Neural-fuzzy control system for robotic manipulators

This article presents a control system structure, as well as a control algorithm, that combines neural networks with fuzzy logic for dynamical compensation of both structured and unstructured uncertainties.

A new fuzzy reasoning method is derived in the neural mechanism and implemented with a cerebellar model articulation controller (CMAC), which outperforms conventional fuzzy controllers by reducing computational complexity and providing a learning ability that conventional fuzzy systems do not have. The overall control system is proven to be stable. The simulation results confirm that the system can track the desired position for both set-point and dynamic tracking in the presence of uncertainties such as changing payload, various frictions, and unknown disturbances.