Autonomous Airborne Docking for Refueling or Recharging

This video shows some recent flight test results from my PhD research on autonomous UAV docking, that was conducted with the Australian Centre for Field Robotics (ACFR) at The University of Sydney. To the best of our knowledge, these experiments are the first time two UAVs have docked in mid-air.

Some of my other videos provide more specifics on the hardware/software, but I'm using my own autopilot, a Beaglebone Black as the formation flight computer and an Odroid U3 to do the onboard vision processing. All the guidance, navigation and control algorithms on all the hardware is developed in Simulink and automatically converted to C/C++.

The IR vision aspect is probably of interest to some of the guys doing auto-landing with IR beacons. We use a C920 with narrow bandpass filter and 940nm LED markers without modulation. It works in all lighting conditions except when looking directly at the sun where the high number of false positives are difficult to distinguish from the actual markers, but there are a number of ways to handle this infrequent situation.

There are some nice airborne shots so watch in HD!

Views: 3739

Comment by Jethro Hazelhurst on April 19, 2015 at 10:56pm

This is remarkable work, a major step forward for autonomous UAV tech, thank you for sharing with the community!


MR60
Comment by Hugues on April 19, 2015 at 11:09pm

20/20 on your PHD with highest mention.

Comment by JB on April 20, 2015 at 12:58am

Awesome stuff Dan! Definitely need this for the next UAV challenge!

I can just imagine retracting a 250 quad into a big wing with it! ;-)

How did you manage to maintain distance and contact with the drogue? Did it have some sort of hook/clamping mechanism or did you do it with just follower plane throttle control? If so even more impressive! :-)

Comment by Thomas Stone on April 20, 2015 at 7:31am

@Dan

Thanks for the reply! I sent you an e-mail. 

That makes perfect sense. I imagine Lidar Lite will be a good low-cost solution for distance-to-target sensing for our precision landings. 

The modulation works well with our sensor's processing capabilities. I agree, if it added a lot of complication, I would probably take another route. I think one of the cool things we can do is acheive 'multiple layers of reliability' against false detections. ... Optical filtering -> Modulation -> Pattern Detection ... Exciting stuff! Please stay in touch.

Best,

Thomas

Comment by Armin Strobel on April 20, 2015 at 2:16pm

@ Dan

Sadly that concept study is not a project. If you know somebody who wants to make a project from that study or someone who does something similar, then let me know.

Thanks for sharing your experience. Did you ever try to dock when there where strong thermal activities? Would be nice to see how robust it is, or how far you can push it. I am a (manned) glider pilot. being towed behind an aircraft in thermals can be very rough. Regarding the wake, down wash and the propeller wash, should be relative easy to avoid if the drogue is light and the cable long enough to stay above those effects. At least that was the idea in the concept that you dock outside those disturbances and then get pulled inside trough those disturbances.

Do you use the makers on the pulling airplane as additional information in your controller or just the markers on the drogue?

Btw. just realized that we have a mutual friend in Munich.


@ Thomas

I am using the lidar lite for my precision landing systems. I was initially worried because of inteferences, but it works like a charm (so far). I use one ir-led (940nm/3W).

Comment by Thomas Stone on April 20, 2015 at 2:59pm

@ Armin

Excellent! That is a good observation. I suspect the reason is that Lidar Lite is operating in a completely different band of wavelengths.... We are also around 940nm, so everything should be good. :)  (will shoot you a message soon; I have to run)

Comment by Justice Bentz on April 20, 2015 at 3:43pm

I'd imagine this could be used for fixed wing landing on moving vehicle.

Comment by Dan Wilson on April 20, 2015 at 6:23pm

@ JB

I suspect you would go through a lot of platforms in the development phase while attempting to retract a quad! 

The drogue has a few magnets on the inside and the follower has a metal nose cone with magnetometer inside, but this system is primarily used to detect docking. The follower tries to punch through the drogue slightly, then target a position forward and above where the drogue was, if docking was detected. This ensures some slack on the tether during station keeping. One of the challenges is accounting for the repulsive force when docking occurs. This is because the drogue is ~40% of the leader's total drag, which is instantaneously removed when docked. At the same time, the follower gains extra drag and mass on the nose so the controllers need to adjust fairly quickly.

@ Armin

That's a shame, I'm sure someone would be interested in giving it a shot, however it may be easier to start with more similarly sized aircraft and attempting to capture one with the other.

We had some very strong thermal activity in summer and as you know, it made things much more difficult. I believe we managed to dock a few times but it's much less repeatable. The same is true in strong winds since we fly in quite a hilly area which creates a lot of turbulence. Larger aircraft would help with this.

Yes we use the leader's markers to estimate the leader-follower relative state at all times, then the drogue markers to estimate the drogue's state, relative to the leader. Before attempting docking, we did a lot of experiments with no drogue and just close formation flight, down to 4 m separation. I'll post some of this videos at some point showing formation flying close to the ground (15 m), past trees and in high winds (50 km/h). 

Marc mentioned he knew you!

Comment by ivan ravinovitch on April 23, 2015 at 5:54am

This is extremely amazing!

Comment by Hasufel on April 27, 2015 at 1:09pm

Very nice work Dan. Keep it up. X-47B style! https://www.youtube.com/watch?v=rIRwsOG_AYQ

Comment

You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service