Development of a Planning Approach for the Transition Between Robotic Walking and Running

Abstract

Bipedal legged locomotion promises improved accessibility and navigation of complex and non-barrier-free terrain compared to wheel-based mobility. The tradeoff for increased versatility, especially for humanoid robots, is slow locomotion. Based on existing methods for walking and running, this work presents a new approach that allows for smooth transitions between the two gaits, and thus faster locomotion. Due to the distinct biomechanics of the two gaits, different mathematical models and control approaches are used. The existing trajectory generation algorithms for walking and running developed at the German Aerospace Center were modified and integrated into a combined trajectory generation framework that generates continuous trajectories for multiple walking and running sequences and their transitions. Due to the coupling of different state variables between walking and running, the complete trajectory in the horizontal direction was generated in a single matrix calculation. By resolving the coupling in the vertical direction, the vertical trajectory was computed in a forward recursion from the first to the last gait sequence. The control strategies are unified by integrating the proposed trajectory generation into an inverse dynamics based whole-body controller. The presented approaches are validated in simulations with the humanoid robot Toro.