Robust adaptive dead zone technology for fault-tolerant control of robot manipulators using neural networks. (English) Zbl 1047.93546

Song, Q.; Hu, W. J.; Yin, L.; Soh, Y. C.

J. Intell. Robot. Syst. 33, No. 2, 113-137 (2002).

Summary: A multi-layered feed-forward neural network is trained on-line by the robust adaptive dead zone scheme to identify simulated faults occurring in a robot system and reconfigure the control law to prevent the tracking performance from deteriorating in the presence of system uncertainty. Considering the fact that system uncertainty can not be known a priori, the proposed robust adaptive dead zone scheme can estimate the upper bound of system uncertainty on-line to ensure convergence of the training algorithm, and in turn the stability of the control system. A discrete-time robust weight-tuning algorithm using the adaptive dead zone scheme is presented with a complete convergence proof. The effectiveness of the proposed methodology has been shown by simulations for a two-link robot manipulator.

MSC:

93C85	Automated systems (robots, etc.) in control theory
68T40	Artificial intelligence for robotics
68T05	Learning and adaptive systems in artificial intelligence

Keywords:

neural networks; robotic fault-tolerant control; adaptive dead zone