Unknown Object Control by Multiple-Arm Robot Using Fuzzy Neural Network

Manipulating an object is one of the most important tasks of robots. Many studies have been carried out on object control algorithms and optimizing control force. Most of those studies, however, deal with objects whose properties are previously known. In this paper, a control method for unknown objects is proposed using a fuzzy neural network. The controller consists of a main controller for trajectory control, and subcontrollers for manipulator force control, of an unknown object. Since manipulators have to apply specified force to an unknown object, the environment estimator is set in the subcontroller in order to avoid large overshooting which might damage the object. Computer simulation was performed to evaluate proposed controller.