Neural Network-based Robot Control Study

This paper is based on the project-"The research of nonlinear control technology". The main task is to discuss the use of neural network intelligent control methods in the control of robot. Composite control combining neuro-based inverse control with FID control is studied. Combined with the robot dynamics, a heuristic learning algorithm is applied to the learning of networks. Theoretical analysis and simulation indicate that the algorithm not only can highly enhance the speed of convergence but also can satisfy the quick, precise and real-time control of robot. Due to uncertain disturbances or unmodeled dynamics, a neuro-based model reference adaptive control method is adopted. Using dynamic recurrent neural networks as the identification and controller, it can achieve high-precision tracking control of robot and excellent anti-jamming. At last, this paper discussed the integrated technology of neural network and fuzzy system, and predicted the application of fuzzy neural networks to the control of robot.