Do you like sumo ?
Do you like wonky physics and hover cars!?
Do you like Reinforcement Learning!?
This is an experiment I bodged together where I tried to train two competitive agents using the same Neural Network and a lot of trial and error to try and get a working policy.