STRUCTURAL NETWORK MOLDELING AND ADAPTIVE CONTROL OF DYNAMICAL SYSTEMS

摘要

23ABSTRACTStructural network modeling and adaptive control of dynamical systemsTwo new parametric network models for multi-degrees-of-freedom mechanicalsystems are presented based on systems' own functions. The networks are only used tomodel the inertia matrix D(q) and the potential energy V(q) (or the gravitational forcesG(q)), which are functions of position q only. Static networks are therefore sufficientto model them. The complete dynamics of robots are constructed based on thenetworks for D(q) and V(q) (or G(q)). Subsequently, a general adaptive controllerbased on the resulting prescribed network models is given. It can be shown that all theclosed-loop signals are bounded and tracking error goes to zero. Further, with theinventive techniques, adaptive controller design for robots and the like becomes simplya construction of networks and there is no need for dynamic modeling.(Figure 3 is to be published.)