Real-time control of biped robots based on a moving horizon strategy and virtual passive walking

In order to solve problems in real-time operation of walking biped robots, a real-time control strategy was proposed; its principles combined both moving horizon optimization and virtual passive dynamic walking. Differential characteristics of the mechanical energy of a seven-link robot under virtual gravity conditions were analyzed, and then input torques in different joints constrained by this energy law were determined. Optimized regulation of each link's virtual inclination and of the expected velocities and state parameters of the system at the end point of a gait period minimized difference between actual robot states and expected values at the very beginning of each period, just after each foot lands. Compared with gait programming methods which apply only moving horizon optimization, the method introduced in this paper needs less optimized parameters. In addition, the length of the moving horizon is variable, which increases the feasibility and optimality of robot gait management. Simulation analysis shows that this approach can provide dynamic walking motion including foot rotation. The initial points of multi-step walking satisfied periodic stability conditions on even ground.