Study Of Intelligent Control Strategy In The Planar Single Inverted Pendulum System

With the quick development of science and technology, new control methods appear unceasingly. The inverted pendulum systems get extensive research as an important experiment means to test the validities of new control theories and methods. In this paper, the planar single inverted pendulum of Googol Technology is selected as a controlled object to study and test the intelligent control method for multi-variables complex system this paper proposed, that is, a control strategy using of Mamdani Neural Fuzzy Inference System based on Rough Set Theory and variables grouping compensation.First, this paper introduces the basic concepts of Rough Set Theory, and has a deep research on its knowledge obtain method and idea, then proposes a fuzzy rules generation algorithm for theplanar inverted pendulum system. This algorithm combines Rough Set theory and Fuzzy Set theory in the Fuzzy Decision Table construction, and selects an algorithm using of attribute significant degree in the Decision Table reduction. The gotten fuzzy control rules are more reasonable and have better control adaptability to the construction of Mamdani Neural Fuzzy Inference System.Secondly, this paper has a theoretical analysis for the Neural Fuzzy Inference System based on Mamdani Fuzzy Model, involving its control idea, its network structure and its learning algorithm. Then based on these theoretical analysis foundations, the paper points out the advantage of Neural Fuzzy Inference System control method, as well as its shortage in the planar single inverted pendulum system control, that is, network structure is huge, network design is complex, network training speed is slow, and it cannot use sample data feature fully.Thirdly, for the shortage of Neural Fuzzy Inference System technology in the planar inverted pendulum system control, this paper proposes a control idea of Mamdani Neural Fuzzy Inference System based on Rough Set Theory. That is, getting fuzzy rulesfrom the planar inverted pendulum system's sample data based on Rough Set Theory first, then mapping the reduced fuzzy rules to a Neural Fuzzy Inference System model directly. This method doesn't need complex learning process of the network structure, and it can improve the network learning speed and the system performance. At the same time, according to the planar single inverted pendulum system's control property, this paper proposes a combination of variables grouping compensation control idea and the RMNFIS control idea. This variables grouping compensation control method can effectively reduce the control variables number for Neural Fuzzy Inference System, and it has higher control reliability for multi-variables complex system. Compared with the general state variables synthetic method in the planar single inverted pendulum system, this method has advantages that its control rules establishment and control method are easier.Finally, the intelligent control method proposed in this paper for the planar single inverted pendulum system is validated, and comparing with the LQR optimum control method, the simulation results show its validity.