FPGA implementation of a hybrid neural fuzzy controller for flexible-joint manipulators with uncertain dynamics

Abstract

In this paper, we propose a VLSI (very large scale integrated) implementation of a hybrid neural fuzzy control scheme on a Xilinx Virtex2 Pro 2VP30 field programmable gate array (FPGA) for a flexible-joint robot manipulators with uncertain dynamics. The control strategy is based on a feedforward artificial neural network that approximates the manipulator's inverse dynamics. An adaptive feedback fuzzy sliding mode controller is used to compensate for residual errors. A systolic top-down hardware design methodology takes full advantage of the neural networks' inherent parallelism that allows the controller to operate at high frequencies. Furthermore, a pipeline strategy was used to speed-up the feedback fuzzy controller's inference process. Numerical simulations and the synthesis of the results highlight the effectiveness of the proposed controller in compensating for the nonlinear unknown manipulator's dynamics.