Optimizing footstep placement for robust and stable running of humanoid robots

Abstract

Bipedal legged locomotion offers improved accessibility and the ability to navigate complex, non-barrier-free environments more effectively than wheel-based systems. However, this increased versatility, particularly in humanoid robots, comes at the cost of slower movement. Based on existing methods for humanoid running, this work extends the analytical concept for biologically inspired center of mass (CoM) trajectory planning with optimal footstep and center of pressure adaptation. The analytical method allows for a planner design, where the desired footsteps are input parameters and the current footstep is adjusted to stabilize the motion. This forces the controller to select suboptimal footstep placements when kinematic constraints between the current and next previewed footsteps are reached. The existing trajectory generation algorithms, developed at the German Aerospace Center (DLR), were adapted and extended by optimizing all previewed footsteps to avoid kinematic limitations and increase robustness to disturbances. During stance, the center of pressure (CoP) is optimally adjusted within the contact area of the stance foot, the optimal adaptation of the footstep is executed in flight. These two strategies combined, enable high robustness to challenging starting robot configurations, allowing locomotion transitions, such as walking to running. The trajectory planning algorithm is integrated into an inverse-dynamics based whole-body controller and validated in simulations with the humanoid robot Kangaroo.