Optimizing the stepping of a humanoid robot for a sequence of tasks

In this paper, we present a method to optimize the stepping plan of a humanoid robot according to a sequence of tasks. There is a gain in efficiency to expect from such a method as it helps reduce the time and energy spent by the robot in locomotion. In order to achieve this, we map a motion of the robot satisfying a sequence of tasks to the configuration of a virtual redundant manipulator. This sparse representation keeps track of the footsteps and a posture of the robot for each of the tasks motivating the motion. It leads us to transform the stepping plan optimization problem into an inverse kinematics problem on the virtual manipulator. We also propose an alternative method following a dynamic programming approach that optimizes the stepping plan by knowing one task ahead at a time. We will illustrate these methods in simulation on the humanoid robot HRP-2.