Planning heavy lifts for humanoid robots

Lifting heavy objects poses a unique challenge for humanoid robots and more broadly, for any robot which is responsible for maintaining its own balance. Configurations that are balanced without supporting a heavy object's weight might not be balanced while the object's weight is being supported, and vice versa. In this paper, we present a series of planning techniques which resolve these issues without relying on real time control methods or extensive force/torque sensing. We introduce the novel concept of the Virtual Task Dimension (VTD) for motion planners, which can handle the transition between balancing constraints. We describe the implementation of these techniques and offer suggestions for obtaining fast and reliable solutions. We also demonstrate the algorithms running on a DRC-Hubo humanoid robot.