The Duke Humanoid Advances Efficient Locomotion

The Duke Humanoid is a 1-meter-tall, open-source platform designed to push the limits of humanoid motion control. Featuring 10 degrees of freedom and human-like leg proportions, it combines passive dynamics with reinforcement learning to achieve efficient, natural walking. By alternating between passive and active joint control, researchers cut locomotion energy costs by 31%, demonstrating a powerful blend of mechanical design and intelligent control for next-generation robotic movement.

Transcript

00:00:01 introducing the Duke humanoid an open-source 10 degrees of freedom extensible research platform for humanoid Locomotion standing 1 M tall and weighing 30 kg the Duke humanoid is the size of an 8-year-old child its lag proportions replicate that of a human lag and its joint range of motion closely matches human

00:00:25 capability inspired by passive Dynamics walking robots our Des design ensures static equilibrium when standing on flat ground the center of gravity aligns symmetrically at the hips providing stability with straight knees this is an essential feature for utilizing passive Dynamics in active control leading to more efficient gate patterns both the knee and Ankle Motors are positioned

00:00:53 above their respective joints and driven by parallel linkage this design elevates the L Center of Mass reduces the lower lag inertia and minimize the torque required by the motors to evaluate the robot's Locomotion capabilities we developed a deep reinforcement learning policy for walking the policy was trained entirely in simulation here we show the progress of the training as the

00:01:20 robot learns to walk we then deployed the train policy zero shot on the physical robot here is a demonstration of a soral transfer at a command velocity of3 m/ second this serves as our Baseline for Locomotion optimization with the goal of attaining better Energy Efficiency in Locomotion we developed a control policy within the RL framework to leverage passive

00:01:48 Dynamics our passive control policy modulates the joint torqus and as a result The Joint switches periodically between a passive and active mode this allows the robot to utilize its natural Dynamics during lag swing thus reducing the need for active motor control to demonstrate the Energy Efficiency of our passive policy we compareed the cost of Transport between the active and the

00:02:13 passive policy real world experiments show a 31% reduction in the cost of Transport at a command velocity of3 m/ second the Duke humanoid serves as an open-source research platform for studying humanoid locomotion the integration of passive action into the reinforcement learning control showcases the potential for achieving more natural and efficient humanoid movements

00:02:40 offering a promising direction for future research please check our website for the open source of both software and Hardware