Dynamic Reinforcement Learning Crowds is a tool for crowd simulation based on the SIGGRAPH paper CCP: Configurable Crowd Profiles. The project uses Unity’s ML-Agents package to let agents learn different crowd simulation behaviors through reinforcement learning. Users can then start a simulation and during runtime modify agent behavior parameters, causing changes in how the crowd moves. The Unity backend is also able to be linked to a Maya environment through a TCP connection, allowing the tool to be used in Maya if preferred. You can read to project writeup or check out the demo video below.