IEEE Power Engineering and Automation Conference
This paper describes a new practical approach for approximating the inverse kinematics of a manipulator using an RBFN (Radial Basis Function Network). This neural network with its inherent learning ability can be an effective alternative solution for the inverse kinematics problem where traditional methods are impractical because the manipulator geometry cannot be easily determined, e.g. in a robot-vision system. However, sometimes a well-trained network cannot work effectively in the operational phase because the initial network training occurs in an environment that is not exactly the same as the environment where the system is actually deployed. In this paper, an on-line retraining solution using the Delta rule is presented for systems whose characteristics change due to environmental variations. Moreover, a “free interference rule” is also suggested to avoid learning interference where the training effect of a current training point may upset some of the weights which were trained with previous points. To verify the performance of the proposed approach, a practical experiment has been performed using a Mitsubishi PA10-6CE manipulator observed by a webcam. All application programmes, such as robot servo control, neural network, and image processing tool, were written in C/C++ and run in a real robotic system. The experimental results prove that the proposed approach is effective.
