Hybrid Fuzzy Fictitious Play based Control for Robotic Manipulators

---------------------------------------------------------------------***--------------------------------------------------------------------Abstract In this work, we propose a hybrid controller based on game theory for robotic manipulators. The controller implements the hybrid strategy using fuzzy inference systems on a continuous time two link manipulator. We strive to formulate what may be called a “safely consistent” Markov theory based controller by a gradual mixing of the Markov game based safe control with a fictitious play based consistent control. The controller has been evaluated on a rigid two link manipulator and its performance compared against (i) Markov game based fuzzy controller and (ii) fuzzy Q controller. Empirical results show that the controller can handle nonlinearities and parameter variations quite effectively.