whoenig / uav_trajectories

Helper scripts and programs for trajectories
MIT License
85 stars 46 forks source link

Drone swarm formation path-planning for collision avoidance. #17

Closed raghavsuri13 closed 1 year ago

raghavsuri13 commented 1 year ago

Hi,

I'm working on a project that involves a large number of drones that are supposed to takeoff from an initial grid, and create a set of formations one after another, at the end of which the drones land at their initial starting positions, how can I use this to plan the trajectories for each of the drones in the formation such that collision avoidance is guaranteed. There are no static obstacles in the simulation, the only obstacles are the drones themselves.

Thanks! P.S. The number of drones may range from 100 - 5000.

whoenig commented 1 year ago

This repo is for single UAVs only. For your use case, I'd recommend Buffered Voronoi Cells or Control Barrier Functions.

raghavsuri13 commented 1 year ago

Hi, thanks for the advice, I'd also come across one of your research papers called "Trajectory Planning for Quadrotor Swarms". I believe that approach would be very apt for my use case considering we need to just preplan the paths for all the drones, we don't require real-time monitoring as the drones would potentially have no onboard sensors or cameras as well as no separate camera monitoring them. Could you suggest the correct approach for the same or would you suggest I go with Buffered Voronoi Cells / Control Barrier Functions. Thanks for the help!

whoenig commented 1 year ago

If you don't have any obstacles, then the paper you mentioned is overkill and BVC/CBF are the better choice.

raghavsuri13 commented 1 year ago

Hi, Could I get in touch with you somewhere else privately, I have a few questions about the implementation and would need to explain the project that I'm working on as well. Again thanks a lot for all the help, I really appreciate it.

whoenig commented 1 year ago

You can find my contact info at https://whoenig.github.io/.

raghavsuri13 commented 1 year ago

Perfect, I guess we can mark this issue closed.