This research project implements deep RL algorithms for motion imitation, enabling simulated characters to learn and perform complex physical skills from motion capture data. The system supports training through adversarial motion priors and multiple imitation techniques. It is now deprecated in favor of the faster MimicKit implementation.