A Fast Decoupling Method for Multi-input Multi-output Nonlinear Continuous-Time System

Neural network inverse (NNI) method has been applied in decoupling control of MIMO nonlinear continuous-time system, but slow learning speed restricts its application. To solve the problem, a new online learning algorithm called modified online regularized extreme learning machine (MO-RELM) is proposed. MO-RELM can learn the training data one by one or chunk by chunk and discard the data for which the training has done. Finally, through connecting NNI based on MO-RELM with the original MIMO system, the constructed pseudo linear composite system (PLCS) can decouple the nonlinear coupled system into a number of independent pseudo linear SISO systems online. Simulation results show that the proposed fast decoupling method realizes online decoupling control of MIMO nonlinear continuous-time system. The method has good application prospect.