Improving MPC Based Teleoperation Systems by Employing Intelligent Optimization Algorithms

Though Model Predictive Control (MPC) is among the promising methods for teleoperation system applications, its time consuming computations is considered as a major limitation. Intelligent algorithms seem to be powerful tools to overcome such issues. Prediction of the process outputs by using a model and solving one constrained linear or nonlinear optimization problem in the time interval between two samples, leads to sever computational complexity in the controller calculations. This limits the application of the MPC algorithm only to the slow dynamic and low dimensional processes. To overcome this drawback, this paper focuses on the optimization part and proposes the use of intelligent methods and efficient optimization algorithms to reduce computation time of the MPC algorithm. First the MPC algorithm will be re-expressed as a standard quadratic programming (QP) optimization problem with bound constraints. Then, a number of common optimization methods which were found suitable for solving this type of optimization problem, namely sequential quadratic programming (SQP), genetic algorithm (GA), active set, interior point, singular value decomposition (SVD) method and recurrent neural networks (RNN), are employed and a comparison is presented. The efficiency of the proposed algorithm will be investigated through applying these algorithms to the teleoperation systems.