microsoft/frigatebird

AI for autonomous soaring

Project Frigatebird

Project Frigatebird is aimed at enabling small fixed-wing UAVs to travel long distances by soaring -- taking advantage of rising air regions the way human sailplane pilots and bird species like frigatebirds do -- using AI planning and learning techniques.

Currenty, this repository contains an implementation of POMDSoar, an algorithm that allows a sailplane UAV to detect and gain altitude in thermals, "columns" of rising air that occur over certain regions of Earth's surface. This approach, based on solving a Partially Observable Markov Decision Process, is described in

I. Guilliard, R. Rogahn, J. Piavis, A. Kolobov. "Autonomous Thermalling as a Partially Observable Markov Decision Process." Robotics: Science and Systems conference, 2018. https://arxiv.org/pdf/1805.09875.pdf

We implemented POMDSoar as a soaring controller integrated into a fork of ArduPlane, the fixed-wing flavor of ArduPilot open-source drone autopilot suite (original ArduPlane code, wiki). The code in the Frigatebird repository allows building this modified version of ArduPlane. Please see BUILD.md for building instructions.

The DATA subdirectory contains ArduPlane parameter files and an example geofenced waypoint course for running POMDSoar on a Radian Pro sUAV as in the experiments in the above paper.

Developer Information

• Github repository: https://github.com/Microsoft/Frigatebird

• The main site (http://ardupilot.org), developer wiki (http://dev.ardupilot.org), and discussion page (http://discuss.ardupilot.org) of the parent project, ArduPilot, is useful for understanding how to modify Frigatebird.

License

Project Frigatebird is licensed under the GNU General Public License, version 3

Maintainers

• Iain Guilliard, main developer of Frigatebird's POMDSoar thermalling controller
• Andrey Kolobov, project lead

Acknowledgements

We would like to say a big "thank you!" to the developer community of ArduPilot. Their code and advice has greatly helped in starting Project Frigatebird.

frigatebird's Issues

build link in readme is bad

This repo is missing important files

There are important files that Microsoft projects should all have that are not present in this repository. A pull request has been opened to add the missing file(s). When the pr is merged this issue will be closed automatically.

Microsoft teams can learn more about this effort and share feedback within the open source guidance available internally.

Merge this pull request

waf based builds

What can I do to help flush out waf build issues? Running Ubuntu / WSL on Windows 10 laptop, and everything of late in ArduPlane / ArduCopter is waf based. It's a bit of a step backwards in this ecosystem to have to use make. Also, hoping to fly Frigatebird code on latest ArduPlane master running on ChibiOS HAL on Matek F405-Wing, on Radian Pro. So a highly relevant config, IMHO. Would really like to avoid having to maintain parallel and different build environments/flows.

Kelly

PR the SilentWings SITL backend for ArduPilot master

hi guys, are you interested in bringing the SilentWings SITL backend into ardupilot master? I'd be happy to look at a PR for that, or I could just copy what you've done.
I haven't tried silent wings myself yet, but I just saw the video from Marco. Looks really nice.

Code is missing attributation to Stack Overflow Q&A

The code at https://github.com/microsoft/Frigatebird/blob/master/libraries/AP_HAL_F4Light/hardware/sd/Sd2Card.cpp#L82-L138 has been lifted from my answer at stack overflow.

https://stackoverflow.com/questions/21593692/convert-unix-timestamp-to-date-without-system-libs

user contributions licensed under cc by-sa 4.0 with attribution required.

https://creativecommons.org/licenses/by-sa/4.0/

Queries

Hi there,

I'm working on a simplified version to act as a starting point to getting the POMDP into the main codebase. For the following settings:

• POMDP_NORM: Normalises the covariance trace (= 1 in param file)
• POMDP_EXT: Extends action duration in lift-exploiting mode (= 3 in param file)
• POMDP_PLAN: Enable deterministic trajectory planning mode (= 0 in param file)

Do you find that a particular configuration gives the best performance?

I'm also wondering about the relationship between N, HORI and K (samples per action trajectory, planning horizon duration and samples per second) - it seems like only two should be needed?

Lastly, the scheduling - there's several entries in the scheduler table calling soaring_policy_computation at 50Hz. This is unlikely to be acceptable - is there another way to achieve this? Would calling it at a higher rate have the same effect?

Cheers!

Segfault when running master on SITL

When running in SITL I get the following segfault:

Program received signal SIGSEGV, Segmentation fault. 0x00000000004ea483 in VarioSavitzkyGolayFilter::prediction ( this=0x7d5c74 <plane+1236>, dt=2.85445094, ekf_buffer=0x7d64b0 <plane+3344>, ekf_buffer_size=300, ekf_buffer_ptr=0, Edot=0x7d6488 <plane+3304>, Edotdot=0x7d648c <plane+3308>) at ../../libraries/AP_Soaring/VarioSavitzkyGolayFilter.cpp:22 22 p1 += SG_FILTER_C[0][i] * ekf_buffer[k][0];

My gcc version is 5.4.0 20160609. The only change I have made is to comment out the soaring-specific mavlink messages.

Mavlink message definitions missing

The soaring specific mavlink message definitions are missing from the repository. Please push the updated mavlink repository and update the submodule spec.

Many thanks.