Design of a compliant bipedal walking controller for the DARPA Robotics Challenge

This paper provides an overview of the bipedal walking controller implemented on ESCHER, a new torque-controlled humanoid designed by Virginia Tech to compete in the DARPA Robotics Challenge (DRC). The robot's compliant control approach relies on an optimization-based inverse dynamics solver proposed in a previous publication. This work presents two unique features to improve stability on soft and uncertain terrain, developed in preparation for the dirt track and stairs task at the DRC Finals. First, a step adjustment algorithm is introduced to modify the swing foot position based on the divergent component of motion (DCM) error. Second, a simple heuristic is introduced to improve stability on compliant surfaces such as dirt and grass by modifying the design of the center of pressure (CoP) trajectory. The proposed approach is validated through DRC-related experiments demonstrating the robot's ability to climb stairs and traverse soft terrain.