A neurocontroller for robot manipulators
Y. Derbal, Mohamed M. Bayoumi · 2002
Feedforward neural networks have been proven to be capable of approximating nonlinear mappings on compact sets. This property has been used in the design of a large number of robot controllers. The overwhelming majority of these NN based robot controllers lack any substantial stability analysis. We propose a neurocontroller where: the unmodelled dynamics are considered; and the closed loop system is L/sub /spl infin// stable provided that certain m assumptions are satisfied.>