Analysis of the effect of backlash and joint acceleration measurement noise in the adaptive control of electro-mechanical systems

Abstract

A new branch of computational cybernetics was developed on the principles akin to that of the traditional soft computing (SC). At the costs of limited expired circle of applications a priori known, uniform, lucid structure of reduced size can replace the normally enormous structures of traditional soft computing. Furthermore, traditional deterministic, semi-stochastic or stochastic machine learning can be replaced by simple and short explicit algebraic procedures especially fit to real time applications since in this way considerable computational advantages can be achieved. The key element of the approach the modified renormalization Transformation supported by the application of a simple linear transformations, and the use of some simple interpolation or prediction technique. It is analysed how the quality of control is influenced by the backlash of the robot's joints and by the noise of the joint acceleration measurements in the presence of unmodeled internal mechanical and electric degrees of freedom. Simulation examples are presented for the control of DC electric motors driven 3 DOF SCARA arm by the use of a special family of symplectic transformations. It is concluded that the method is surprisingly robust against considerable noise and backlash.