Abstract：Based on the ANN αth order inverse system method, this paper presents a novel method to approximately decouple the controlled nonlinear MDF robot into a number of independent SISO linear subsystems. Then well developed linear system theory is used to design an additional feedback linear controller to successfully realize the real time tracking of any arbitrary trajectory starting from arbitrary initial conditions even if the desired trajectory is not known in advance. Experiment results show its very good decoupling control and tracking performance. With the character of simple structure and without knowing the control system’s precise model, this method can be generalized to a class of MIMO nonlinear system in real field.
[1] Xianzhong Dai,Dan He,Ann αth-order Inverse System Method for Control of MIMO N onlinear Systems and Its Application in Robot Control. Proceedings of The Fourth International Conference on Motion and Vibration Control, ETH Zurich,Switz land,1998,3： 1177-1182 [2] Xianzhong Dai,Jun Liu,Chunbo Feng,Dan He. Neural Network αth-order Inverse System Method for the Control of Nonlinear Continuous Systems. IEEE Proceedings-Control Theory and Applications,1998,145(6): 519-522 [3] Johansson R. Adaptive Control of Robot Manipulators Motion. IEEE Trans on Rob otics and Automation,1990,(6): 483-490 [4] Lilai Yan, James Li C. Robot Learning Control Based on Recurrent Neural Netwo rk Inverse Model. Journal of Robotic Systems, 1997,14(3): 199-212
