Learning to Walk through Imitation

Programming a humanoid robot to walk is a challenging problem in robotics. Traditional approaches rely heavily on prior knowledge of the robot's physical parameters to devise sophisticated control algorithms for generating a stable gait. In this paper, we provide, to our knowledge, the first demonstration that a humanoid robot can learn to walk directly by imitating a human gait obtained from motion capture (mocap) data. Training using human motion capture is an intuitive and flexible approach to programming a robot but direct usage of mocap data usually results in dynamically unstable motion. Furthermore, optimization using mocap data in the humanoid full-body joint-space is typically intractable. We propose a new modelfree approach to tractable imitation-based learning in humanoids. We represent kinematic information from human motion capture in a low dimensional subspace and map motor commands in this lowdimensional space to sensory feedback to learn a predictive dynamic model. This model is used within an optimization framework to estimate optimal motor commands that satisfy the initial kinematic constraints as best as possible while at the same time generating dynamically stable motion. We demonstrate the viability of our approach by providing examples of dynamically stable walking learned from mocap data using both a simulator and a real humanoid robot.