Drone violent sprinting

[Tim Robertson]

I've recently been integrating the drone onto a larger hex system with a pixhawk running the latest firmware.

During a test flight, the loss of RTK lock was followed by a very quick sprint to another location followed by an equally rapid deceleration.

We've determined the action occurred after RTK lock was lost, we also noted a distinct bias in GPS coordinates from the 3DR gps and the RTK GPS system.

Three questions:

• 1) Is an incorrect RTK Survey position (+/- a few meters) likely to cause this issues?
• 2) Other forum questions imply some amount of GPS jumps during flight behavior, is this a problem that is resolvable by further attempts at turning the RTK parameters to fit the drone?
• 3) Can you provide any insight on ways of maintaining Rtk lock while in flight?
  

[Swift Navigation (Josh Kretzmer)]

Tim,

It sounds like what you are experiencing is the drone switching back and forth between the 3DR GPS and the Piksi. The type of behavior you are seeing is the drone attempting to correct its position - and due to discrepancies in the position solutions provided by each GPS this is manifesting itself as the drone rapidly moving to a new location.

• 1) Is an incorrect RTK Survey position (+/- a few meters) likely to cause this issues?

• Yes. If your base station coordinates are not accurate, you can expect the position of the drone to change significantly when switching between the 3DR GPS and Piksi.

• 2) Other forum questions imply some amount of GPS jumps during flight behavior, is this a problem that is resolvable by further attempts at turning the RTK parameters to fit the drone?

• If you are able to maintain RTK fix during flight, you should not see this behavior.

• 3) Can you provide any insight on ways of maintaining Rtk lock while in flight? 

• The most critical variables in maintaining RTK lock in flight are going to be your communications link and your sky view. Since you are using Pixhawk as your platform, you may find some helpful information in our Pixhawk Integration Guide. What is the distance between your base station and drone?

Thanks,

-Josh

[Tim Robertson]

Thanks for the reply Josh,

-Our base station and drone were ~20m from each other at the time we saw the sprinting behavior, we are using the 3dr radios.

As a follow up

1) Is there a way to get a 'good' survey location without needing a surveyor/cartographic marker? Perhaps their is a better GIS map that is publicly available?

2) Does paint mess up the antenna behavior? I noticed the puck antennas have a 'do not paint' marker on them. Would this imply a cover/canopy is out of the question as well?

3) What about time of day dependence? There are some satellite trackers which suggest the dilution of precision changes with respect to the location of sats over a location (ie. different times are better for gps precision). Does this apply to the RTK system as well?

[rai gohalwar]

Hi Tim,

1. The cheapest/best way I have found my base location is using a Ublox m8 GPS. The position I get after 20-30 mins of run time is usually under a meter of accuracy.

2. For sure. With paint or canopies, it depends on the material. For Piksi's antenna, I would recommend you not cover it with anything. Refer to the link Josh Kretzmer shared with you for images of ideal UAV hardware setup and recommendations.

3. Most definitely. Since Piksi is a GPS only L1 GNSS unit, time and location can really make a big difference. Good and bad. Trimble has a tool which can track sattelites and map their trajectory. 

There is however a fundamental question you must answer. Must you use RTK for navigation on your UAV? From my work with RTK systems on UAVs, I have found that there is little to no performance advantage. 

Best

Rai Gohalwar

[Boris Lipchin]

Raj your comment about performance advantage is interesting, and I wanted to dive deeper into that. How do you define performance in that there is no difference between flying with GPS RTK versus a standard GPS unit?

Boris

[rai gohalwar]

Boris,

What I meant to say is that adding a RTK GPS on your UAV will not significantly help you follow your planned path or hover at the exact spot. Its not the gps, but the control loops in APM. They are not optimized for really accurate positioning data. Not to say that RTK doesn't help at all. During my research, I found that a UAV's ability to follow a specific path or hover exactly at one spot was effected by not just the accuracy of your GPS data, but also the PID settings, wing/gust conditions, vibrations on the UAV, and lastly CG of the machine. 

Best

Rai Gohalwar

PS: my name is Rai, not Raj. 

[Boris Lipchin]

Oops, apologies about the name Rai! Misread the letter. =)

That is a fair statement, the quality of the navigation is not necessarily the bottle neck in the system and the bandwidth of the actuator or the control loop might actually be what is limiting performance. My hope is that one of two things is true: either the navigation /is/ the bottle neck and I will see a direct improvement or a better navigation might give me more headroom in terms of margins to then tighten up the PID gains and improve performance there. If the PID loop for this plant is already perfectly optimized given the physical constraints of the system (moment of inertia of the plant and the props, the CG, etc.) given the bandwidth of my motors and ESC's then yeah I think I'd have to move more drastic changes.

This is at least my thought process with regards to performance gains due to RTK. Personally I need it for hover precision in general and less so in terms of path following.

[rai gohalwar]

Boris,

No problem at all. When someone calls me Raj, I feel like I am in a bollywood movie. haha

Anyhow, I agree with you. You will definitely see some performance difference with the addition of an RTK system. What kind of precision/accuracy are you aiming for  ? 

Best

Rai Gohalwar

[Boris Lipchin]

For me it's a the more the better sort of thing, up and to the right, etc. Comes down to a minimum precision requirement given some wind speed. A 10cm error is roughly my design point. At what wind condition can I effectively achieve this is going to be, like you said, a factor of a bunch of stuff over some of which I have little direct control over.

B

[rai gohalwar]

Boris, 

1.5 years ago when I was doing research on this very topic, my team had a hard rule of not testing in gusts over 15 k/hr. Since we were trying to hover within 15cm radius of a marked peg on the ground, we optimized with that number. We saw our UAV's performance being effected dramatically with winds above that. Now, these numbers may be different depending on the UAV. We were flying the old 3DR x8 drone and it was pretty well tuned and optimized. 

Best

Rai Gohalwar  

PS: Just to spin some more thoughts in your mind, I soon realized that making the interrupt time shorter for the Motor-PWM loop in APM, and changing some EKF values on the IMU and GPS filters, along with new ESC technology (OneShot) there could be a potential with RTK for positioning. This is a lot of work though and I would go broke and homeless trying to do it. lol