Neuro-Computing Aspects in Motor Planning & Control

Planning robot movements in an unstructured environment is investigated from the point of view of common sense reasoning as the simulation of a complex dynamical system. We first show how motor redundancy can be solved by simulating passive motion (what we call Passive Motion Paradigm). We then extend this concept in terms of Abstract Force Fields, that allow to express a variety of problems in addition to trajectory formation (force/position control, obstacle avoidance, concurrent tasks, etc.). We show the analogy between this approach and Hopfield nets. We define the notion of Hybrid Motor Schema as the composite ensemble of an analogic component (which consists of the “mental” simulation of model dynamics) and a symbolic component (which contains rules and methods for instantiating the analogic component in a given context: goal, environment, expectations, constraints,…). Simulation results of preliminary implementations are presented.