search

larics / icuas22_competition

The main repository for the ICUAS 2022 UAV competition.
17 stars 4 forks source link

UAV flies off to the sky after some time #16

Closed EceChaik closed 2 years ago

EceChaik commented 2 years ago

Hi guys,

I am running into a repeated issue with the competition stack that I haven't been able to solve. The UAV arms and takes off successfully, it also successfully listens to commands on the /red/tracker/input_pose topic. However, after some time in the simulation, without fail it starts climbing infinitely. There doesn't appear to be a clear error that triggers this behavior, but I am attaching a screenshot of the main terminals when this occurs. Screenshot from 2022-02-23 17-12-48 I think it has something to do with the 'height' messages appearing in the ardupilot console, though I have no idea where they originate.

fpetric commented 2 years ago

@EceChaik thanks for the report, we will look into it

lmark1 commented 2 years ago

Hello @EceChaik

Thanks for the report!

The issue should be resolved with PR #18 and by updating the Docker images or the natively installed simulation stack as described in Troubleshooting.

Long story short - the internal altitude estimation in Ardupilot firmware started diverging and forced the UAV to start climbing when not using GPS. The issue seems to be fixed when updating the Ardupilot firmware to v4.1.

EceChaik commented 2 years ago

Hi @lmark1 , I thought that in these simulations the uav is flown in GUIDED_NOGPS mode and you control the position with your own position controller, then feed the attitude commands to Ardupilot. If that's the case, why should the internal altitude estimate matter, since you're hijacking the internal position controller? Unless you are still using the ardupilot's internal positioning estimate

lmark1 commented 2 years ago

We do control the UAV with our own position controller and in this case the Gazebo odometry sensor feedback. However...

By default the Ardupilot firmware interprets the thrust command (in GUIDED_NOGPS mode) as climb rate. When using 'thrust' as climb rate Ardupilot still requires some altitude information in order to know if it's climbing up or down.

By changing the parameter https://github.com/larics/icuas22_competition/blob/8c4d15b49b82a3bfb8c7561b76caeb5724d60da9/parameters/arducopter_nogps.params#L361 we are able to tell the firmware to interpret thrust as thrust (and not as climb rate) thus eliminating the reliance on internal altitude estimates.

EceChaik commented 2 years ago

Ooohh, very nice