During the course of my research this semester I needed to implement the Powerlaw Collision Avoidance model. This is a force-based model to intelligently repel agents based on the time-to-collision, or TTC. There is already an official python implementation however it does not support static obstacle avoidance. In order to implement this into my code I chose to mimic the Python ORCA Wrapper and write my own Cython wrapper for the official C++ Powerlaw model.
The code is on my GitHub along with a simple 2 vs. 1 agent configuration example that mimics some C++ example code. The only issue I have found is that python has differing precision in floats than C++ (at least on my machine). This is only an issue when specifying the agent's initial parameters and does not have any actual baring on the underlying simulation.