Grid Search Optimized SVM Method for Dish-like Underwater Robot Attitude Prediction

Abstract

The control of dish-like underwater robot motion is complex. It involves many kinds of influencing factors and it's also a nonlinear process. The model of attitude motion control is very important for the accuracy control and self adapting predictive control. For establishing the attitude motion model and predicting the attitude, SVM algorithm was used to construct a MIMO identifier in this paper. Moreover, in order to improve the effect of the identification and prediction, the grid search method was adopted to optimize the key parameter C and g in SVM. At last the effects were contrasted with GA and PSO optimized SVM algorithm by the data from the experiments in the pool, the results proved the superiority of grid search method in both calculating time and optimizing results. The results show the well performance of this GS-SVM on the identification and prediction for the attitude of dish-like underwater robot.