Error Compensation for Long Arm Manipulator Based on Deflection Modeling and Neural Network

Abstract

Long arm manipulators are designed to work in special conditions including aviation, engineering, and other scenarios that require a large span operation. However, since the long arm will cause a large flexible error, the manipulators maintain large terminal absolute error and difficulty in control. In addition, testing in a real machine is time and economic consuming, and obtaining enough data is untoward. Under such conditions, this paper proposes an error compensation method for a long arm manipulator combining deflection error modeling and neural network, using a specially designed long arm manipulator. By using this method, better results are achieved than traditional error modeling alone since non-traceable errors are also compensated for. The result is also better than neural network compensation alone since in the case of less training data preferable results can still be achieved.