Model predictive control integrated with multi-agent particle swarm optimization-based SVR

The kernel function's selection has a great impact on the performance of support vector regression (SVR). A new method of nonlinear model predictive control (NMPC) based on polynomial kernel SVR is put forward, and multi-agent particle swarm optimization algorithm is introduced to obtain the optimal control law of rolling optimization in NMPC. Compares with the NMPC based on quadratic polynomial kernel SVR, the simulation results show that the characteristics of our method, such as, overshoot, volatility and tracking are superior to those of quadratic polynomial kernel SVR.