IMRCLab / crazyswarm2

A Large Quadcopter Swarm
MIT License
117 stars 64 forks source link

Axis flips when increasing the number of crazyflies #461

Closed abyvinodm closed 7 months ago

abyvinodm commented 8 months ago

Hi,

We are trying to scale up an application we got working with two crazyflies to four crazyflies. But we are facing unstable crazyflie flight and axis flips.

What worked: I had two crazyflies and two non-crazyflie robot working with crazyswarm2. I was using crazyswarm2 to control the crazyflies in my motion capture volume, as well as get the positions of the non-crazyflie robots.

Problem: I would like to extend this application to four crazyflies and two robots. However, in all of my experiments so far, I saw the crazyflies flip and crash.

Attempts so far:

  1. Use uniform marker configuration for the crazyflies and libobjecttracker for the crazyflies. This is our preferred configuration, but it has the axis flips (at least one out of four).
  2. Use (modified) hello_world script to test: The flips continue to occur (at least one out of four). Here we took off and hover for 20 seconds, and land them. This and the following attempts were made without non-crazyflie robots.
  3. Use (modified) hello_world script and remove the AI decks on board of the crazyflies: This eliminated the flips, but we are not sure why.
  4. Use (modified) hello_world script and use single marker and librigidbody tracker: This removed the flips, but we are no longer getting orientation for the non-crazyflie robots. See https://github.com/IMRCLab/crazyswarm2/discussions/460.

Other notes:

  1. The flips disappear when we go back to two crazyflies and two non-crazyflie robots (with AI deck).
whoenig commented 8 months ago

How close are the CFs flying to each other? And are you using the wifi on the AI deck when flying with the AI decks? Finally, do you use a fairly recent version that has the /status and /connection_statistics topics? If so, what output do you get on these topics (no flying required - just while on the ground)?

abyvinodm commented 7 months ago

Thank you @whoenig. Sorry for the late reply.

  1. How close are the CFs flying to each other? 50 cm.
  2. And are you using the wifi on the AI deck when flying with the AI decks? Yes
  3. Finally, do you use a fairly recent version that has the /status and /connection_statistics topics? We upgraded to the latest commit and switched to using single marker from uniform marker configuration.
  4. If so, what output do you get on these topics (no flying required - just while on the ground)? We didn't check yet.

To work with ground robots, we switched to using single marker for the non-crazyflie ground robots as well, and did some minor computations for getting the yaw of the ground robots.