Forrest

Overview

I work with RoboTUM on our flagship bipedal humanoid project, aiming to build the world’s fastest walking robot, called Forrest. My focus is on reinforcement learning policy development and large-scale training in IsaacLab, developing controllers and RL pipelines for our own designed humanoid using the IsaacSim/IsaacLab and RSL-RL stack.

Key Highlights

• Contributing to RoboTUM's goal of building the world's fastest walking bipedal humanoid robot
• Developed reinforcement learning locomotion policies with large-scale training in IsaacSim/IsaacLab using the RSL-RL stack
• Designed a CPG-based (central pattern generator) controller for locomotion
• Implemented terrain adaptation for robust walking across varied surfaces
• Developed sim-to-real transfer methods to deploy trained RL policies on the physical robot