bipedal-locomotion

Reproducible Microduck RL experiments, policies, simulation assets, and hardware add-ons

Biped robot locomotion with deep RL: Isaac Lab + MuJoCo, custom PPO/SAC/DDPG/D4PG/MPO, MLP/LSTM/Transformer, motion imitation, sim-to-real

Project Eleven: Developing an agile quadruped robot using physics-based simulation and reinforcement learning. Starting with a bipedal prototype to understand locomotion dynamics, then scaling to four legs. Includes complete CAD-to-simulation workflow (SolidWorks → URDF → MuJoCo XML), control systems implementation, and sensor integration for gait

RL-based framework for robust bipedal locomotion on uneven terrains and staircases using the Unitree H1 humanoid robot in the Genesis simulator with PPO, curriculum learning, gait phase generation, and domain randomization.