Artificial Locomotion Control: from Human to Robots
Résumé
This paper concerns the simultaneous synthesis and control of walking gaits for biped robots. The goal is to propose an adaptable and reactive control law for two-legged machines. The problem is addressed with human locomotion as a reference. The starting point of our work is an analysis of human walking from descriptive (biomechanics) as well as explicative (neuroscience and physiology) points of view, the objective being to stress the relevant elements for the approach of robot control. The adopted principles are then: no joint trajectory tracking; explicit distinction and integration of postural and walking control; use of evolutive optimization objectives; on-line event handling and environment adaptation and anticipation. This leads to the synthesis of an original control scheme based on non-linear model predictive control: Trajectory Free NMPC. The movement is specified implicitly through coherent physical inequality constraints. Dynamic model and internal limitations of the system are part of the problem constraints. This work is validated by simulation results obtained for the Bip and Rabbit biped robots in various walking and standing situations and compared to human data recorded in these same situations.