GPU RTT implementation in Issac Gym

I’m implementing the Rapidly Exploring Random Trees algorithm in Issac Gym, and I’d like to keep everything on the GPU.

If you’re not familiar with RRT, that’s okay. I’m stuck at two major subproblems:

  1. Checking collisions between a provided mask and a line defined by its endpoints. If I can generate a mask for the line, then this is as easy as an AND operation. But how to generate a line mask quickly?
  2. Finding the closest point from a list of points (around 100)

As you can see, these are quite trivial to do efficiently on cpu, but not so on gpu. I’m not adverse to delving into CUDA code, but I’d like to keep everything in Python if possible.

Thanks!