In this paper, a fuzzy coordination method based on Interaction Prediction Principle (IPP) and Reinforcement Learning is presented for the optimal control of robot manipulators with three degrees-of-freedom. For this purpose, the robot manipulator is considered as a two-level large-scale system where in the first level, the robot manipulator is decomposed into several subsystems. In the second level, a fuzzy interaction prediction system is introduced for coordination of the overall system where a critic vector is also used for evaluating its performance. The simulation results on using the proposed novel approach ,for optimal control of robot manipulators show its effectiveness and superiority in comparison with the centralized optimization methods.
I. INTRUDUCTIONThe optimal control is one of the most important topics in control theory and optimization of large-scale systems. The problems such as complexity, high dimensionality of variables, geographical separation of subsystems, etc., usually are the burdens for solving the overall problem in a centralized fashion. That is way during past three decades, many approaches such as coordination strategies in multi-level systems and decentralized schemes have been proposed by researchers.In decentralized methods, the system is decomposed into several subsystems where their optimization only depends on local variables while the effects and interactions of other subsystems are either ignored, or considered resulting in robust decentralized sub-optimal control schemes.In the coordination methods, similar to the decentralized approaches, the system is first decomposed into several subsystems, while the effects among them are compensated through a coordinator. In this approach, the control of largescale systems is done by using the hierarchical multi-level control scheme. So hierarchical multi-level control is a common approach that has been presented as an important and efficient method in control of large scale systems.The basic principle of hierarchical control is decomposition of a given large-scale system into several smaller scale systems and then coordination of the resulted sub-systems to reach the optimum solution. In an attempt for improving this strategy, Mesarovic et al. presented one of the earliest formal quantitative treatments of hierarchical systems by the coordination of large-scale systems are mainly based on these two principles.In [3]-[6], using these two principles (IPP & IBP), two new gradient based coordination schemes are introduced that have much faster convergence rates than the classical methods. In [7], [8], two new neuro-fuzzy reinforcement strategies are introduced for intelligent coordination of large-scale systems based on IPP and IBP, where critic vectors are used for evaluation of their performances. In [9]-[11], using the new gradient based coordination schemes, the optimal control of robot manipulators have also been considered. In this paper, by using the novel strategy [7], the optimal control of robot manipulators is investigated. ...
