Abstract:
This research proposes the development of adaptive algorithms for the dynamic modeling and control of a robot arm for industrial applications. The mathematical model for a six-degree-of- freedom 6-DoF R700 robot arm was developed to accommodate different parameters with a view of quantifying the effects of different parameters on the manipulated and controlled variables. The manipulated variables inputs were the electrical input voltages to the servo motors while the inclined angles made by the joints were (vertical Shoulder, horizontal Shoulder, elbow, the End-effector Yaw, End-effector Pitch and the End-effector Roll) of the developed 6-DoF robot arm and they constituted the controlled variables (outputs). The developed mathematical model of the 6-DoF R700 robot arm was perturbed and simulated to generate the input-output data for neural-Network training. A NN-based (ARLS) algorithm is proposed for the NN-training to obtain the dynamic NN model of the R700 robot arm, On the basis of the NN-model of the R700 robot arm, a dynamic NN-based adaptive control algorithm was developed based on feedback linearization. Four scenarios were proposed, as case studies to generate the developed trajectory to be tracked by the proposed adaptive control scheme for performance comparison, was developed the proposed NN-based ARLS was compared with another online Incremental Back-propagation (INCBP) while the proposed NN-based adaptive control algorithm was compared with the standard proportional- Integral- Derivative (PID) controller. The simulations showed that the proposed ARLS has less computation time, also has small values of the Akaike’s final prediction error (AFPE) estimate which indicates that the trained networks by ARLS algorithm captured the underlying structure and dynamics of the robot arm trajectory better than the INCBP
