A 5-DOF biped with a crank-slider leg extension and adjustable center-of-mass height, built around macaroni penguin inertial properties. In MuJoCo and on hardware with motion capture, penguin-style torso-over-stance-leg motion with a high COM was 1.8× faster with 44% lower cost of transport than an upright gait at μ = 0.12, reversing at μ = 0.45. CPG-RL and imitation RL policies now climb 6° slopes at μ = 0.1 in simulation.

The first steerable single-motor biped. I built a PID attitude controller for directional control and analyzed 2D contact dynamics to optimize gait efficiency and limit-cycle stability.

Leading the Unitree G1 humanoid velocity-tracking and LEAP Hand in-hand reorientation tasks: converting non-differentiable reward and termination terms into smooth surrogates, tuning first-order training, and benchmarking against zeroth-order baselines (PPO) on sample efficiency and wall-clock time.

How should a biped's design change with its size? Surveying existing bipedal robots shows mass scales as L² rather than the isometric L³. Simulating Mugatu (0.15 m legs) and Zippy (0.025 m legs) in Drake across a 60× range of leg lengths, walking speed follows dynamic similarity (v ∝ L^½), joint torque scales as mL, and feasible foot size scales linearly with leg length.

Details coming soon.