ABSTRACT
A simple yet elegant approach to a visual servoing control scheme has been presented in this chapter. The control action has been learned as a function of the error in the visual space using a T-S fuzzy framework. Thus the parametric space of the controller consists of locally valid PD gains in a distributed fashion. The initial values of these parameters are learned while mimicking the model based controller in the first phase. An online adaptation scheme has been proposed that can fine-tune the controller parameters to compensate the uncertainities associated with the system model and environment. The conceptual novelty in this proposed scheme is that the manipulator can be controlled without having to compute its own inverse Jacobians. Thus the model reflects a more cognitive learning architecture just like a child learns to actuate his/her hands and legs without the need for understanding the complexities of the involved kinematics. The proposed scheme has been validated through exhaustive experimen tation on a 7 DOF robot manipulator. The controller input-output data have been used to learn the initial parameters of the distributed fuzzy PD controller which has been termed as a non-adaptive fuzzy controller in this chapter. These controller parameters have been fine tuned using the proposed adaptation scheme. It has been shown that the controller is effective in visual servoing for both static and moving targets.
