Post mortem

[wholder]

So, what with having missed this year and the lack of a video feed for much of the early rounds, I'm curious to get a 911 on the new track layout and how everyone's "best laid plans" worked out in practice.  Specifically:

1. What new challenges emerged from the narrower, longer distance track.

2. How well did the GPS vs non GPS crowds do?

3. How did the new sensors (i.e. Bosch BNO055) perform in combat conditions.

4. Did anyone make effective use of vision?

5. Did anyone try collision detection and avoidance?  If so, how'd it work out?

Wayne

[Minuteman]

On Monday, June 22, 2015 at 4:27:03 PM UTC-6, wholder wrote:

So, what with having missed this year and the lack of a video feed for much of the early rounds, I'm curious to get a 911 on the new track layout and how everyone's "best laid plans" worked out in practice.  Specifically:

1. What new challenges emerged from the narrower, longer distance track.

It was noticeably tighter, but not too bad. 

 

2. How well did the GPS vs non GPS crowds do?

Non-GPS ruled the day, regardless of the bonus. Daisy Rover, JRover Maxx, Rosy, Roadrunner, Minuteman and Quixote were all top finishers, and were non-GPS. I think that goes for Auton as well. I think all the categories were won by non-gps bots, and maybe second place as well. And, I think that would have stood without the bonus. 

In the pits, I saw many 3DR modules and people using the route planner. It seemed like every other bot was using ardupilot. In general, they were not successful.  

 

3. How did the new sensors (i.e. Bosch BNO055) perform in combat conditions.

JRover Maxx had good success with that sensor, I believe. The MPU-6050 still does the job for Roadrunner and Minuteman ( and others, I believe). 

 

4. Did anyone make effective use of vision?

There was a team from google computer vision (unofficially). They abandoned the idea of navigating with CV at some point, I believe. They seemed very capable, which means that it is likely a very difficult task :) 

5. Did anyone try collision detection and avoidance?  If so, how'd it work out?

I don't think there was much (successful) collision detection. JRover Maxx seemed to be able to just ignore collisions...big beast! I just tried to go faster. 

Nathan

[Ted Meyers]

1. Certainly added to the challenge, but only in the barrels area (very tight to avoid!).  A larger bot like Maxx had no issues here =)

2. Non-GPS is a huge advantage (even more with the bonus) in my opinion.  I believe that Auton used a GPS.

3. The BNO-055 worked remarkably well, even correcting for some solid collisions.

4. No vision here, too many other things to get working.

5. I had added some collision detection/recovery using accelerometer and odometer data, but ended up removing it, I did not have a quadrature wheel encoder this year and my position got too far off when backing up (even when trying to use throttle to determine reverse/forward).   This would not be worth taking the code out, but I had a very real concern of false positives (definitely going to work this out for next year).  I ran out of time and wasn't able to get my IR rangers working for collision avoidance.

My biggest problem this year was trying to drive inside the barrels and miss the stanchions (there was something like 6 feet between them).  I had the precision, but could only get my initial starting angle within about +/- 1 degree which I figure is about +/- 5 feet at that distance.  Once past the barrels it was much easier.  I figure the new track is as narrow as 18 feet in several places and there is a lot less room around the corners than before.  Also, the starting line is very tightly packed with 5 bots starting at a time.  I feel sorry for the little bot that Daisy drove over, but he did try to t-bone Daisy.

Ted

On Monday, June 22, 2015 at 4:27:03 PM UTC-6, wholder wrote:

[Thomas Roell]

Ad (1) The narrower track was more unforgiving for the smaller bots. I did like that a lot.

Ad (2) GPS SUCKED. We ran 2 bots, one with a UBLOX7 GPS, and one with a UBLOX8 GPS/GLONASS setup. The run to run differences were in the order of 4 meters (in words "four"). And this with 9 satellites used and DGPS (WAAS) on the GPS setup. I have not gone throu the GPS/GLONASS data logs yet, but looking at one of them briefly, same issue. I have never ever seen this before, and I did test on a few parking lots around noon close to industrial buildings (well, easy if you have a daughter playing indoor volleyball).

Ad (5) After realizing the GPS issue, I added quickly some collision detection. Brute force, and simple. If the GPS reported low speed, but the RPM sensor reported high speed, it was classified as a crash. Since the course is a right turn course, the rover would back up in reverse to the left, and then forward to the right before handing control back to the normal guidance system. That worked well as you can see in the AVC live stream and Killer Kitty's 3rd run. However another software bug that was triggered by the barrel collision a few moments before, that cure was temporarily, and just delayed the death in the fence. It seems that simple pure pursuit does not work post collision, irregradless of whether it's just a deflection of a barrel of a full stop in the fence.

Looking back, I found it very frustrating, as mainly little software gremlins that never raised their head in testing really killed us in a big way. Also the choice of rover based made a difference. I had planned to jump the discumbudingens. But the short wheel based 1/18 buggy was too instable (that was part of my gamble going for GPS).

As to whether no-GPS or GPS is better, I still don't know. Except for JRoverMAXX (tm), none of the dead reckoning rovers could recover well from a crash of deflection.

- Thomas

PS: I am short on visual data analysis. Got a busted knee from a Skiing accident, so I let my kiddos run around the course. Their analysis skills do need to be improved quite a bit ;-)

On Monday, June 22, 2015 at 4:27:03 PM UTC-6, wholder wrote:

[Ted Meyers]

Ad (1) The narrower track was more unforgiving for the smaller bots. I did like that a lot.

I will ++ that point!

Ad (5) After realizing the GPS issue, I added quickly some collision detection. Brute force, and simple. If the GPS reported low speed, but the RPM sensor reported high speed, it was classified as a crash. Since the course is a right turn course, the rover would back up in reverse to the left, and then forward to the right before handing control back to the normal guidance system. That worked well as you can see in the AVC live stream and Killer Kitty's 3rd run.

I noticed that several of the GPS bots doing something like this.  It is a bit more challenging for non-GPS, for sure. 

 

 It seems that simple pure pursuit does not work post collision, irregradless of whether it's just a deflection of a barrel of a full stop in the fence.

JRover Maxx and Daisy Rover both had incidents where they deflected off of obstacles and continued on; don't know about Maxx, but Daisy was using pure pursuit.

 

As to whether no-GPS or GPS is better, I still don't know. Except for JRoverMAXX (tm), none of the dead reckoning rovers could recover well from a crash of deflection.

Yes, my bot could only recover from a deflection, not a head on unless it managed to knock a barrel out of the way (did happen once in practice).  Even then, the heading and/or wheel encoders had a good chance of being off.

BTW, Thomas, your daughter is quoted in this mornings Denver Post article on Sparkfun.

Ted 

[Thomas Roell]

Comments embedded.

- Thomas

On Tue, Jun 23, 2015 at 9:08 AM, Ted Meyers <ted.m...@gmail.com> wrote:

 It seems that simple pure pursuit does not work post collision, irregradless of whether it's just a deflection of a barrel of a full stop in the fence.

JRover Maxx and Daisy Rover both had incidents where they deflected off of obstacles and continued on; don't know about Maxx, but Daisy was using pure pursuit.

On the 3rd run with Killer Kitty the rover got effectively deflected to the left around a barrel. The next target waypoint was right of the barrel (the course was set to pass the barrel on the right). That put the pure pursuit into a state where the position was beyond it's lookahead and outside the radius to find the next subsequent waypoint. The result was simply that the rover would follow the line between the last waypoint and the current waypoint, which lead into the fence. IMHO pure pursuit needs to be upgraded to handle the case where the rover leaves the lookahead radius, and/or passes the target waypoint. Looking at other folks source code, the same issue is present in most snippets I looked at. 

As to whether no-GPS or GPS is better, I still don't know. Except for JRoverMAXX (tm), none of the dead reckoning rovers could recover well from a crash of deflection.

Yes, my bot could only recover from a deflection, not a head on unless it managed to knock a barrel out of the way (did happen once in practice).  Even then, the heading and/or wheel encoders had a good chance of being off.

What did you use as encoders and where ? I opted for an RPM sensor off the motor. It was a bad choice as there was a lot of variance between readings. Perhaps a Hall sensor off the main drive shaft is better (but offers less pulses per rotation. Still need to find an optical encoder that I could drive off the motor. 

BTW, Thomas, your daughter is quoted in this mornings Denver Post article on Sparkfun.

Cool. Need to get the paper version then.

 

Ted 

--
You received this message because you are subscribed to a topic in the Google Groups "diyrovers" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/diyrovers/W3iqBv1KIKw/unsubscribe.
To unsubscribe from this group and all its topics, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CADnZWKM_SAK0UaNoAHTCq1Y1_0XaNJBOwP%3DqK%3DcCgwYS9tDTRQ%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

[Jon Watte]

Looking back, I found it very frustrating, as mainly little software gremlins that never raised their head in testing really killed us in a big way.

That's why there are competitions, no? :-)

Perhaps a Hall sensor off the main drive shaft is better (but offers less pulses per rotation. Still need to find an optical encoder that I could drive off the motor.

If you have a sensored brushless, you can use that. I saw some team glue a black/white cylinder to their drive shaft and use a photodiode at RoboGames a while back; that seemed to work, too.

cI still think vision *could* do it all. Having the Google team not manage to do it, though, shows how hard it is.

I bet a camera pointed down, with strong illumination, and short shutter time, could do pretty good odometry and direction based on pattern matching between frames if you had enough computing oomph. The RPi camera goes to 90 Hz in 640x480, and the GPU might allow you to do some number of cross convolutions to find "best match" deltas.

Sincerely,

jw

Sincerely,

Jon Watte


--
"I find that the harder I work, the more luck I seem to have." -- Thomas Jefferson

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.

To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CAEhxOcn35BxbL2VB1J6CogpeGEbPB19u%2BN%3D04e9FpqgvNoxHMA%40mail.gmail.com.

[Wayne Holder]

> Ad (2) GPS SUCKED.

In looking over the photos and videos, I'm fairly convinced that the chain length fence and its proximity to the course may have been a factor in GPS reception, as it's a nasty source of multi-path reception.  And, the proximity to the large, two story Sparkfun building would be a factor, too.  Frankly, last years location at the lake was a better venue for GPS-based bots.  It's almost ironic that Sparkfun chose to penalize the use of GPS in this new venue.

Personally, I've always preferred a GPS-based approach, as the AVC is my poor man's version of the DARPA challenge and I'm interested in a general purpose, ground-based navigation solution.  But, if future AVCs are going to now be at Sparkfun HQ, I may have to abandon that goal, as I think it's rather GPS hostile.

Wayne

--

You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/b7d80d82-33c2-478d-b867-bdbe0895626c%40googlegroups.com.

[Ted Meyers]

The fence theory makes a lot of sense.  The fence was quite high and all the way around the course with no way to get far from it.

A GPS based challenge would be a fun and interesting contest; I'm thinking of something on a much larger course with a fair amount of cross country and off-road legs, maybe in a large park.  I'd be willing to organize a contest like that in Colorado, if anyone is interested.

Ted

[Thomas Roell]

Ted,

that is a stellar idea. I was toying with letting Marie's Killer Kitty (not mine anymore it seems) go around Washington Park (Denver). That's a 2.4 mile course ;-) The obstacles are randomly moving (bikers ;-))

A big parking lot (empty) with perhaps a bike-trail next to it might be a good idea.

- Thomas

--
You received this message because you are subscribed to a topic in the Google Groups "diyrovers" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/diyrovers/W3iqBv1KIKw/unsubscribe.
To unsubscribe from this group and all its topics, send an email to diyrovers+...@googlegroups.com.

To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CADnZWKOXJZnRs9M8JSgmUJMTm2vS6eHB2TCazcLHDbS20de3Mw%40mail.gmail.com.

[Ted Meyers]

Okay, cool, I'm going to start working on this.  Maybe a bit shorter than 2 miles (like 1/2 to 1 mile), but something like that.  Need to find a good location that isn't too populated.  Maybe some time this fall.

Ted

--

You received this message because you are subscribed to the Google Groups "diyrovers" group.

To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.

To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CAEhxOckQcSDWuc7Ss78vix6CP1F9QYk6mX3ESoZUGzqdKb7Cwg%40mail.gmail.com.

[Thomas Roell]

A mall that is closed for renovation might be a good target.

- Thomas

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CADnZWKN36n3dbkHJfkzFbzmfOUhUb%3DoF5QcfSr24XGVtqksO1A%40mail.gmail.com.

[Thomas Roell]

Or a highschool during the fall-break.

[Michael Stephens]

honestly a track and field event... one competition is the track (you could even line follow) and another is the field where there are way points. It would be fun to have like 3 qual events throughout the year and maybe a big final at the end. That way the robot juices keep flowing.

Michael Stephens

engrst...@gmail.com

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CAEhxOc%3DVZOTj%2Bv8XG8EuC9iBsUstL3mVje5_PZotTF8YdoMSnw%40mail.gmail.com.

[Thomas Roell]

The field is kind of the tricky part for us who have smaller rovers (1/18 scale). I tried to run ours in the park over grass, and that just will not work.

- Thomas

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CAJ4bJXQN1Hg6fpRWvvuO3aWHZ4YrjVCEBN0z6X%2Bx1ZR%2Bd6Utmw%40mail.gmail.com.

[Wayne Holder]

A big, outdoor course would be a wonderful venue for an autonomous race.  Heck, even a smaller course in a GPS-friendly environment would be fine, too.  Ideally, it would be great to have a course that provided opportunities to use different strategies and technologies for different portions of the course.  One thing that has always bugged me about the Sparkfun course is that some of the obstacles, such as the hoop, are hard to deal with other than via precise navigation, or dumb luck.  I'd like to see obstacles labelled with unambiguous glyphs so that vision systems might have a reasonable chance at locating and tracking them.  Likewise, I'd place similar markers on other parts of the course so that someone might have a chance to use an all vision-based solution.

Wayne

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CADnZWKOXJZnRs9M8JSgmUJMTm2vS6eHB2TCazcLHDbS20de3Mw%40mail.gmail.com.

[Jon Watte]

A GPS based challenge would be a fun and interesting contest; I'm thinking of something on a much larger course with a fair amount of cross country and off-road legs, maybe in a large park.  I'd be willing to organize a contest like that in Colorado, if anyone is interested.

That's what RoboMagellan is, pretty much! (Not THAT much longer, but still more.)

Sincerely,

jw

Sincerely,

Jon Watte


--
"I find that the harder I work, the more luck I seem to have." -- Thomas Jefferson

On Wed, Jun 24, 2015 at 10:10 AM, Ted Meyers <ted.m...@gmail.com> wrote:

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CADnZWKOXJZnRs9M8JSgmUJMTm2vS6eHB2TCazcLHDbS20de3Mw%40mail.gmail.com.

[Ted Meyers]

Okay it is similar, but I was thinking more along the lines of a race course, instead of trying to find cones.  But I'm open to changing it up.

Ted

[Minuteman]

This does sound similar to robo magellan, which I would like to compete in. However I wouldn't advocate Sparkfun doing robomagellan, only because it would become a repeat of other events--not unique.

I have played around with GPS's before, but never used one for AVC. I've thought a lot about how I would implement it, if needed, and I would like an excuse to do that. So, I would also welcome a GPS-based challenge, as long as it would be something more than what an out-of-box ardupilot could handle. For some reason I was a little put-off by all the 3DR equipment I saw at this year's competition. Thankfully, they didn't win, for the most part.  I think the aerial competition had already gone that way, with a proliferation of off-the-shelf quad copters and such. Not much innovation there.

I think I will pull out my Emaxx and start adding some guidance to it. I want for it to autonomously rove my neighborhood. That would be fun.

[Thomas Roell]

Last year I had suggested to Robert C as to whether there couldn't be constructor points. Just using the 3DR hardware and software takes really the fun out of the competition.

- Thomas

--
You received this message because you are subscribed to a topic in the Google Groups "diyrovers" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/diyrovers/W3iqBv1KIKw/unsubscribe.
To unsubscribe from this group and all its topics, send an email to diyrovers+...@googlegroups.com.

To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/b6cc5e8f-05a1-4463-92c4-f9165209b78f%40googlegroups.com.

[Wayne Holder]

I don't mind the 3DR-type entrants, as I think that a good entry point for new entrants and kids.  I was able to teach my daughter the basics of waypoint navigation and this made the challenge more comprehensible to her.  But, I would prefer a race course where waypoint-based navigation was just a starting point.  Ideally, a challenging course should require a bot to handle collision detection and avoidance, and/or recovery using relatively inexpensive technology, such as SONAR, LIDAR or vision.   The AVC barrels work OK, but I'd have more than one segment have obstacles, not just one short segment.  Also, bonus challenges such as the Hoop should be made in a way that renders them more visible to SONAR, LIDAR and vision techniques.  Just using a thicker pipe, such as 3 inch PVC drain pipe painted a distinctive color would make this a more doable and interesting challenge for ranging sensors, as would adding a unique glyph to the top center of the arch for vision systems to spot and home in on.  Here are some other ideas that might make for an interesting race course:

1. The discombobulator was an interesting idea, but wold be more interesting if it rotated at slower rate and perhaps was a 90 degree turn rather than a straight across obstacle.  This would defeat jumping, but still make for a very interesting obstacle.  It might even be programmed to turn on and off and reverse direction at random intervals.

2. A section of track (perhaps a shortcut) with high, detectable walls that started out 8 feet wide, but narrowed down to perhaps 2 feet at the exit.  This could be tackled with ranging sensors, or perhaps vision if the exit area contained something visually unique to track.

3. Borrowing from miniature golf, perhaps a gate that opens and closes at timed intervals.  Again, made such that vision and ranging sensors could determine its state.

4. This is a bit farfetched, but a loop would be a very interesting challenge (see https://www.youtube.com/watch?v=nUEPC66KplI for some attempts using RC cars), as it would require speed and precision to pull off.  To make it more doable, perhaps paint a high contrast stripe down the center so a high speed line-following algorithm could be used to run it.

5. Likewise, rather than a simple ramp, perhaps a pair of back-toback ramps (one up, one down) that crosses a chasm, or obstacles of some sort.  Kind of an Evel Knievel-type, jump the buses thing.  Would add an element of danger, as does the loop.

Ideally, the course would be arranged so that one could complete the longest version of the course using simple waypoint navigation as the entry-level challenge for new guys.  But, coring any serous points would require building a bot that can handle the more advanced challenges like I described above.

This might all be just to difficult and time consuming to even try to even consider, but it could be a lot of fun, too.

Wayne

[Ted Meyers]

Wayne,

All of those obstacles would be really cool and fun, but the downside is the difficulty in building them.  Ideally we want obstacles that are easy to construct, so that everyone can test with them.  It is much more difficult to build a successful robot if it cannot be tested.

[Rich B (Roadrunner)]

Discombobulator - I ran roadrunner and wile e. coyote both in practice and in competition. I found roadrunner (1/10th scale) performed flawlessly on the course and in jumping the discombobulator once its course was dialed in. It had 3 flawless runs in the competition.

wilie e. coyote (1/16th scale) was a different matter. It would run the course just fine, but wouldn't consistently launch over the discombobulator. I think the car's shorter length and higher speed needed to completely clear the obstacle resulted in it launching at different angles and that resulted in inconsistent results. During heat 1, it slammed into one of the discombobulator's walls and failed to complete the course. In between heat 1 and 2, I increased the speed and changed some waypoints. The last practice test resulted in a high speed crash with the fence. After hot gluing the front suspension back together, heat 2 resulted in the car the just spinning around in circles...turns out the high speed crash killed my gyro. After replacing the gyro, but doing no testing, heat 3 resulted in a low speed crash with a different fence. Lesson learned...smaller cars should avoid the discombobulator.

1. What new challenges emerged from the narrower, longer distance track.

Not much. The track is similar length to previous years, so I didn't notice any differences.

2. How well did the GPS vs non GPS crowds do?

Covered already. Non-GPS won the day IMO.

3. How did the new sensors (i.e. Bosch BNO055) perform in combat conditions.

Didn't use it.

4. Did anyone make effective use of vision?

Covered already.

5. Did anyone try collision detection and avoidance?  If so, how'd it work out?

Haven't used it in 5 years and I don't plan on starting. There are other was to avoid crashes that are easier than using sensors.

[Minuteman]

I like to think of ways that the competition could be changed with innovative new events. Then I am reminded of the fact that the majority of the entrants don't even get off the starting line, year after year. It reminds me that robotics is hard, especially for people just starting out.

I also compete in mini sumo robotics events. That is another competition with relatively simple objectives, but even that task has endless complexities that makes it fun and interesting.

Every small tasks has a potential to be very complicated to solve. I think providing too many tasks would mean even fewer successful bots.

[Ted Meyers]

Where do you compete in mini sumo?  I might be interested.

[Ted Meyers]

In regards to robots getting off the starting line, I've seen these rules for building robots:

Rule 1:  Get something to work.

Rule 2:  Get rule 1 to work really, really well.

Rule 3.  (optional) Get something else to work.

Rule 4:  See rule 2.

My goal this year was to follow these rules, and I had much more success than ever before.

Ted

On Saturday, June 27, 2015 at 8:10:05 AM UTC-6, Minuteman wrote:

[Minuteman]

We have had a few competitions locally (in Los Alamos) and I have a friend here that has very competitive bots, so we help each other. We have competed in the UNI send-in competition for the past couple of years: http://www.narobotics.org/

This year we also went out to the robo games in California. The competition is very tough from the international competitors (not so much from the US ones).

Here are a couple of videos of my bot: 

https://youtu.be/WV2rmmB4uTY

https://youtu.be/228yAyQGVqo

Rich (roadrunner) lives in Denver, and will be working on getting some local competitions going in mini-sumo. I would love to see Sparkfun add that to AVC. It is the perfect scope of project for beginner through advanced robotics. Cheap, simple hardware, lots of room for innovation.

Nathan

[Minuteman]

Yup, that basically what I go for as well.

[ColinC]

Wayne, I hear you and I feel your pain, but, really, if you feel that that venu is gps hostile, one has to wonder what would be considered gps friendly...maybe the middle of the great salt lake.

Trees or spectators lining the course could be detrimental. Or just about anything. I think the gps environment tested should be somewhere near typical, or it's just not useful.

But good work, in any case :-)
Colin

[Wayne Holder]

My theory about the chain length fence is still just a theory, even though I can point to some articles mentioning the effect, so I plan to investigate that in more detail when time permits.  But, last years venue was actually pretty good, IMO.  It's actually pretty hard to use GPS on a rover given that being too close to the ground can also cause multi-path issues and with the limited accuracy of consumer GPS modules.  I don't expect a perfect GPS environment, just one where some considerations are taken, such as avoiding, or height limiting the use of metallic fencing close to the track and keeping away from tall buildings (the original course at the old Sparkfun Bldg was an absolute nightmare for GPS.)

At the same time, I'd like to see the course made a little more interesting and challenging for everyone by getting away from a simple square shape, which seems to be becoming fairly trivial for the dead reckoning bots to handle.  Even the rounded rectangle shape of classic  racetracks like Indy would be an improvement over a square-edged rectangle.  Or, how about throwing in some with diagonal runs and switchbacks to keep those gyros on their best behavior.  But, I'm just musing.  Sparkfun will do what they do, so It'll be up to me to adapt.  I may have to give up on my "general purpose", GPS-based solution.

Wayne

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/d2ccbb7c-8ac8-41c2-a0ac-06b6e51b5c98%40googlegroups.com.

[Minuteman]

Actually, I believe dead reckoning would shine even more with a complex path, especially if the turns were tight, or close together.  It would have trouble with paths that aren't pre-defined. But I think GPS would also suffer in that case as well. 

Having said that, I do plan to try GPS this year (probably not for AVC). I will treat it as a long-term correction to the dead reckoning so that I can have my bot drive around the neighborhood.

[Ted Meyers]

Notice the caption on the GPS clock in the Sparkfun blog today (6/30/15) reads: "GPS may not be ideal for getting your ‘bot around the AVC track, but for keeping time it’s just fine."

[Ted Meyers]

Hah!  I just found the first major bug in my code for this AVC: I was originally running the BNO-055 in NDOF mode, but switched to using IMU mode (no mag), however I added the code to set the mode to IMU _before_ it was setting the mode to NDOF and I failed to remove that line, so it was running in NDOF (sort of).  Actually it wasn't really, because to run in NDOF the magnetometers need to be calibrated by spinning them and I wasn't spinning the bot (on purpose).  But this could explain some odd and previously unexplained behavior that had really freaked me out in testing a few days before the AVC (and at least once during).  Uh, it really shouldn't be this hard.

Ted

On Monday, June 22, 2015 at 4:27:03 PM UTC-6, wholder wrote:

So, what with having missed this year and the lack of a video feed for much of the early rounds, I'm curious to get a 911 on the new track layout and how everyone's "best laid plans" worked out in practice.  Specifically:

1. What new challenges emerged from the narrower, longer distance track.

2. How well did the GPS vs non GPS crowds do?

3. How did the new sensors (i.e. Bosch BNO055) perform in combat conditions.

4. Did anyone make effective use of vision?

5. Did anyone try collision detection and avoidance?  If so, how'd it work out?

Wayne

[ColinC]

As for changing up the course, the only thing I would like to see added is one decent, honest-to-god, hill. I'm thinking maybe upper and lower level parking lots joined by a short road.

Cheers,
Colin

[Jon Watte]

Turn the discombobulator into a see saw instead of a glorified jump ramp?
Or just extend the on ramp to be much longer. And get some gearing for the engine...

Jw

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/fbed3cd8-b73b-429b-b0f5-e12d19bf44b8%40googlegroups.com.