Bitcraze Forums

Hello!

I'm currently working on a collision avoidance mechanism in a swarm of Cfs that works with the HL-commander (piecewise_poly).
I already managed to send to all members in the swarm the position of its neighbors, but now I'm a little lost from where I can implement a method that uses this information to guarantee a collision free motion of the swarm.

Do you guys have any idea where it can fits best in the current firmware?

My setup is
Crazyflie 2.1
FlowDeck v2
LPS TDoA 2

Appreciate any help or advice from where I can start!

Perhaps https://github.com/bitcraze/crazyflie-f ... issues/567 will help. In the discussion, there are also a few links with example implementations (with the goal being to integrate a version of them into the official firmware in the future.)

whoenig wrote: ↑Thu Apr 16, 2020 4:37 pm Perhaps https://github.com/bitcraze/crazyflie-f ... issues/567 will help. In the discussion, there are also a few links with example implementations (with the goal being to integrate a version of them into the official firmware in the future.)

Thanks for the link Whoenig!

I did not see that issue, and it's good to know that other people are already working on the same thing!

I'm doing my master thesis and the original objective had to be changed given the pandemic situation that we are living.
I'm searching for a new objective for the thesis and I have an interest in this collision avoidance mechanism! I'm not having any feedback from my professor supervisor and I'm kind of lost with which algorithms or approaches I can use to implement this feature in the firmware!

Do you think that some algorithms that you guys have talked about like buffered Voronoi cells (BVC), optimal reciprocal collision avoidance (ORCA) or maybe artificial potentials are good topics to be researched and implemented in the Crazyflie? Do you have any idea of what can be of interest to develop in this thesis to collaborate with the platform?

I know that this is maybe off the technical discussion that this post is supposed to have, but if you guys can help me with some papers to been seen or advice, I will appreciate it!

I have an example from my own research history: https://link.springer.com/article/10.10 ... 018-9760-3.

This enabled avoidance based on the RSSI of bluetooth connection with collision cones. Maybe you can exploit the peer to peer communication of the crazyflies to enable to do the same with the crazyflies?

I think just implementing BVC, ORCA, or Barrier Functions (see e.g., https://arxiv.org/abs/2002.11807) might not be enough for a MS thesis since that has been done. To my knowledge, all of those proofs-of-concept require a motion capture system though. There could be some merit to leverage the peer-to-peer communication (as Kimberly suggested) or the multi-ranger deck (https://www.bitcraze.io/products/multi-ranger-deck/) for drone-to-drone collision avoidance.

kimberly wrote: ↑Mon Apr 20, 2020 9:39 am I have an example from my own research history: https://link.springer.com/article/10.10 ... 018-9760-3.

This enabled avoidance based on the RSSI of bluetooth connection with collision cones. Maybe you can exploit the peer to peer communication of the crazyflies to enable to do the same with the crazyflies?

Hi Kimberly!

First of all, I appreciate all the help that you and whoenig are giving! Thank you!

I liked your paper, and I'm doing some research to see the best way to implement this in the Crazyflie. Do you think that its possible to implement this feature with the DWM1000? I know that this chip does not have the RSSI feature, but I was thinking that maybe the usage of ToF will do the job to get the relative distance between one another. Maybe some mechanism that interrogates the other drone and it sends back the response with the timestamp that came from the interrogation message.

There are some work going on to do "Inter ranging between multiple crazyflies by using the Loco decks" (https://github.com/bitcraze/crazyflie-f ... issues/576). This might br of interest in your application?