Combined Controllers that Follow Imperfect Input Motions for Humanoid Robots

Humanoid robots have the potential to become a part of everyday life as their hardware and software challenges are being solved. In this paper we present a system that gets as input a motion trajectory in the form of motion capture data, and produces a controller that controls a humanoid robot in real-time to achieve a motion trajectory that is similar to the input motion data. The controller expects the input motion data not to be dynamically feasible for the robot and employs a combined controller with corrective components to keep the robot balanced while following the motion. Since the system can run in real-time, it can be thought of a candidate for teleoperation of humanoid robots using motion capture hardware. Type of Report: Other Department of Computer Science & Engineering Washington University in St. Louis Campus Box 1045 St. Louis, MO 63130 ph: (314) 935-6160 Combined Controllers that Follow Imperfect Input Motions for Humanoid Robots Gazihan Alankus, O. Burchan Bayazit Washington University in St. Louis {gazihan, bayazit}@cse.wustl.edu Abstract— Humanoid robots have the potential to become a part of everyday life as their hardware and software challenges are being solved. In this paper we present a system that gets as input a motion trajectory in the form of motion capture data, and produces a controller that controls a humanoid robot in real-time to achieve a motion trajectory that is similar to the input motion data. The controller expects the input motion data not to be dynamically feasible for the robot and employs a combined controller with corrective components to keep the robot balanced while following the motion. Since the system can run in real-time, it can be thought of a candidate for teleoperation of humanoid robots using motion capture hardware. Humanoid robots have the potential to become a part of everyday life as their hardware and software challenges are being solved. In this paper we present a system that gets as input a motion trajectory in the form of motion capture data, and produces a controller that controls a humanoid robot in real-time to achieve a motion trajectory that is similar to the input motion data. The controller expects the input motion data not to be dynamically feasible for the robot and employs a combined controller with corrective components to keep the robot balanced while following the motion. Since the system can run in real-time, it can be thought of a candidate for teleoperation of humanoid robots using motion capture hardware.