This project was completed as part of the Optimal Control and Reinforcement Learning Course at Carnegie Mellon University. We used direct collocation to generate an optimal trajectory for a Unitree G1 Humanoid to kick a ball at a target, and used inifinte-horizon and time-varying LQR to balance the robot and track the desired kicking trajectory. Note that there is an initial balancing period prior to the kicking motion.
