Intelligent Control Method Based On CMAC Theory And Its Application On Microgravity Compensation System

Cerebellar model articulation controller (CMAC) neural networks has the advantages of fast learning and is very suitable for real-time nonlinear systems. However, it has some drawbacks in choosing learning rate, quantification precision and sampling precision and etc. Moreover, its generalization ability hasn't been theoretically studied. Research is still ongoing in terms of how to choose its center of Fuzzy CMAC. In the application of CMAC, its robustness still needs to be strengthened. In this dissertation, the theory of CMAC is further studied from its learning rate's and quantification precision parameters' optimization, generalization ability, the center of FCMAC. We have also studied a robust FCMAC algorithm, and explored the application of CMAC in 863 project "Microgravity Compensation System".In this dissertation, the nonlinear system is the microgravity compensation system. We propose the constant tension method and the semi-active compensation method to simulate microgravity environment. In the two systems, we cannot derive exact mathematical model because of friction and other uncertain factors. Therefore, the gratifying compensation result can't be obtained by traditional methods. Using CMAC, we can get good result because CMAC has the ability of approaching nonlinear system. The experiment results demonstrate that the performance of the control system can be improved by using CMAC and verify the correctness of the modified theory of CMAC.The main contribution in this dissertation can be summarized as follows:1. For the microgravity system, the kinematics formulas of whole system are deduced. According to the calculation results, the mathematics model is then deduced.2. We proposed a novel method of optimal selection based on the adaptive genetic algorithm to select learning rate. Without suitable learning rate, the system will be instable or its convergence may be slow. Compared the adaptive GA with the traditional method, the CMAC using adaptive GA can get faster convergence and have better stability. It has been verified by the simulation results.