Intelligent modelling of flexible manipulator systems

Abstract

Intelligent techniques, such as genetic algorithms (GAs) and neural networks (NNs) have attracted the attention of the wider control community due to their various advantageous features in relation to system identification and control. The major advantage of utilising GAs for system identification is that they simultaneously evaluate many points in the parameter space and converge towards the global solution. On the other hand, the use of NNs is inspired by their ability to mimic the capabilities of the brain such as learning, adaptation, association and generalisation. More importantly, NNs can address the nonlinearity of a system. This paper presents an investigation into the use of GAs and NNs to model a single-link flexible manipulator. The GA and NN based identification of the system are realised in this investigation by minimising the prediction error of the actual plant output and the model output. However, to allow interactive and user friendly features, that are desired especially in computer aided teaching and research, be incorporated a modelling, simulation and control environment is developed in this work for flexible manipulators using Matlab and Simulink. To this end the authors have developed an interactive and user-friendly environment referred to as SCEFMAS (Simulation and Control Environment of Flexible Manipulator Systems) [1]. As an on-going development process, the SCEFMAS environment is enhanced by the addition of intelligent modelling using NNs and GAs. © 2006 Springer-Verlag Berlin Heidelberg.