID loop

12 bit  is 4096 counts per turn

Peter

[Andrew Mckee]

revwarguy7

Thanks for posting the CHARLI pdf link, I'd never seen it before and had loads of interesting info to compare against numbers I was getting earlier with respect to the DarwinOP & Kondo mechanical specs. I've recorded all calculations so far on both, and subsequently added the CHARLI-L & CHARLI-2 after you posted the link to that interesting paper.

I'm going to post 4x consecutive posts so anyone who's interested can cast their eye over it or glean any extra info I didn't pick up on or calculate. I'm sure there's a few grammatical errors & possibly 1 or 2 mistakes in my math, if you do spot any encourage pointing it out so it can be rectified. I'll post in following order; DarwinOP, Kondo, CHARLI-L & CHARLI-2. Notice the torque increase on the CHARLI-2 btw

Andy

[Andrew Mckee]

DarwinOP (3x)

DarwinOP (2700% Upscaled)
Overall Height: 1365mm
Overall Width (arms ext): 1626mm
Leg Length: 658mm
Foot Width: 204mm
Weight: 2.8kg

Regular OP Servo Torque: 2.5Nm Dynamixel MX-28 (Half duplex asynchronous serial communication)
0.025Nm per 1% body weight
Servo Weight: 72grams
Dims: W35.6xH50.6xD35.5
Number of Servo's: 20
Total Servo Weight: 1.44kg
Rpm: 55 (@12V)
Regular OP Weight: 2.8kg
Servo's accountable weight: 51.4%

2700% OP upscaled Servo Equivalent Torque: 67.5Nm

Proposed Actuator:
Link;
http://www.robotshop.com/uk/198rpm-1200-oz-in-planetary-gearmotor.html
Price: $69/£57
Rpm:198
1200oz-in
Additional Gear Ratio: 7:1
Increasing Torque to: 59.3Nm
Reducing Rpm to: 28.3
Weight: 725grams
Number of Servo's: 20
Total Servo Weight: 14.5kg
2700% OP Equivalent Weight: 75.6kg
Servo's accountable weight: 19.18%

Note at this point the gulf between the standard sized DarwinOP & the upscaled DarwinOP 'Servo Accountable Weight'. Significantly reducing the '2700% OP Equivalent Weight' will balance this out proportionally again. The new concern becomes "how significantly does this affect the swinging mass (inverted pendulum)?". Time will tell,..

Revised upscaled weight: 28.21kg
0.675Nm per 1% body weight
Revised Servo Spec;
Dims: W106.8xH151.8xD106.5
Weight: 725grams
Total (20x) Weight: 14.5kg
Torque: 67.5Nm
Rpm: 55-60

[Andrew Mckee]

CHARLI-L

CHARLI-L (100% scale)
Overall Height: 1410mm
Overall Width (arms ext): 1354mm
Leg Length: 775mm
Foot Width: 140mm
Weight: 12.7kg

CHARLI-L Servo Torque: 10.5Nm
(2x per upper leg joint, plus 5Nm w/4-bar linkage giving 26Nm)
Dynamixel EX-106+ (RS485 Asynchronous Serial)
0.26Nm per 1% (127g)body weight
Dims: W40.2xH65.1xD46
Servo Weight: 154g
Number of Servo's: 18 (legs only)
Leg Servo Weight: 2.77kg (legs only)
Arm Servo (x8) Weight: 0.28kg/0.5kg
(3.6Nm/6.3Nm each)
Head Servo (x3) Weight: 0.22kg (3.6Nm each)
Total Servo Weight: 3.77kg
Servo's accountable weight: 29.7%
Rpm: 69.93 (@18.5V)
Lower body weight: 6.8kg (53.5%)

[Andrew Mckee]

CHARLI-2

CHARLI-2 (100% scale)

Overall Height: 1410mm
Overall Width (arms ext): 1354mm
Leg Length: 775mm
Foot Width: 140mm

Weight: 12.1kg

CHARLI-2 Servo Torque: 10.5Nm
(2x per upper leg joint/1:3 gear ratio giving 63Nm)

Dynamixel EX-106+ (RS485 Asynchronous Serial)

0.63Nm per 1% (121g) body weight

Dims: W40.2xH65.1xD46
Servo Weight: 154g
Number of Servo's: 18 (legs only)
Leg Servo Weight: 2.77kg (legs only)
Arm Servo (x8) Weight: 0.28kg/0.5kg
(3.6Nm/6.3Nm each)

Head Servo (x2) Weight: 0.14kg (3.6Nm each)
Total Servo Weight: 3.69kg
Servo's accountable weight: 30.5%
(Note: Servo's accountable weight marginally higher than CHARLI-L due to 0.6kg weight reduction in CHARLI-2 version)
Rpm: 69.93 (@18.5V)
(Bear in mind Rpm is from EX-106+ datasheet spec, not AFTER the 1:3 geardown/ratio used on CHARLI-2 which would result in 23.3rpm)
Lower body weight: 6.8kg (53.4%)

[Andrew Mckee]

Kondo didn't upload first time round,..

Kondo KHR-3HV (4x)

Kondo KHR-3HV (6400% Upscaled)
Overall Height: 1604mm
Overall Width (arms ext): 1524mm
Leg Length: 778mm
Foot Width: 173.6mm
Weight: 1.5kg

Reg KHR-3HV Servo Torque: 1.37Nm
KRS-2552RHV (ICS3.5 protocol)
0.0137Nm per 1% (15g) body weight
Servo Weight: 41.5grams
Dims: W22xH41xD30.5
Number of Servo's: 22
Total Servo (22x) Weight: 0.92kg
Rpm: 71
Regular KHR-3HV Weight: 1.5kg
Servo's accountable weight: 61.33%

6400% KHR-3HV upscaled servo Torque: 87.7Nm

Proposed Actuator:
Link;
http://www.robotshop.com/uk/12v-512rpm-681-oz-in-planetary-gearmotor.html
Price: $69/£57
Rpm: 512
681oz-in
Additional Gear Ratio: 17:1
Increasing Torque to: 82.35Nm
Reducing Rpm to: 30rpm
Weight: 635grams
Number of Servo's: 22
Total Servo Weight: 13.97kg
6400% KHR-3HV Equiv Weight: 96kg
Servo's accountable weight: 14.55%

Note at this point the gulf between the standard sized Kondo & the upscaled Kondo's 'Servo Accountable Weight'. Significantly reducing the '6400% KHR-3HV Equiv Weight' will balance this out proportionally again. The new concern becomes "how significantly does this affect the swinging mass (inverted pendulum)?". Time will tell,..

Revised upscale weight: 22.77kg
0.877Nm per 1% (228g) body weight
Revised Servo;
Dims: W88xH164xD122
Weight: 635grams
Total Servo (22x) Weight: 13.96kg
Torque: 87.7Nm
Rpm: 70

[Andrew Mckee]

CHARLI-2 stats

Overall Height: 1410mm
Overall Width (arms ext): 1354mm
Leg Length: 775mm
Foot Width: 140mm
Weight: 12.1kg

CHARLI-2 Servo Specs
LEGS
Servo used: EX-106+
Leg Actuators quantity: 18
Leg actuator Weight: 0.154kg
Total Leg Actuator Weight: 2.77kg
Torque each: 10.5Nm
Torque x2 per joint: 21Nm
Rpm: 69.93
Torque after 1:3 gear reduction:63Nm
Rpm after 1:3 gear reduction: 23.31

ARMS
Servo used: RX-64
Arm RX-64 Actuators quantity: 4
Arm RX-64 actuator Weight: 0.125kg
Total RX-64 (4x) Weight: 0.5kg
Torque each: 6.27Nm
Rpm: 61.73


ARMS
Servo used: RX-28
Arm (RX-28) Actuators quantity: 4
Arm RX-28actuator Weight: 0.072kg
Total RX-28 (4x) Weight: 0.29kg
Torque each: 3.6Nm
Rpm: 79.36

TORSO/WAIST
Servo used: EX-106+
Torso Actuators quantity: 1
Torso actuator Weight: 0.154kg
Total EX-106+ (1x) Weight: 0.154kg
Torque each: 10.5Nm
Rpm: 69.93

HEAD
Servo used: RX-28
Head Actuators quantity: 2
Head actuator Weight: 0.072kg
Total RX-28 (2x) Weight: 0.144kg
Torque each: 3.6Nm
Rpm: 79.36

The remaining mechanical/structural & electronics accounts for 68.1% (8.24kg) of total weight.

All-Up-Weight of body parts
LAYOUT: Each/%/Total/%
Thigh: 1.729/14.29%/3.458/28.58%
Shin: 0.783/6.47%/1.566/12.94%
Foot: 0.886/7.32%/1.772/14.64%
Torso: 3.565/29.46%
UpperArm:0.372/3.07%/0.744/6.15%
Forearm: 0.411/3.4%/0.822/6.79%
Head: 0.212/1.75%

All-Up-Weights (Kg/%)
Entire legs weight: 6.796kg/56.17%
Torso: 3.565kg/29.46%
Entire arms weight: 1.566kg/12.94%
Head: 0.212kg/1.75%

Servo Weight Totals (Overall%)
Legs: 2.77kg (22.9%)
Torso: 0.154kg (1.27%)
Arms: 0.79kg (6.53%)
Head: 0.144kg (1.19%)
TOTAL: 3.86kg (31.9%)

Structural Weight-kg/% (no servo's)
Entire legs: 4.03kg/33.3%
Torso: 3.411kg/28.19%
Entire arms: 0.776kg/6.41%
Head: 0.068kg/0.56%
TOTAL: 8.285kg (68.47%)

Summary;
Legs structure: 4.03kg/33.3%
Legs servo's: 2.77kg (22.9%)
Legs AUW: 6.796kg/56.17%

Torso structure: 3.411kg/28.19%
Torso servo: 0.154kg (1.27%)
Torso AUW: 3.565kg/29.46%

Arms structure: 0.776kg/6.41%
Arms servo's: 0.79kg (6.53%)
Arms AUW: 1.566kg/12.94%

Head structure: 0.068kg/0.56%
Head servo's: 0.144kg (1.19%)
Head AUW: 0.212kg/1.75%



Suggested actuators.
LEGS;
http://www.robotshop.com/en/planetary-gearmotor-720rpm.html
Quantity: 12 (6 per leg)
Cost: £60.46each (£725.52)
Weight: 0.79kg (9.48kg)
Standalone Torque: 2.12Nm
Gear Ratio required: 1:31
Achieving Torque: 65.67Nm
Achieving rpm: 23.22
PROBABLY TOO HEAVY?..

Or

http://www.hobbyking.co.uk/hobbyking/store/__31822__RJX_FS_0521HV_Metal_Gear_Digital_Servo_Ultra_High_Speed_and_Torque_21_3kg_0_05sec_68g.html
Quantity: 18 (9 per leg)
Cost: £59.62each (£1073.16)
Weight: 0.136kg (2.448kg)
Standalone Torque: 2.09Nm/Servo
4.18Nm/2x servos
Gear Ratio required: 1:15
Achieving Torque: 62.7Nm
Achieving rpm: 13.33
PROBABLY TOO SLOW?..

Or

http://www.robotshop.com/en/12v-780rpm-2400-oz-in-planetary-gearmotor.html
Quantity: 12 (6 per leg)
Cost: £46.37each (£556.44)
Weight: 0.635kg (7.62kg)
Standalone Torque: 4.81Nm
Gear Ratio required: 1:13
Achieving Torque: 62.53Nm
Achieving rpm: 39.38
GETTING CLOSER,..

TORSO;
http://www.robotshop.com/en/12v-780rpm-2400-oz-in-planetary-gearmotor.html
Quantity: 1
Cost: £46.37 (£46.37)
Weight: 0.635kg (0.635kg)
Standalone Torque: 4.81Nm
Gear Ratio required: 1:7
Achieving Torque: 33.67Nm
Achieving rpm: 73.14

ARMS;
http://www.robotshop.com/uk/roboard-rs-1270-digital-servo-motor.html
Quantity: 8
Cost: £62.34 (£498.72)
Weight: 0.07kg (0.56kg)
Standalone Torque: 3.43Nm
Gear Ratio required: 1:3
Achieving Torque: 10.3Nm
Achieving rpm: 90.9

HEAD;
http://www.robotshop.com/uk/roboard-rs-1270-digital-servo-motor.html
Quantity: 2
Cost: £62.34 (£124.68)
Weight: 0.07kg (0.14kg)
Standalone Torque: 3.43Nm
Gear Ratio required: None
Achieving Torque: n/a
Achieving rpm: n/a

Servo Costs:
Legs: £556.44
Torso: £46.37
Arms: £498.72
Head: £124.68
TOTAL: £1226.21

Servo Total Weight:
Legs: 7.62kg
Torso: 0.635kg
Arms: 0.56kg
Head: 0.14kg
TOTAL: 8.955kg

Note: As it stands the total servo weight is 5.1kg heavier (over double) than the original CHARLI-2 total servo weight (3.86kg)

[Andrew Mckee]

http://www.robotshop.com/uk/roboard-rs-1270-digital-servo-motor.html

Specs for 1x Servo
Torque: 3.43Nm
Rpm: 90.9
Weight: 70g
Price: £62.34p


Specs for 4x Servo gang
Torque: 13.72Nm
Rpm: 90.9
Weight: 280g
Price (per dof): £249.36p

Specs for 4x Servo Gang (w/1:4.5)
Torque: 61.74Nm
Rpm: 20.2
Weight: 280g (exc 1:4.5 geardown)

Dims: 40.4x20.1x40.3mm (per servo)
Quantity needed for legs: 36
Total Servo Weight of legs: 2.52kg
Total leg servo expense: £2244.24p

Or

http://www.hobbyking.com/hobbyking/store/__64645__RoboStar_SBRS_5314HTG_280_Digital_Metal_Gear_High_Voltage_Robot_Servo_53_1kg_0_14Sec_81g.html

Specs for 1x Servo
Torque: 5.2Nm (@11.1V)
Rpm: 71.43
Weight: 81g
Price: £25.51p

Specs for 3x Servo gang
Torque: 15.6Nm
Rpm: 71.43
Weight: 243g
Price (per dof): £76.53p

Specs for 3x Servo Gang (w/1:4)
Torque: 62.4Nm
Rpm: 17.86
Weight: 243g (exc 1:4 geardown)

Dims: 40.2x65.1x46mm (per servo)
Quantity needed for legs: 28
Total Servo Weight of legs: 2.27kg
Total leg servo expense: £714.28p

Or

(Bit of Wild-card but good specs)
http://www.goodluckbuy.com/sxsv50-80-24v-metal-servo-80kg-cm-large-torque-servo-for-rc-large-robots.html

Specs for 1x Servo
Torque: 7.84Nm (@V)
Rpm: 80 (@24V)
Weight: unknown?
Price: £59.65p

Specs for 2x Servo gang
Torque: 15.68Nm
Rpm: 80
Weight: unknown
Price (per dof): £119.30p

Specs for x Servo Gang (w/1:4)
Torque: 62.72Nm
Rpm: 20
Weight: unknown (exc 1:4 geardown)

Dims: 95x60x110mm (per servo)
Quantity needed for legs: 18
Total Servo Weight of legs: unknown
Total leg servo expense: £1073.70p



It's important to recall the CHARLI-2
leg servo specs at this stage. This should be kept in mind to ensure the solution/alternative remains cost effective, and as such worthwhile pursuing (as opposed to just buying 18x dynamixel EX-106+ instead).

Dynamixel EX-106+
Dims: 43x32x32.5mm (per servo)
Price (Per Servo): £337.30p
Quantity needed for legs: 18
Total Servo Weight of legs: 2.77kg
Total leg servo expense: £6071.40p

Thoughts, opinions, feedback on any of these options?..

Thanks,
Andy

[gael langevin]

Hi everyone!
This is a very interesting thread that keeps me busy reading!
I will let you guys keep evaluate and work computing options for the walk of legs, you seem to be knowing well what you are talking about!
There is nothing more motivating than "believing" into what you are making. InMoov shall walk!!

Just to let you know what I am working with for the legs. I will not be currently using servo motors for the legs unless I change of direction.
I'm using DC motors with encoders mounted on the motor shaft, and added angle positioning potentiometer.
I have been testing many solutions with various linear actuators and servos, but it has everytime been unsatisfactory regarding the four variables that I want to keep for the legs.




​
​I had found a super good deal in China for these, but they are too slow for the purpose unfortunatly, I also tried to modified the inner leadscrew to raise the speed but it was then not powerfull enough to sustain the InMoov body.

Low price, speed and force. An affordable motor that has speed and enough torque to sustain the weight of the InMoov body is not common at a low price.
The fourth important variable is, supply. It is important that we can get those motors easily from almost everywhere.

I had bought a bunch of second hand 18v drills for 1 euros each, that is a super deal for one that builts his own robot, but who will be able to supply the exact same drills?


These motors are great and powerfull, I have been able to use them to do all kinds of tests!

[gael langevin]

I have been testing also some electric scooter motors, but the weight of the motors become a problem at some point.


​

So I'm using planetary gear motors with encoders from AndyMark.
The price, speed, and torque is quit reasonable.

Another problem that I needed to solve was the space. Those motors are big, the total length with the encoder and planetary module is about 16cm to 18cm depanding on the gear box. Adding pulleys or lead screw to the shaft adds even more length.
But I have designed something to solve that problem that is pretty cool.

[gael langevin]

I have tried using Hbridge but was having problems to make it work properly with the potentiometer, I think there is some recent improvement done on MRL for that.  Recently I used some motor controller found on eBay which are pretty handy, but I'm trying to set the potentiometer on the circuit board to add the angle position.

Not sure yet how this is going to work.

I'm also using accelerometers for each motor like these:

​

​

Mmmh Posting all these pictures of my progress gets me into the mood of using my vacation for doing some tests instead of writing, ehehe!!

[gael langevin]

Ahh very nice Google, all my pictures are broken...

What a time lost...

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--

You received this message because you are subscribed to the Google Groups "InMoov" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+un...@googlegroups.com.
To post to this group, send email to inm...@googlegroups.com.
Visit this group at https://groups.google.com/group/inmoov.
For more options, visit https://groups.google.com/d/optout.

[Drew Rogge]

That's a drag Gael. Are the photos online somewhere? I'd really like to see them.

Drew

On 12/22/15 9:57 AM, gael langevin wrote:
> Ahh very nice Google, all my pictures are broken...
> What a time lost...
>
> Gael Langevin
> Creator of InMoov

> InMoov Robot <http://www.inmoov.fr>
> @inmoov <http://twitter.com/inmoov>
>
>
>
> 2015-12-22 18:50 GMT+01:00 gael langevin <hair...@gmail.com <mailto:hair...@gmail.com>>:

>
> I have tried using Hbridge but was having problems to make it work properly with the potentiometer, I think there is some recent improvement done on MRL for that. Recently I used some motor controller found on eBay which are pretty handy, but I'm trying to set the potentiometer on the circuit board to add the angle position.
> Not sure yet how this is going to work.
>
>

> I'm also using accelerometers for each motor like these:
>
> ​
>
> ​
>

> Mmmh Posting all these pictures of my progress gets me into the mood of using my vacation for doing some tests instead of writing, ehehe!!
> --
> You received this message because you are subscribed to the Google Groups "InMoov" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+un...@googlegroups.com <mailto:inmoov+un...@googlegroups.com>.
> To post to this group, send email to inm...@googlegroups.com <mailto:inm...@googlegroups.com>.

> Visit this group at https://groups.google.com/group/inmoov.
> For more options, visit https://groups.google.com/d/optout.
>
>

> --
> You received this message because you are subscribed to the Google Groups "InMoov" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+un...@googlegroups.com <mailto:inmoov+un...@googlegroups.com>.
> To post to this group, send email to inm...@googlegroups.com <mailto:inm...@googlegroups.com>.

> Visit this group at https://groups.google.com/group/inmoov.
> For more options, visit https://groups.google.com/d/optout.

--
Drew Rogge
dr...@dasrogges.com

Phone: 8934OOO629OO4829631OOOOOOO

[gael langevin]

Gee no, unfortunately!!

I need to re-do it all again.

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

[wilco van toorn]

Gael, take a break, have vacation, enjoy the season with your wife and children, go see family, and when you return after your break, you will pop up the legs in no-time.

But serious, i still am having a hard time believing my inmoov will walk, seeing the cost and it is a hobby agreed but i cant pay 1000+euros on servos, motors even if the result is a standing robot. But dont get me wrong, i believe it can be done with time, patience, working together and trying other methods outside the box.
Could something used in an exo skeleton be something, using springs, cables, pulleys and shafts. I think we had a discussion about it on mrl.
Because a dc motor, with cables and pulleys can be used to lift a car motor from the car, with not much force.

[wilco van toorn]

@Gael, do you know if the designer of the servoboard stil makes them and sell them? I know you did buy a couple, and i want to buy some to do testing with old cordless drill motors and encoders, servoboard and mrl, to make something like the dgital robotics servos that can be coupled in seperate adresses.

[revwarguy7]

@ Gael,

​I had found a super good deal in China for these, but they are too slow for the purpose unfortunatly, I also tried to modified the inner leadscrew to raise the speed but it was then not powerfull enough to sustain the InMoov body.

You may already be aware of it, but a great way to increase the speed of a linear actuator is to use a multi-start as opposed to a single-start leadscrew. Downside is they need a matching multi-start nut for the leadscrew as well, but these are also available.  Just by replacing a single-start to a 5-start leadscrew makes the actuator travel 5 times further with the same turn of the motor within the same space and unit of time.

[Andrew Mckee]

Revwarhuy7,

What effect would that method of speeding up linear actuators have on its standalone torque?..

Thanks,
Andy

[Andrew Mckee]

Hi Gael,

It's great to see you on this thread! Your work is an inspiration! I'm of the firm belief that because you've created InMoov, made the project open and accessible & built up a large community around InMoov, it's not a question of 'if', it's more a question of 'when' InMoov will walk!

Undoubtedly we all have an uphill struggle ahead of us to accomplish this mighty feat, but the journey there will be just as exciting as the goal itself.

Thanks for breaking fullscale robotics out of the laboratory & into the lives of ordinary people with a passion for robotics.

Thanks,
Andy

[Andrew Mckee]

http://www.hobbyking.com/hobbyking/store/__64645__RoboStar_SBRS_5314HTG_280_Digital_Metal_Gear_High_Voltage_Robot_Servo_53_1kg_0_14Sec_81g.html

LEGS

Specs for 1x Servo
Torque: 5.2Nm (@11.1V)
Rpm: 71.43
Weight: 81g
Price: £25.51p

Specs for 3x Servo gang
Torque: 15.6Nm
Rpm: 71.43
Weight: 243g
Price (per dof): £76.53p

Specs for 3x Servo Gang (w/1:4)
Torque: 62.4Nm
Rpm: 17.86
Weight: 243g (exc 1:4 geardown)

Dims: 40.2x65.1x46mm (per servo)
Quantity needed for legs: 28
Total Servo Weight of legs: 2.27kg
Total leg servo expense: £714.28p

TORSO

Specs for 1x Servo
Torque: 5.2Nm (@11.1V)
Rpm: 71.43
Weight: 81g

Torque (1:2): 10.4Nm
Rpm (1:2): 35.71
Price: £25.51p

ARMS

Specs for 1x Servo
Torque: 5.2Nm (@11.1V)
Rpm: 71.43
Weight: 81g
Price: £25.51p

Quantity: 8
Total weight: 0.648kg
Total cost: £204.08p

HEAD

Specs for 1x Servo
Torque: 5.2Nm (@11.1V)
Rpm: 71.43
Weight: 81g
Price: £25.51p

Quantity: 2
Total weight: 0.162kg
Total cost: £51.02p

Servo weight/%total of 12.1kg/cost
Legs: 2.27kg/18.4%/£714.28p
Torso: 0.081kg/0.67%/£25.51p
Arms: 0.648kg/5.36%/£204.08p
Head: 0.162kg/1.34%/£51.02p
TOTALS: 3.161kg/25.77%/£994.89p

Would it be possible to downsize the InMoov upper body to accommodate the CHARLI-2 dynamics, albeit InMoov will be around the height of an average 10-11year old, however this would still be a tremendous step forward

Using these potential servo's we can estimate the weight envelope of each of the scaled down InMoov bodyparts/regions.

CHARLI-2 AUW (inc dynamixels)
Entire Legs: 6.796kg
Torso: 3.565kg
Entire Arms: 1.566kg
Head: 0.212kg

InMoov structural weight (-Robostar servo weight)
Entire Legs: 4.526kg
Torso: 3.484kg
Entire Arms: 0.918kg
Head: 0.05kg

Clearly we can see from these calculations the total structural leg & torso regions are generally workable weight-wise. The arms & head not so much, however if we think in terms of burying the 2x servos per shoulder (4x total/0.324kg) on the inside of torso side of shoulder, furthermore if the 2x head actuators (0.162kg) can be located at the trunk of the neck our structural weights increase to a more workable region. See revised structural regions with changes,..

InMoov structural weight (-Robostar servo weight)
Entire Legs: 4.526kg
Torso: 3.16kg
Entire Arms: 1.242kg
Head: 0.212kg
Total: 9.14kg

Weight of downsized InMoov; 9.14kg
Entire Servo (39x) Weight; 3.16kg
Entire AUW; 12.3kg
101.65% of original CHARLI-2

If we alternatively maintained the CHARLI-2 structural weight, however using Robostar Servo's (39x) it looks like this,..

CHARLI-2 structural weight; 8.285kg
Entire Servo (39x) Weight; 3.16kg
Entire AUW; 11.445kg
93.05% of Downsized InMoov
94.586% of original CHARLI-2

Very interested to see current InMoov structural weight & perhaps if someone knew what factor InMoov would need scaled down to to maintain humanlike proportions at overall height of 1410mm, then we can calculate the downsized structural weight per bodypart/region & take it from there.

Thanks again, Andy

[revwarguy7]

There is no free lunch, but 4 start leadscrew vs a single start leadscrew is more efficient.  You often see them used with steppers because steppers have more torque at lower speeds than a regular induction motor, which has its max torque at higher speeds.

[peter heim]

Hi All

I think the problem with all electric motors is size and weight to get the joint to move fast enough with torque to do the job will be expensive

then the batteries will also heavy. I think you will need something like micro hydraulics  

https://www.youtube.com/watch?v=vvaYTVBHHYU

On Saturday, December 19, 2015 at 7:48:12 PM UTC+10, Andrew Mckee wrote:

Hi all!

Just wanted to start a thread to tease out the fundamentals required for bipedal locomotion in the hopes that over time we can build up a useful resource for anyone embarking on the challenging task of making a large/lifesize humanoid walk.

I must first highlight, I'm under no disillusionment with regards to the difficulties in achieving this goal, the number of post grad engineers (mechanical, electronics, software, et al), the funding budgets such teams have to achieve this, the manufacturing capabilities universities have under one roof (or at least at that disposal). However, despite all this we all know at least CHARLI-L (by RoMaLe team) was built on a budget of $20,000, so we can deduce that there are some moderately advanced full-scale bipedal solutions almost within reach.

I hope we can collectively contribute & discuss ideas and concepts that we possibly may never have even thought about.

If this thread picks up enough momentum possibly Gael could organise discussion into the main sub-categories (ideal actuators for bipedal locomotion, biomechanic discussion, energy efficiency of bipedal locomotion, materials science -best biped materials to use, considerations of the upper body in bipedal balance, Important Bipedal Terminologies&Concepts-ZMP/Inverted Pendulum/Passive Dynamic locomotion/etc) to better organise/disseminate critical info & especially so anybody, with any interest whatever their background can come in, start reading from the basics, until they are up to speed and could very well contribute previously unthought of solution in one of the more complex areas.

That's the pitch, I hope this interests many, I'm certain there are many like myself who like the prospect of exploring solutions to what appears an insurmountable challenge, it wouldn't be stimulating otherwise.

Ps, I did a bit of digging around affordable actuators that 'may', with a bit of modification be potential candidates for this task. I'd like to leave this post up for a bit first to see if it garners enough interest (esp from Gael himself) and then I'll throw them out there for debate (also, to include them now would make an already long post messy, confused and/or unreadable).

Thanks for reading & I look forward to your contributions!

Regards,
Andy

[Don Bio-Sensingbot]

Greetings all. With regards to balance, bending, walking, climbing stairs, walking obstacle compensation and so forth I would like to suggest a way I've been thinking about for a while with my own bot by using accelerometers and gyros. There are several applications as examples like Asimo from Honda https://www.youtube.com/watch?v=RRVZjdk5OsQ, DarwInOP http://www.robotis.com/xe/darwin_en as well as LIT Motors in the US where an inventor created a free standing untippable motorcycle with car cabin features. http://litmotors.com/ As you will see in the videos of this remarkable vehicle in th elink and on youtube https://www.youtube.com/watch?v=xWchOGECOno the inventor is very aggressive in his efforts to tip over the motorcycle but it maintains it's balance. The same could apply to our walking InMoov project. As mentioned in others posts the most cost effective means of walking stability is the goal. There are lots of ideas and relevant information in these links to glean from. I will begin researching here in the US for low cost options. I hope this helps.

[Don Bio-Sensingbot]

While researching alternative actuators, motors and servos I came across a few ideas that may have potential. 

Artificial muscles using a variety of different materials. With some additional research and testing something with these methods may work.

https://www.youtube.com/watch?v=yqCwHzK81dA

https://www.youtube.com/watch?v=Tba8Nf02OSI

https://www.youtube.com/watch?v=uw8FLgiXsmk

[Don Bio-Sensingbot]

On Thursday, December 24, 2015 at 3:30:41 AM UTC-8, Don Bio-Sensingbot wrote:

While researching alternative actuators, motors and servos I came across a few ideas that may have potential. 

Artificial muscles using a variety of different materials. With some additional research and testing something with these methods may work.

https://www.youtube.com/watch?v=yqCwHzK81dA

https://www.youtube.com/watch?v=6xZK2PeMWrE 

https://www.youtube.com/watch?v=Tba8Nf02OSI

https://www.youtube.com/watch?v=uw8FLgiXsmk

[Andrew Mckee]

Hi Don Bio-Sensingbot,

Welcome to the discussion & thank you for you contributions/links. I haven't been able to view all of the links provided yet (home wifi is acting up atm) but I agree there are some very interesting areas of novel actuation methods being worked on recently. One that especially peeked my own interest is this; http://www.riserobotics.com/mechanism
However beautiful no doubt this mechanism is, I'm sure significant torque could be obtained in the correct configuration, however an aspect I'm not soo sure of (and this goes for other current novel ideas, like artificial muscles) is the response time.

Even most hobby servo's have response time of 20ms, only some of the slightly pricier hobby servo's have 2ms response time.

In contrast the Dynamixel EX-106+ support up to 1Mbps communications speeds.

How 1Mbps translates to response time I'm unsure, but it's an important factor.

The last thing we need is to commit ourselves "All-In" with an actuator, blinkered by Torque, power, speed & cost requirements, and get blindsided by being ignorant of the response time.

If we found ourselves overcommitted to an actuator with inadequately slow response time this will cause tremendous headaches when it came to trajectory planning. Albeit, it could eventually be made to work via issuing predictive commands to actuators (to offset response time), however this would resign us to perfectly flat surface/known environmental working.

Well, technically issuing predictive commands to actuators could work on unknown surfaces with the aid of some extremely fast visual processing, although that will invariably over complicate an already difficult process.

Interestingly though, you can imagine how well humans nearly subconsciously use this method, although we can still walk with our eyes closed or in a pitch black room, just much slower & cautiously. This does however highlight that depending on a combination of motor control & trajectory planning alone is akin to asking someone to walk through an unknown room in the dark. It would be unreasonable to expect someone to blindly navigate an unknown environment whilst maintaining their usual (unblind) gait.

Another interesting quick read on this wiki;
https://en.m.wikipedia.org/wiki/Motor_control

Explains "it normally takes 0.4 - 0.6 seconds for the nervous system to activate available motor units to contract", furthermore it earlier explains that "Muscle fiber type/contraction speed;
I(slow)/100milliseconds
IIa(fast)/50milliseconds
IIb(very fast)/25milliseconds".
It's not entirely clear whether the 0.4-06 seconds is inclusive of the 25-100milliseconds?

The Kondo KHR-3HV that we've all seen Dr Guero's amazing humanlike walking gait, including traversing uneven terrain & reacting to unexpected outside disturbances, using KRS-2555HV servos over ICS 3.5 protocol at 1.25Mbps standard.

Thanks again Don Bio-Sensingbot, without your posts I'm not sure I would've even begun thinking about response times etc.

I suppose it's time to switch off for a day or two so enjoy your holidays everyone & I hope you get the socks/boxer shorts you've been hoping for all year! 😂

Thanks,
Andy

[Kevin Watters]

Quick comment.  I meet the guys from RiseRobotics.  They had a table next to me at the MIT Mini Maker faire.  I spoke with them about the InMoov project, they were interested in helping out.  I saw their actuator and even tested it out.  It was much stronger than me!  I attempted to keep it from moving and was unable.  It's a very impressive design.  I don't know what they've been up to lately, if we want to go down that path, I'm happy to try to reach out to them.

[Don Bio-Sensingbot]

I hope all of you are having a great time during your holiday season. Got home from visiting family and was excited to do a little more research for this project while the little woman is downstairs watching TV. :)  

Found something late last night of interest, nitinol/flexitol made from nickel-titanium. When an electric current or heat is applied it behaves like an actuator/muscle.

Nitinol/flexitol example: https://www.youtube.com/watch?v=IgIR76HyQT0

Linked is the white paper from Dynalloy Inc. http://www.dynalloy.com/ on the characteristics of this unique product. http://www.dynalloy.com/pdfs/TCF1140.pdf

This looks like a good contender for many of our actuator/servo needs at a reduced cost in electric current and overall electrical hardware.

I'm going to order some different sizes and lengths and see what results I get. If anyone else is willing to do the same we can compare notes. 

Again just another option to consider.

Best wishes.

[Kevin Watters]

Hi Don Bio,

  I ordered up some Nitinol and did some tests with it.  It was a quick realization after a few tests that it does not actually "contract" with electricity.  Actually what happens is, when it's heated up, it returns to the shape that it was at that temperature previously.  It's a memory alloy.  Basically, if you bend it into a shape and heat it up and then let it cool.  When you heat it back up, it will return to that shape.  The way you heat it up is by adding an electrical current to it.  The best you might be able to come up with is coiling it up like a spring and heating it up.

  This presents some challenges, 1. reaction times are not super fast.  2. in order to return to the initial state it needs to cool.  3. it's not a very strong movement.

Just wanted to share some experiences, I did look into this path also.

-Kevin  

[juerg maier]

Hmmm, what nice news to read about at these relaxing times of the year! Thanks Andrew for starting this!!!

Can't really add much to the discussion at the moment but wanted to make sure that you know others are watching this thread too! Once Gaël can make his pictures available and we have a bit more facts about dimensions, force and speed requirements and accuracy of sensors we probably could create a list of potential designs and configurations and have different people assigned to them. This would spread time and costs over the many of us and give us an idea which path to take when comparing costs and results? And maybe we can find a better way to represent and find the information about this subject then this rather long list of mixed entries (e.g. something like onenote or a wiki)?

- Juerg

 

[juerg maier]

@Kevin

Had a look at the Rise Robotics web page and it looks convining. Made me think about printing one to test it out ;-)

[Andrew Mckee]

Hi Kevin,

The RiseRobotics actuator is an amazing bit of engineering, I'd also guess it has a great power-weight ratio.
As they received the patent for their design recently I'm unsure in what way they'd be comfortable contributing.
I wouldn't wager they're going to upload their stl./obj files online anytime soon, but who knows, possibly if they released it under the same licenses Gael did with InMoov?

However, the RiseRobotics website does mention helping/contributing/empowering makers, hackers & enthusiasts alike so I suppose it couldn't hurt reaching out to see if they can see a way of contributing to InMoov community?

If RiseRobotics can see a way they can contribute it goes without saying. it would be of immeasurable help!

The differential conical drive would in all likelihood meet/exceed the actuator requirements of Gael's current leg design that he's been working on. It would probably be very mutually beneficial to both InMoov & RiseRobotics to collaborate in some way.

[Andrew Mckee]

@Juerg Maier,

Welcome to the discussion! You touched on a very good point with regards to finding a better format to present all relevant information. It's probably more of a fundamental aspect we should all give serious consideration & agree upon.

There are several reasons to this.

#1- The topic/challenge is such an undertaking, with soo many categories each branching out again into many more subcategories. This makes it necessary to format discussions in such a way that keeps information categorised, clear & on-topic.

#2- There are some (many) fundamental crossroads where consequential decisions are required, as such there would need to be a separate discussion area to a) raise those fundamental questions, b) discuss available options, & c) agree upon best available option moving forward.

#3- There will invariably be differing opinions on what is 'the best this', or 'the best that', or anything else being chosen to work with, work on, or otherwise pursue. It would be ideal to create a space for any person/s to pursue any ideas of a differing nature than to that of the main body of work. This will hopefully minimise discussions going off-track and provide anyone an opportunity to work on alternatives to the same problems, if alternative solution is successfully proven it may well be adopted as the solution in the main body of work (majority decision does not necessarily mean right).

#4- Initially this won't be a problem, once a substantial body of research, testing & analysis information is under our belt then it does become somewhat of an elephant in the room. That is to say, someone joins the group for the soul benefit of mining the key info of all our efforts, tweaking it just enough to pass as original, packaging it up with their company logo to sell on in the commercial marketplace, or better yet, knocking together a rough copy and launching a crowdfunding campaign as product in prototype/proof-of-concept stage.

That last point is sad & unfortunately true, many of us can probably remember in the not-so-distant past, a certain grey-haired fox that hung around the early 3d printer scene (repr community) that had what could only be described as an 'enterprising nature'. This point is key to ensure everyone enjoys contributing, with enthusiasm, and without the thought/worry of copycat artists.

Can I just stress that in no way am I suggesting anybody here is that unsavoury nature, although it has to be raised to maintain the free flow of information.

Has anyone any good suggestions/recommendations for best way to categorise, maintain & manage collaborative access to what will eventually be (hopefully) quite a large body of work?

Would it be worth suggesting each of us contributing to our own area of expertise & having access to that areas/category only, with possibly Gael being the overarching custodian of all categories? That way everybody is safe in the knowledge that each category contains at least some 'secret ingredients', but not 'all ingredients' plus all the kitchen utensils necessary to cook it.

Thanks again all for reading,
Andy

[peter heim]

There was a early design on thingiverse

http://www.thingiverse.com/thing:589587/

[Andrew Mckee]

Hey Peter,

I seen that Thingiverse upload myself. Interestingly it has since been removed.

Personally, I'd urge everyone to respect RiseRobotics IP and not pursue any efforts to reverse-engineer any of their work, unless of course they explicitly express permission to do so.

There is the potential for some agreement to be reached between Gael & Blake Sessions/Arron Acosta, however those discussions would probably be lengthy, and would also private. I wouldn't like to think that someone attempting to recreate their conical drive actuator potentially scuppering those discussions.

In the meantime, there's no issue whatsoever continuing discussions on other solutions. If Gael received contact from RiseRobotics I'm sure he'd update us if there was anything positive or potential for collaboration to report. If no word received back on that front at least there's been no time wasted twiddling our thumbs (waiting with abated breath on outcome).

Looking forward to hearing other opinions & perspectives on this.

Thanks,
Andy

[peter heim]

Hi Andy

It's a interesting design but if I was to spend any time on legs it would be with hydraulics the newer systems are far superior

to what we used in aviation 

Peter

[Andrew Mckee]

Hey Peter,

My knowledge/experience of hydraulic's is zero, but I think Altas is hydraulic driven, and PETMAN is pneumatic driven (if memory serves correct?). Bear in mind ATLAS has a TCOT (Total Cost Of Transport) of around 20, to put that into perspective humans have a TCOT of 0.2.

To get a sense of what is possible mechanically speaking (because let's face it, the human body is an incredibly efficient engineering feat in itself), during this years DRC away from the main event was a lower-key endurance challenge, one of the contenders was DURUS by SRI International, that achieved a TCOT of 1.5, and the team behind it is confident they can tune it to come in a just under 1.0. Moreover, when asked about what type of actuators they were using they were quoted to saying,..

"everywhere there’s sliding friction [like in a traditional gearbox], we tried to turn it into rolling friction instead, leading to an improvement in efficiency that is a few orders of magnitude better than traditional gearboxes"

This achieved 97% efficiency in the actuators, and here's the best part, using low-cost motors.


Personally speaking, I'd lean towards same direction as Gael suggests, using dc motors, geared down with feedback sensing added. I think the best solution lays somewhere therein.

However I find myself torn between the dc motor solution (that would take some lengthy & focused work to perfect and make repeatable for others), and the simpler 2-3 ganged servo approach. It's maybe ignorance to say 'simpler' as I suppose the servo's would need stripped, feedback sensors added & reassembled to working order, but my intuition tells me it would be quicker to put together a tangible test-bed, and begin gait testing.

That doesn't detract from the fact I think the dc motor route, if done right, is the better option.

Thanks,
Andy

[Amby Ambroise]

the link no work???

[peter heim]

"everywhere there’s sliding friction [like in a traditional gearbox], we tried to turn it into rolling friction instead

nice words sounds like a harmonic gearbox.

I played around with then a bit they are interesting but hard to make with just a printer they work OK with a geared input

but using a high speed motor they just vibrate and grind(needs bearings and balance) and higher precision overall  

Robbie;s new gearboxes have bearings which improves rigidity and friction but I still have the double gearing(geared motor and worm drive)

efficiency loss must be nearly 70%(most in the worm gear) we can gain back some by reducing friction (mostly in the worm and gear)

all this will add cost and weight 

Peter

[Andrew Mckee]

I interpreted the "everywhere there's sliding friction, we tried to turn it into rolling friction" as, for example in the ankles appear to absorb energy in a 'sliding motion', where that energy may well be stored and subsequently released, however in a rolling mechanism elsewhere.

I read into it as essentially saying, any area we were unnecessarily loosing energy (via, heat, noise or otherwise), absorbing that energy through sliding friction, and redistributing/translating that energy to rolling motion.

Of course it's wide open to interpretation, until SRI International reveals any further details, but one thing is for sure, they've cracked a winning combination to building the worlds best TCOT humanoid.

Considering Harmonic Drives and/or Cycloidal Drives, both with their pro's & con's, whatever way you chop it up, if you're creating a high ratio gearbox/drive personally I wouldn't attempt it without access to an industrial cnc mill, and preferably a lathe (manual is fine) too, a spark-eroder would be of benefit to try a few alternative approaches to traditional manufacturing approaches.

Because I don't own any of those (a Solidoodle Press only) that is why I prefer the 'almost direct drive' of the servo option, however that's a bias view based on tooling (or lack thereof) I have at my disposal. That's not to say someone here with the tooling could pursue a very suitable solution to the leg drives/gearboxes.

In actual fact, in a sense that brings up the collaborative work method once again. If people have any of the more expensive tooling at their disposal it would allow for a 'Machine Shop' area whereby anyone with an idea worth exploring could forward it to be machined to prove out concept & functionality.

Out of curiosity, does anybody here have machining capabilities above & beyond 3d printing?

Thanks,
Andy

[Andrew Mckee]

http://www.robotshop.com/en/banebots-first-cim-motor.html

Banebots FIRST CIM 12V 5280RPM 345oz-in Brushed DC Motor

$28USD (£18.92p)

Diameter: 62mm
Length (w/o shaft): 112mm
Shaft Length: 33mm
Shaft Diameter: 8mm
Efficiency: 65%
Stall Torque: 343oz-in
Nominal Voltage: 12V
Voltage Range: 6-12V
Kt: 2.98oz-in/A
Kv: 442.5rpm/V
Weight: 1.3kg

Additional Info;
5280 no load rpm at 12V (4A)

Torque(oz-in)-Speed(rpm)-Current(A)-Power(Wo)-Efficiency(%)
Free Load: 0-5342-2.4-0-0
Normal Load: 64-4356-23-206-75
Max Efficiency: 45-4649-16.9-155-76
Max Power: 173.4-2671-58.2-343-49
@STALL: 346.9-0-114-0-0

Proposed Gear-Down @ 67:1

Free-Load
Torque: 0
Speed: 79rpm
Current: 2.4A
Power: 0
Efficiency: 0

Normal-Load
Torque: 30Nm
Speed: 65rpm
Current: 23A
Power: 206
Efficiency: 75%

Max-Efficiency
Torque: 21Nm
Speed: 69rpm
Current: 16.9A
Power: 155
Efficiency: 76%

Max-Power
Torque: 82Nm
Speed: 40rpm
Current: 58.2A
Power: 343
Efficiency: 49%

@Stall
Torque: 164Nm
Speed: 0rpm
Current: 114A
Power: 0
Efficiency: 0


CYCLOIDAL-DRIVE EFFICIENCY

Single-Stage: 93% (Typically)
Double-Stage: 86% (Typically)

Single-Stage typically available up to ratio of 119:1.
Double-Stages typically available up to ratio of 7569:1.

A great benefit to cycloidal gears in robotics applications is their ability to absorb shocks without high risk of damaging the gear drive. They also suffer from a relatively small amount of backlash.

However Cycloidal drives are not typically back-drivable. Also, due to the off-centre nature of the internal mechanism, if not counterbalanced with a 2nd disk or counterweight, it will create excessive vibrations through drive causing increased wear on teeth & bearing of drive.

Factoring in an average efficiency of 90% for a single-stage cycloidal-drive to reflect the 10% loss (namely of Torque), however this time using the gear ration of 89:1. In brackets to right is Nm minus 10% lost through drive for a truer picture of torque achieved.

Proposed Gear-Down @ 89:1

Free-Load
Torque: 0Nm (0Nm)
Speed: 60rpm
Current: 2.4A
Power: 0
Efficiency: 0%

Normal-Load
Torque: 40Nm (36Nm)
Speed: 49rpm
Current: 23A
Power: 206
Efficiency: 75%

Max-Efficiency
Torque: 28Nm (25.2Nm)
Speed: 52rpm
Current: 16.9A
Power: 155
Efficiency: 76%

Max-Power
Torque: 109Nm (98.1Nm)
Speed: 30rpm
Current: 58.2A
Power: 343
Efficiency: 49%

@Stall
Torque: 218Nm (196.2Nm)
Speed: 0rpm
Current: 114A
Power: 0
Efficiency: 0%

Furthermore if we want the motor to be kept within its most efficient range of operation, we can modify the gear ratio to allow more useful torque.

Proposed Gear-Down @ 155:1

Free-Load
Torque: 0Nm (0Nm)
Speed: 34rpm
Current: 2.4A
Power: 0
Efficiency: 0%

Normal-Load
Torque: 70Nm (64Nm)
Speed: 28rpm
Current: 23A
Power: 206
Efficiency: 75%

Max-Efficiency
Torque: 49Nm (44.1Nm)
Speed: 30rpm
Current: 16.9A
Power: 155
Efficiency: 76%

Max-Power
Torque: 189Nm (170.1Nm)
Speed: 17rpm
Current: 58.2A
Power: 343
Efficiency: 49%


@Stall
Torque: 379Nm (341.1Nm)
Speed: 0rpm
Current: 114A
Power: 0
Efficiency: 0%


This motor is no doubt capable of delivering plenty of force at a reasonable stepped-down speed. Although running a max efficiency, one actuator will be spiking around 20A each, times anywhere up to 12 for full lower legs, that's maybe asking too much from any power distribution/management system, no?

Thoughts, opinions on power requirements,..

Thanks,
Andy

[juerg maier]

Not meant to disappoint anybody - but I get  the feeling we won't make it very far if we can not separate and channel the different issues about this subject.

@Gaël: we haven't heard much from you beside your try to give us some insight into

 existing layouts. 

1) Are you willing to try to share the pictures again with us?

2) Could we get a Wiki set up as part of your web site so we could at least start an index about the subsubjects of this? We will also need to agree on moderators for the different subjects or ask people to take over that part.

@Kevin: Read your referenced gait document and from what I understand from it and with a proposed mechanical design the travel distances, speed and torque could be calculated / estimated for the different joints? This could lead to minimum specs for the actuaters?

@Andy: Your very detailed specs about the motors need to go into a table or a set of tables and without knowing what we need I can't gain much from it. Beside distance, speed, torque and power we also will have weight and space restrictions?

We might also need actuators that use a combination of fast and rough positioning combined with a slower/finer positioning especially when the actuator has to take the requirements of keeping balance into account?

So to start off or to get "initial contact" why not agree on some geometrical basic facts of - I assume already existing - leg parts and see how far we can get with affordable parts in theory and then maybe also instanciated? For the practice part I assume it is necessary to have different members purchasing and testing the parts in order to also split the costs and time over many shoulders?

Juerg

[Andrew Mckee]

Hi Juerg,

I tend to agree. The thread in its current format has only really served to access the number of people with the common interest of achieving this goal.

I think it's fair to say this goal will be extraordinarily difficult (near impossible) for any one person to accomplish, and if it even were it would take an extraordinary length of time.

Spreading Costs wise, because it's a common goal we'd be working towards I'd guess the best (fairest) way to distribute costs would be once a proposed item of hardware is chosen to test and evaluate, the members of this 'Walk the Walk' group contribute an equal (fraction) amount to Gael, or whichever member may be performing test/evaluation work (via PayPal). It's the only fair way distributing costs, we'd be talking dollars or pounds per individual.

One things for sure, there's enough of us here with the interest to make this work, but it without a doubt requires organisation in a clear format.

Thanks,
Andy

[bruno....@free.fr]

Hi All,

Newbee in InMoov, newbee in bipedalism…

but i want to give you this link because the project is realized with 3D printer, servos with external potentiometer ;-) and some gyro helicopter model. it's walking !!

[bruno....@free.fr]

Oupps !!

https://www.youtube.com/watch?v=5oVN2puw59Q&list=PLpwJoq86vov-NYLfOhBh0JKrwfB1_i-RU

Regards

[Andrew Mckee]

Hi Bruno,

Welcome to the discussion & thanks for linking to XRobots great work!

Ah yes, James Bruton's work is incredible, especially when you read as far back to when he began the project before the advert of cheap desktop 3d printers, and he singlehandedly designed, chose components, bought, fabricated, assembled, tuned and implemented gait pattern.

James' undoubted wealth of knowledge in many area would be greatly beneficial to this group. I contacted him recently to see if he could offer any oversight in key aspects of bipedalism, do's & don't, what does work & what definitely doesn't. Hopefully he finds time to offer some helpful knowledge, we'll just have to wait.

Thanks,
Andy

[Andrew Mckee]

It's a bit ham-fisted but it will perform its duty to organise & categorise/subcategorise the multitude of topics requiring indepth analysis/discussion.

It is still very much a work in progress and not every topic will spring to mind, so any areas of importance you feels been left out just mention on the 'general discussions' and it'll be added fairly quickly.

In all likelihood a number of admins will be required to manage the forum, it would be preferable that potential admins had strong knowledge, experience or otherwise grounded understanding in topic/area of administration.

By no means do I lay any ownership to this forum, I may be putting the bare-bones of the site together & beginning to populate topics, boards, etc, but that is not to say that 'if' for example, I suffer from common forum admin issues becoming drunk with power (lol), I may be banned by a collective of admins that deem so fit.
This forum has NOT been set up any one individual, it is set up for the collective advancement toward a common cause, to document that work & hopefully ease the burden on future folks who share a similar interest.

Please feel free to visit the forum now, start an topic off, encourage anyone else you may know keen to contribute, as mentioned previously-post any suggestions you may have on the 'general discussions' board. I will probably be occupied for a while adding all suggested topics/boards so forgive me if I'm unable to contribute to the discussion for a brief period.

If you need to reach me directly you can e-mail me at; openingb...@outlook.com

Here's the url to the forum;
http://opening-bipedalism.freeforums.net

Thanks,
Andy

[juerg maier]

Hi Andrew

Great news, I hope we can motivate at least some of the core members to add their knowledge!!!

Thanks for this anyway - will watch it and hopefully contribute some.

Regards 

Juerg

[Andrew Mckee]

Hey Juerg,

I'm trying to complete as many 'Brief Overview' entries on the 'Case-Studies' board for quick comparison & easy reference.

It's on occasion very time consuming, however after a healthy repository of case-studies are collected it will probably make sense to create a standardised format to further compile each example into.

The hope is not only to collate as much relevant information, and identify 'commonality' between them, but to also importantly enable anyone in future to review, study & analyse this info much much easier.

Naturally it's still somewhat rough around the edges, however even with only five brief case-studies so far you can almost begin to see some vague patterns, or commonalities between what works or is otherwise successful, and what doesn't, won't or can't work.

I think, over time these 'patterns' and/or 'commonalities' will become more & more readily apparent, especially when utilising a standardised format for comparison.

thanks,
Andy

[Jaxon Klein]

Agreed on the strategy Andy. A bit of organization will go a long way with a topic as broad and complex as this. Thanks for setting up the forum. Good call.

I'm not sure how best to fit it into the topic structure, but for me, the one thing that has most been missing from this conversation so far is a discussion about our goals. In my mind, we either need to choose a unified goal, a few different goals, or choose a set of progressive milestones and goals. Bipedalism/walking is the broad goal, but there is a lot of specificity to determine therein.

Some examples of goal components in my mind are: top speed, robustness of consumable parts (I.e. maybe replacing a high friction gear after every X hours of use would be fine for our community and use-cases), general agility (do we want walking, running, stair climbing, rough terrain, etc), smoothness of movement, endurance, cost, etc.

Again, I'm not exactly sure how the conversation and decisions should be structured, but they will potentially drive a lot of the R&D focus, so good to get them out of the way early and avoid wasting any of the few man/woman hours we have available as a group.

I am going to do as you did when you started this thread and hold off on offering my opinion for now. I look forward to any thoughts you or the group have.

Best,
Jaxon

[Andrew Mckee]

Hi Jaxon,

Welcome to the discussion, and thanks for highlighting an extremely important point.

Indeed it is a vast topic, and insomuch probably why the important area of goals has been an oversight to date. And as you quite rightly point out, such a simple yet fundamental oversight as it is, would leave a collective effort such as this wide open to incompatible goals & expectations tearing it down the middle.

Entirety agree, the goals need set & expectations managed from the beginning. Again, thanks Jaxon for highlighting this point. I've been completely blindsided with trying to put together as many case-study overviews for ease of reviewing on the opening-bipedalism forum.

We could probably continue using this thread as a soundboard to discuss each & everyone's goals, hopes & expectations. Having been discussed & achieving a general consensus then a 'goals statement' can be drafted, previewed by everyone first before transferring statement across to the opening-bipedalism main page, so the goals/aims are crystalline from the outset.

thanks,
Andy

[gael langevin]

Hello Andrew,

If you search on the forum with the key words, "leg" "feet" and "walk", you should find many threads that have been posted since 3 years. Many of them contains links, images, suggestions which could be added to your case study. For exemple you will see that James Bruton has been part of some of our discussions over this subject.
In fact many of the links and technics suggestions that have been posted in this thread, are already posted in some other threads.

Re organising the InMoov forum has been something I wanted to do since almost two years now. It is one of our goal, with Sebastien whom re-designed the website, to integrate a good forum.

The main problem is to be able to import all the topics into the new forum in a way so we don't lose the work that has been already done.

Sometime ago, there has been someone who started to make another InMoov forum at some point, but it turned out to be for his own interested. Commercial banners and entrance fee was his goal.

If I knew how to do it myself, it would be done a long time ago.

Does someone know and can help us to do that?

I know this a bit off topic here...

-Sebastien creates a good forum

-Set together the various categories and subjects and what is primary important for the community.(we need everybody suggestions)

-Import all the Google threads into the new forum. (this is where we need help)

-Integrate the new forum into the InMoov website (That is not the difficult part)

If anyone knows how to proceed contact me: hairygael at gmail dot com.

Best wishes for this New Year to all of you!!!

Gael

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--
You received this message because you are subscribed to the Google Groups "InMoov" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+un...@googlegroups.com.
To post to this group, send email to inm...@googlegroups.com.
Visit this group at https://groups.google.com/group/inmoov.
For more options, visit https://groups.google.com/d/optout.

[Andrew Mckee]

Hi Gael,

I've searched all bipedal related words/terms and read those threads, however all have ceased to be active.

This was a main reason to start this thread, was to try to re-ignite that enthusiasm & keep an active discussion alive on the subject. However, even this threads flame has been a flickering ember at times.

The important thing that's come from this thread and agreed is because of the complex nature of the challenge, no single thread would be capable of managing every aspect of the subject & progress of work in a clear meaningful way. So its great we all agree we need a more structured system to manage this work.

Regarding that person who set up an InMoov forum for profit in the past, I never heard about it. I remember seeing those idiots on kickstarter using InMoov as part of their marketing campaign. People operating in such fashion will forever have bad luck imo.

Regarding the Opening-Bipedalism forum, I trust you believe me when I write there is absolutely no personal, financial, fame or otherwise motives setting it up. We had discussed the need for the more organised approach a number of times, I think I maybe even mentioned it toward the end of the opening thread post actually, but there was no real movement beyond talking about it so i thought if I didn't do it, it wouldn't happen at all. So you are safe in knowledge it was setup out of genuine interest. As you can see on the forum also, there is no mention or attempt of affiliation to the InMoov Project, this Google group, the InMoov website, any photos or videos or anything other that could be construed as connected you either yourself or InMoov.

Since I've been adding to the new Opening-Bipedalism forum (btw, can we just refer to it as the 'OB forum' from now on? :) much quicker typing) I've realised, there are that many scattered efforts currently working on bipeds to one degree or another, but without any solid tie-up between those efforts, I'd like to continue the OB forum to document all the current ongoing efforts/progress as best as possible in one central repository (not affiliated to any one group).

The amount of time & effort, searching, reading, searching, reading whilst running numbers on a calculator throughout it makes sense in my mind to capture that information which may have taken a day or two to gather, read & digest, and spend a further number of hours distilling as much as possible into something which will take the next person 3-5mins reading & digesting. That's not just efficiency in motion, it's to lay it out for people who otherwise just do not have the time or luxury of reading a 100-200 odd page long thesis very often. So in this sense, I hope I'm keeping at least a few people interested & enthusiastic about where technology is headed.

While on the topic of the Opening-Bipedalism forum, do you have or know 'off-hand' the weight figures for each major region of InMoov (waist up)? Because I have search here, your blog, and the Internet in general and cannot find a breakdown of some basic specifications. The reason I ask is because I was keen to create a Case-Study of InMoov and create a very speculative guesstimate of potential options to make him walk.

Thanks,
Andy

[gael langevin]

Hello Andrew,

Totally agree about re-igniting enthusiasm of the community about structuring a leg format through a new thread.

I really think it is also a good thing to create the OB forum, to gather information and pre digest all of the information available on the net, to help whatever community whom might be interested in bipedalism for robots.

Attempts of other forums:

​And the second, which has been put down:
https://groups.google.com/forum/#!searchin/inmoov/new$20forum/inmoov/9Ug9mLmfmag/v3K9JBRds5MJ

I currently do not have the weight of each region of InMoov's upper body. But, next time I need to dismantle him for traveling, I can weigh each part.

The upper body currently weighs 17,5kg with batterie in the back, computer and all devices. This was done with measured with the hips parts (Mid Stomach and Low Stomach parts)


​
​

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

Andy

[Don Bio-Sensingbot]

Greetings Gael, Andy and everyone else in this fantastic robotics community.

Regarding Gael and others suggestion in creating more enthusiasm, interest and involvement with this forum and various projects I would like to submit a few personal reasons which I am sure some of you share as being part of this community.

I would first like to say that this life experience while very challenging has been very good to me in so many ways. So to be able to give back to others from a spiritual and technical perspective means the world to me. 

As some of you know or suspect Mankind is in the midst of some serious paradigm shifts directly related to many aspects of our daily lives. From how we communicate to current and future job markets, global financial structures and health care. We are incrementally starting to experience more biological, environmental, financial, social and advancing technological stresses.

As a law researcher and student for many years I have come across interesting information that has pointed to these global shifts.

Every year the U.N., http://esa.un.org/unpd/wpp/publications/files/key_findings_wpp_2015.pdf and http://www.un.org/esa/population/publications/worldageing19502050/pdf/207unite.pdf most Nations and many Universities study global dynamic changes in ecosystems, economics, human behavior and various other demographics/statistics. When it comes to particular areas like those dealing with technology, like AI and Robotics displacing workers and the growing number of those aging out of the workforce without adequate numbers of new people to fill the gaps it hints heavily there are certain issues that need attention to avoid serious local, regional and global problems.

For my part I am learning to incorporate AI, Robotics and Bio-sensing for a number of projects that are directly related to enhancing quality of life for those who are now and will be joining those aging out of the work force (this includes many of us in this group). Fact is there will not be enough people in the coming 10 to 20 yrs to support the huge retirement numbers plus other dependencies which by 2025 is estimated at 58.6% of the US population. The increase burden on hospitals and medical facilities could reach a sort of critical mass causing even more problems. Technology coupled with compassion is a sure road to travel to aid many and help reduce some of these stresses. 

I spent 20+ yrs in the computer industry where I specialized in voice activated home and office automation. This is an area that will become a lot more valuable to those who want to remain at home longer with higher quality of life before greater care is needed. Bio-sensing is a critically important field to couple with. Imagine being able to not only mind text but control various devices around the home or call for help by thought. 

A couple of areas worth pursuing which Gael has started and generously offered is advancing Robotics and AI that can interact, perform household tasks and effectively communicate providing companionship to those who have no one else. Another is robotic home assistants who can perform needed biological and environmental scans and tests for those in the home to better monitor vital signs, airborne contaminants, security, etc.. and link directly to physicians and or monitoring stations for faster response times based on various parameters. There are so many sensors currently available and being developed that will provide these and other benefits. We who are the few globally interested in robotics development can provide quality solutions to these and other issues if we continue share our research and findings and work together to overcome obstacles. Open source is one of the few avenues to help ensure a more benevolent and sincere approach to global problem resolution.

Best regards,

Don Kieffer

[Andrew Mckee]

Hi Gael,

Thanks for understanding & appreciating my motives behind the OB forum, it means a lot.

That forum you linked photos of, I used to visit that website and I always was under the impression it was your/an official InMoov site? Just goes to show.

That's great you will weigh each individual part/area of InMoov the next time he's disassembled. Even with knowledge that the All-Up-Weight at 17.5kg, this is very promising news to me, I always thought he would have weighed more (20+kg), at 17.5kg I don't even think that scaling down is even necessary now.

The prospect of InMoov is becoming very exciting, I wrote a post discussing body weight distribution based on human proportions, and although not specifically wrote for InMoov (used theoretical humanoid for investigation), it certainly provokes many further ideas & design parameters (and limitations).

I hope you get an opportunity to review it and provide you thoughts/opinions, and anyone else for that matter.
Link;
http://opening-bipedalism.freeforums.net/thread/9/human-inertia-weight-distribution-design

thanks,
Andy

[gael langevin]

Hi Andrew,

You are correct, my official website looked almost the same, and the forum was going to be integrated into my website, unfortunately, we haven't been able at the time to find a solution to import all the Google threads. 

About the weight, if someone else could do the same, put InMoov on a scale, to see if we come up to a similar weight, it would be interesting.

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

Andy

[g.investi...@gmail.com]

Probably the other InMoov body will weight less (because most ppl not beeing able to afford the whole/same sensor/actuators/PC controller) so there won't be much trouble (those of us that will be lucky can add to their InMoov some protective stuff xP)

Going through all the information I seen here I will like to add some option for invert pendulum stability, not an original idea but this video inspired me:

https://www.youtube.com/watch?v=-wLgLVtQm0w

What James Bruton achieved here (I think he is not fully aware of what effect is making, because he didn't add a sping to the opposite force) is a dynamic stabilization thanks to the Kapitza's inverted pendulum.

I think, that is a half Kapitza's pendulum when its gounded but once is at ZMP it loses its oscillation so you will need stronger oscillations (stronger springs on both force directions, best would be the full +-x,+-y axis).

That of course is not elegant solution but it might be worth trying theorically. (Well maybe you can even advance that feature for robo-barman innate shaking skills XP )

Gael, would love to know why you give up the linear actuators for legs, James Bruton seems to have make the opposite of you, hes last test seem to be around linear actuators for legs right??

And about the PID controller for the dynamic movement, if theres no one working on it I would lovely give the equations if you can tell the model you are working with now (distances and radius are needed, if model simplified to rotational and linear moves), or if there is yet an official leg for it...

https://www.ez-robot.com/Community/Forum/Thread?threadId=7984

Could that be one? Is there in Github .stl files??

If so I coudnt find it, plz if someone got that files could you add the link?

Thanks for the great work, keep it up!

[gael langevin]

@investigation,
I did not abandonned linear actuators at all.
I am working with linear actuators but I am not using servos. I have tried various linear actuators from diffenrent shops, but I was disapointed with speed and price.


I don't know if you have this link in your :
https://github.com/LesUsinesNouvelles/InMoovLegs
It might be interesting for some of you. Files are not available and I haven't tested them.
The last time I have meet them, they had a leg printed and motorized, but I didn't see it working.

The main goal of the students on this project was to use printable parts, with very easy to find technology. If you check the shapes, you can tell they used InMoov shoulder for the hips and went on drawing their own parts. Some ideas are good. When we look at the video, you can tell that it is not fast moving enough for a leg.
Though it could very interesting to modify the bicep in that direction of thought.

[Bob H]

Hi Gael, Are you planning on re-posting the pictures? We would be interested to what track you are on with the legs. I've been working on an idea that involves planetary gears.

[love4...@gmail.com]

Thanks for the link!!

Will dive there as soon as I can :D

With a quick view seems a model that needs really a rework :S

The motors they are using seem a bit slow and weak.

Im not french, neither I undertand it well so maybe I failed to read something.

For my point of view, best will be some 30+kg servos paired for each joint gear and some 9DOF Imu from Bosch sensor, like sparkfun https://www.sparkfun.com/products/11486.

Altho of course 9DoF will get short, maybe for a start you can split the model into, hip joint IMU connecting the hip and low stomach servos to a IMU so you got your Inverted Pendullum balanced, then you could make for each foot joint (with leg) so you know the deviation for the sinusoidal/ZMP walking algorithm in case of disturbances. 

That way you can correct leg relative position for a known floor disturbance as well as correct the inertial force of the inverted pendullum in case of inertial disturbance (how well it works will depend on controller speed as well as torque as well as encoder precision).

That is the real hard part of walking methods, take account of disturbances, along with the dynamic inertial forces.

I think, I read somewhere that you were starting some kind of foot IMU for floor disturbances?

Maybe is not so bad idea...!

Cheers, keep it up!

[Andrew Mckee]

Hey Love4,

Being entirely honest, bipedal trajectory planning is a subject I know absolutely zero (and that's being generous to myself) about. My strengths are more aligned in the areas of mechanical engineering, my electronics is basic but enough to get by with, without killing myself.

Regarding the 2-gang servo's, indeed this is regularly done in humanoid builds, however for an adult sized biped, and an extraordinarily light one at that, 30+kg servo's ganged in pairs would still be woefully under-actuated.

Consider briefly the CHARLI-2, around 1.4M tall, 12.1kg total weight, main leg servo's (2-ganged on major load bearing joints) with just over 10Nm torque each, both on a 1:3 pinion on gear, we're looking at over 60Nm torque on said joints.

In comparison 2x 30+kg servo's ganged together would yield circa 5-7Nm (ganged), consider that was driving a 1:3 gear/pinion we'd be looking anywhere between 15-21Nm. Bear in mind InMoov is weighing in and around 17.5kg upperbody as we speak.

im not trying to be Mr Negatory btw, I just like to think out & plan for the highest chances of success. Mechanically speaking, I see no reason for InMoov to not walk. Judging by previous projects & examples InMoov for all intents & purposes can walk. The tricky but is identifying the correct combination to make this happen.

Warm Regards,
Andy

[Mats Önnerby]

Just to make some simple calculation. In this calculation I'm ignorant to the weight of the legs, so we need to addd that at some stage. I just want to show how to dimension. 

I measured my legs and the distance between my hips to my knees is about 40cm. But I'm tall, so lets calculate with a little shorter lengt of 30 cm => 0.3 meter. The weight of InMoov upper body is 17.5 kg => 175N, So the knees needs 175*.3 = 52.5 Nm => 525 kgcm. ( The unit normally specified for a servo ).

Since we have two legs, each leg would need 52.5 / 2 => 26.25 Nm. 

But that's not really a problem. You can gear any motor to get that torque. 

To make realistic dimensioning, we also need to take time into consideration. So we need to talk about power.

Power (watts)  = Torque (Nm) * speed (radians / second) .

If we define how fast the legs need to move 90 degrees, then we can also calculate how much power is needed.

So what is your guess. How fast do we want InMoov to move from sitting to standing ? How long does it take for you ? 

The formula above is only for constant speed. For acceleration we need more power. We can get back to that in a later post.

And I'm only calculation for the extreme case where the knees are bent 90 degrees. With well designed joints, the torque needed to stand up 100% balanced shold be 0.

Please correct me if you find any errors in the calculations above.

[peter heim]

Hi Mats

Think your torque calculations are a bit simple for the robot to walk it will have to balance on one leg so you need full torque.

Also 52.Nm is the minimum with no friction at max velocity, for a smooth walking gait you will need a lot more torque. For instance for my robots(http://escaliente-robotics.blogspot.com.au/) arm  to move smoothly to a position and then be able to adjust the final position with very small moves(1 to 2 mm) I need 100Nm.

I have tried with less but I have a lot of bounce and hunting.

Peter

[peter heim]

here is a calculator its designed for arms 

http://www.societyofrobots.com/robot_arm_calculator.shtml 

[Mats Önnerby]

My point was: To make proper dimensioning, we also need timings. 

 

[Mats Önnerby]

I think we need a simple model of the legs, with joints and distances, so that we can do similar calclations as you show for the arms.

[love4...@gmail.com]

Well thanks a lot for all the information!

@Andy: It seems a may be missing something!

 "2x 30+kg servo's ganged together would yield circa 5-7Nm (ganged)" that  is based upon the lenght of the model? or how you get those calculation if you dont mind explaining ? (im really lost at calculating absolute inertial numbers)

Thanks anyway Andy!

@Mats: I totally agree with you Power consuption will be a problem. I been cheking some PS of 12V and 30A to work with a downsteper 12-24V to 6V 30A also, can find those in ebay/amazon for 30$ each component more or less.  For that configuration we get 180W, maybe really not good enought? pairing a couple of those on parallel will lead you to 360W altho using 2x baterries ofc having that we can see...360W=26.5Nm*w ; w= 360(Nm/s)/26.5(Nm)=13.58 Nm/s ....:(  wich makes it even worse if we have 180W (know that the eFuel unit gives 20A...lower even!) w=180/26.5=6.8 Nm/s 

And as Mat said that is only for linear velocity! :'(

@Peter: Awesome to know you need 100Nm in one arm in order to make a smooth movement on those babes!

Ok so best scenario to acomplish will be something like speed of 50-100Nm/s?? (50Nm for linear smooth and 100Nm for the exponential smooth maybe?)

Thanks a lot for all responses!! 

Please notice that the tech is out there, right now! (ok maybe only for those with $$ XP)

Keep it up!

[love4...@gmail.com]

Oh man how could I miss Peter calculator link!

http://www.societyofrobots.com/robot_arm_calculator.shtml 

Thanks a lot, will have to study a bit to understand the simplifications! XP

Andy, do you know how to simplify a dynamic model into other dynamic model? Is all about DoF equations and searching linearity?? 

Thanks!

[John (Pole) Owen]

I have been following this thread for quite some time and seeing InMoov walk is of great interest to me as well.  I have very little background in robotics (mechanical or electrical) but I do have a solid background in coding and hope to be of some use in the future once my InMoov is more complete.  I have been following this thread and learning as much as I can from it along the journey.  I have also joined the OP board so I can follow along as this discussion develops.

One thing that may or may not have been mentioned so far is that the "total weight" of the upper body will likely need to include another power supply or possibly a larger power supply.  There has been some talk about power requirements thus far, but I don't think I have seen anyone mention the added weight.  

We might be able to offset some balance issues added weight would cause by locating a second PSU lower on the robot, thus lowering the center of gravity. 

Anyhow I just thought I would toss that out there for consideration.  I would be happy to be of help in any way I can. I know a bit about web hosting and forum maintenance if help is needed in that department, and I am always willing to research issues and help find solutions in that department.

[Andrew Mckee]

Hi Love4,

No worries, the 30kgcm servo's can move 30kg at 1cm from its turning axis centre, so a 2cm diameter wheel or gear on the servo would equal this 1cm from centre, and subsequently be able to affect a mass of 30kg.

To calculate kgcm to Nm try this calculator; http://www.convertunits.com/from/kg-cm/to/N-m

The 2x 30kg servo's when ganged will double output torque (minus small % of efficiency losses, but it'll get us in the ballpark & give us a good indication). If you enter 60kgcm on the kgcm-Nm calculator it works out at 5.88Nm.

The changing one dynamic model to another question, honestly I am not the person to ask as I don't have the first idea how to do that? Perhaps somebody else has experience and could help you convert what you need.

Andy

[Andrew Mckee]

Hi John,

Thanks for joining the discussion. Thanks for joining OB forum, I apologise I haven't been able to post any threads the last couple of weeks, usual story, back to work after holidays and getting myself back into 5th gear so updating threads/topics has unfortunately suffered.

I would not be too concerned regarding little mechanical or electronics experience, my electronics is very basic, my coding level is zero, so I only have mechanical experience to rely on, so essentially we're just in different corners of the same room :)

And a solid coding background is an essential ingredient in making InMoov walk a success, so every single person helping out and contributing is as equally important as the next person helping out and contributing in another discipline.

Very interesting points regarding the weight distribution of another primary source of weight (only after actuators) in the batteries. The RoMeLa team behind the CHARLI-L & CHARLI-2 recognised the importance of weight distribution and maintaining minimal inertia, they moved the batteries to the thigh/upper leg region to, as you said lower the centre of gravity & concentrate the inertia around the mid section. Another point regarding power we haven't touch on, again employed by the RoMeLa team was to partition the power between servo's & computational electronics, this was to minimise & isolate electronics 'noise' moving between systems & allow main board electronics do their computational work unaffected.

Andy

[Andrew Mckee]

Hi everyone,

Has anyone seen this guys open sourced esc (electronic speed controller to the uninitiated)?

http://vedder.se/2015/01/vesc-open-source-esc/

Some people have probably had a favourite pastime in multirotors & will be especially familiar with working out a good prop, motor, esc & lipo combo. The ESC was/is fundamental to making multicopter flight possible because of their extremely precise control of motor speed, enabling the craft to quickly respond to environmental disturbances & maintain level flying conditions.

This guy has become involved with an e-skateboard forum & using this ESC he created himself & open-sourced, built an e-skateboard (longboard? Whatever,..) has a video of himself travelling at considerable speed around his neighbourhood, block, street and kindly added an OSD (again, On-Screen-Display to those unfamiliar with the term) of the setups performance (ie speed, mAh consumed, Current draw, Voltage etc).

Initially it lead me to wondering would it be possible to use this ESC in combination with a brushless inrunner or outrunner motor, geared down to a usable torque & speed for our needs?..

Unfortunately I am ill equipped to answer that as I am neither an electronics nor software expert.

It would perhaps take someone with a mixed mechanical/electronics understanding as I worry brushless motors may perhaps have a poor starting torque curve that may eliminate it as a possible route to investigate.

It would be interesting to hear any and all opinions if anyone has the time to check it out.

Andy

[Andrew Mckee]

Here are another couple of interesting open-source motor control/servo projects;

Open Servo Project,
http://www.openservo.com/StepByStep
Or,
http://openservo.org/OpenServoV3?action=show&redirect=version3

DIYservo Project,
http://diyservo.com/home

And the aforementioned OS-ESC,
http://vedder.se/2015/01/vesc-open-source-esc/

A good motor at a reasonable price, requiring a reasonable gear-reduction is not especially difficult to find, however controlling the beast is an entirely different area I'm unqualified to assist with. I just guess that if there's an open-source control board, that will allow us to add the necessary sensors for feedback & control, then we've made a first LARGE step forward in finding out solution.

The majority (if not all) big problems can be overcome once they're broken down into their constituent components and tackled separately.

The old adage "divide & conquer".

thanks,
Andy

[Anton Fosselius]

Hello, i am currently working on a biped on my spare-time, i am starting with the foot and will work upwards

This is some of my current status: (just a draft, its quite under-powered right now)

I am using load cells (cheapest ones i could find on ebay) to detect contact with the ground and to track ZMP

Here are some good links:

http://www.takanishi.mech.waseda.ac.jp/top/research/wabian/

http://journal.frontiersin.org/article/10.3389/frobt.2015.00014/full

http://www.romela.org/charli-cognitive-humanoid-autonomous-robot-with-learning-intelligence/

http://www.cs.cmu.edu/~cga/walking/grad.html

http://letsmakerobots.com/robot/project/dc-motor-biped-12-dof-legz-3d-printed

https://www.youtube.com/user/DrGuero2001

[Andrew Mckee]

Hi Anton,

Thanks for joining the discussion! Very interesting design, the design of the toes reminds me of The Poppy Project's ones, they used a spring-loaded passive mechanism to reduce weight in no added weight of a servo there.

It is interesting because most people would probably think to begin with the main legs, and cater to the feet as an afterthought, both you and Gael began at the feet/ankles and work upwards.
But then the feet are probably the most difficult part of the lower body because there are so many considerations, achieving enough dof's and at the same time making them weigh essentially 0kg, it's undoubtedly a tricky design task.

If you have the time to read it have a skim over this thread with regards to weight distribution based on the human as a benchmark & bearing in mind maintaining a minimal moment of inertia from the design stage. By not means is it authoritive or a peer reviewed write up, but hopefully you find it insightful or find something useful within it.

http://opening-bipedalism.freeforums.net/thread/9/human-inertia-weight-distribution-design

Andy

[Anton Fosselius]

With the linear weight increase there is a quadratic cost increase, so i am trying to build an ~ 160cm tall < 20kg weight humanoid. I am also trying to minimize the amount of 3d printed parts and use threaded rods and pipes as much as possible. This will make it cheaper and faster to build. I am not so sure that we need that much torque on the motor if we use lead screws as actuators, because much of the force will be absorbed through the linear element and will not load the threads of the gears directly. I will play around with a design similar to "XRobots compliant linear actuator". I am not sure how to calculate the load on the leadscrew and the gearing properly but will try to do that later when i have built my first prototype. Got my 3d printer today, its a Delta kit so it will take a while before i have anything actually printed. 

I like the idea of using stepper motors for linear elements instead of BLDC because steppers are designed to have a high holding torque. i will update with some results from my experiments when i have done them.

[Andrew Mckee]

Hi Anton,

Very interesting! Yes, it's quite difficult ascertaining the exact torque requirements we need from the waist down joints.Eespecially when the human body utilities a number of muscles to produce a single movement.

I'm no electronics expert but I was also intrigued with the use of stepper motors for a while, however I'm under the impression that steppers have large power requirements, drawing power even when not in motion? The threaded rods advantage of reducing stress on the actuators is of course a very desirable property for our purposes. I'm very intrigued to know how you get on.

Ps, you gave me a good idea for a new thread on the OB forum. I'll post a link when I've researched it further & condensed the info.

thanks,
Andy

[Mats Önnerby]

When you calculate the w ( omega ), the result is the rotation speed expressed in radians per second.

So w=360W = 26.5Nm*w ; w = 360(Nm/s)/26.5(Nm) = 13.58 radians per second = 13.58 / 2*PI => about 2 rotations per second.

That is a fast movement for a leg. Torque is rotation, not linear motion.

However I think you are close to realistic numbers. A human beeing on a bike can deliver about 200-300W. InMoov should be able to deliver in the same range or a little less because of lower weight.

For a comparision, a starter motor for a car can deliver about 1400w, so that is a to big motor. 

A wiper motor for the windshield is rated at about 80W, so it is in the low end, but perhaps it can work. With one in each leg we get 160W.

The size of the motors for the strongest parts of the legs should be somewhere in that range.  For a foot, perhaps 20-30% of that could be enough to be able to stand on the toes.   

[Andrew Mckee]

Hi Matt,

Your handle on how power translates into moment force will be very helpful in our progress here. I don't have a firm grasp of this translation, perhaps you could write a 'dummies guide' type thread in the actuators-mechanical/electronics section on the OB forum for those of us unfamiliar???

I am however familiar with torque ratings & how they relate to a force about a distance moving around a point/centre. When I referenced the difficulty ascertaining the torque requirements of the lower body I was imagining a human body, and because there are many different muscles involved in the same movement of 1x DOF, it is easier to approach these DOF's as simple joints/force about a centre, as opposed to investigating the Newton force each individual muscle exerts in a linear push/pull motion and then calculate back from that. I suppose this is a personal bias as I haven't yet approached any linear actuator solutions, nonetheless I find it less brain/processor intensive to work that way. Of course others will prefer linear actuation & naturally prefer to think in terms of Newton force.

Andy

[Andrew Mckee]

Hi all,
I did a rough sketch (ie, not to scale) of human form with the inclusion of approx Weight of each region of the body as a percentage of the total body weight, and approx height (or length) of each body part as a percentage of the overall human height. It is only just as a simple quick reference template before I begin calculating more meaningful values in regards to requirements.

Whilst it only has simply % ratios for weight & height, and before I begin scribbling calculations all over it (and essentially making it look like the work of a raving lunatic) I thought it may be of use to anybody also investigating different design options/choices that are based along human like proportions (and perhaps saving time). I hope because there's only % ratios atm it can easily be used by others for general scaling purposes.

Apologies for last deleted post, realised the lower leg weight % value was for both legs when it should only have been per body part (leg). See amended picture (better lighting this time too).

Andy

[Anton Fosselius]

Take a closer look at http://journal.frontiersin.org/article/10.3389/frobt.2015.00014/full

especially this: http://www.frontiersin.org/files/Articles/130437/frobt-02-00014-HTML/image_m/frobt-02-00014-t002.jpg

and

http://www.frontiersin.org/files/Articles/130437/frobt-02-00014-HTML/image_m/frobt-02-00014-t001.jpg

and here is an visualization of mass distribution:

http://www.frontiersin.org/files/Articles/130437/frobt-02-00014-HTML/image_m/frobt-02-00014-g003.jpg

And from the conclusion:

"In this article we presented the latest progress of the development of our anthropometrically correcthumanoid robot Herbert. We have shown all the DOFs of the Herbert’s completed design, with a total weight of 35.5 kg and the breakdown of the €35,000 total cost."