Neural optimal control of robotic manipulators with CMAC networks

In this paper, an adaptive GA CMAC-based inverse kinematics solution of a robotic manipulator is presented. Real-time control of the end effectors of a robot requires computationally efficient solutions of the inverse kinematics. The inverse kinematics of a robotic manipulator is nonlinear and in some cases it cannot be solved in closed form. Some traditional solutions such as iterative and geometric are inadequate if the manipulator is more complex. Neural network such as CMAC approaches are studied in solving the inverse kinematics problem. But conventional CMAC can not give us appropriate values for the parameters of CMAC module such as the learning rate of CMAC and the size of generalization. Without suitable parameters, the convergence speed of CMAC can be slow. In this paper, we adopt the adaptive GA to search for the optimal parameters. Furthermore, the developed method is applied to a two-joint robot. Finally, the effectiveness of the proposed adaptive GA CMAC model control system is verified by simulation experimental results.