A design of inverse control system based on multiple support vector machine

The paper works out an online self-learning control plant for multiple support vector machine inverse control. The multiple support vector machine model applies subtractive clustering algorithm by which the input space is divided into several small local spaces. By means of least squares support vector machine, the sub-models are established. The prediction output of each sub-model is connected by principal components regression method so that identification of the inverse dynamics model of the system is achieved. Combining inverse model of the system as a system controller with the controlled plant, a SVM direct inverse control system is constituted. In order to overcome the influence of inverse model identification error, a SVM direct inverse control system with the PID compensation is designed in the paper. The simulation research proves that the control strategy can provide the system with good tracking performance, resistance to interference and a better robustness.