Improving the performance of industrial robot manipulators by neural networks

Abstract

Robot manipulators are nonlinear systems. Centrifugal and Coriolis forces as well as the influence of gravitation and friction depend on the state variables of the system. In the presence of strong nonlinearities, linear PID controllers for the individual joint axis drives, usually employed in industrial applications, cannot provide satisfactory performance due to their inherent limitations. Model-based schemes have the disadvantage that they require accurate system models, which are difficult to obtain. The problem is solved by using multilayer feedforward neural networks, which do not rely on a system model. They are used as an addition to the existing linear individual joint control structure. The convergence of the system is proved using Lyapunov's stability theory. Experiments obtained on a two-degree-of-freedom manipulator demonstrate the effectiveness of the proposed technique.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>