Evolving Neural Controllers for Robot Manipulators
We examine here the feasibility of using evolutionary techniques to produce controllers for a standard robot arm. The main advantage of our technique of solving path planning problems is that the neural network (once trained) can be used for the same robot, with a variety of start and target positions. The genetic algorithm learns, and encodes implicitly, the calibration parameters of both the robot and the overhead camera, as well as the inverse kinematics of the robot. The results show that the evolved neural network controllers are reusable and allow multiple start and target positions.
KeywordsNeural Network Genetic Algorithm Target Position Robot Manipulator Inverse Kinematic
Unable to display preview. Download preview PDF.
- D. H. Ballard and C. M. Brown. Computer Vision. Prentice Hall, Englewood Cliffs, NJ, 1982.Google Scholar
- I. Harvey. The Artificial Evolution of Adaptive Behaviour. PhD thesis, University of Sussex, April 1995.Google Scholar
- P. Husbands, I. Harvey, and D. Cliff. Analysing recurrent dynamical networks evolved for robot control. In Proceedings of the Third IEE International Conference on Artificial Neural Networks (ANN93). IEE Press, 1993.Google Scholar
- J. T. Ngo and J. Marks. Spacetime constraints revisited. In Computer Graphics Proceedings, pages 343–350. SIGGRAPH, 1993.Google Scholar
- R. Salama and P. Hingston. Evolving neural network controllers. In Proceedings of the IEEE International Conference on Evolutionary Computation. IEEE, Dec 1995.Google Scholar