Trajectory Tracking with RBF Network Estimator and Dynamic Adaptive SMC Controller for Robot Manipulator

The accuracy in trajectory tracking for robotic manipulators, both in joint and cartesian space, is essential when deputed for industrial applications. The main challenge in controller design is to obtain an accurate model and to guarantee convergence even with external disturbances, friction and model uncertainties. The Radial Basis Function (RBF) neural network effectively overcomes the issue of parameter variations and model uncertainties. Moreover, RBF networks have fast learning ability and better approximation capabilities. A dynamic sliding mode controller provides fast convergence and eliminates the chattering issues associated with conventional sliding mode controllers (SMC). An adaptive dynamic SMC with RBF estimation compensates for modelling uncertainties and provides excellent trajectory tracking. The Lyapunov approach is used to show both the convergence of the RBF weight adjustment rule and the stability of the closed-loop control system. The proposed control strategies are simulated using a 2-DOF manipulator model, and the results are compared with the classical SMC.