Real-time walking step timing adaptation by restricting duration decision for the first footstep

This paper presents two model predictive control formulations for automatically deciding walking step duration for humanoid robots. In the proposed schemes, constrained optimization problems are solved online to simultaneously decide center of mass jerks, foothold positions, and step timing while respecting actuation, geometry, and contact constraints. The main contribution of this work is showing that restricting the duration decision to the first footstep is an effective strategy to achieve real-time computational time. Simulated and real robot experiments were undertaken to present and discuss the capabilities of the proposed formulations.