Abstract
A model-based approach to the monitoring of assembly processes is presented. In assembly operations, efficient control strategies cannot be generated by single control laws because the task condition varies in the process. Robot manipulators must monitor the process state in order to modify task strategies depending on the state. In this paper, we develop a new method of generating procedures that identify the state of assembly processes. The procedure interprets displacement and force information acquired in the process in order to identify the process state. In particular, the system is capable of monitoring the contact state; at which surface or edge the assembly parts are contacting to each other.
We first analyze kinematic and static behavior of assembly parts at each contact state by applying the theory of polyhedral convex cones. The theory enables us to treat constraints by mechanical contact in a systematic manner. Identification procedures that discriminate contact states are formulated with use of the polyhedral convex cones. We then develop a new method of generating the procedures automatically based on geometric models of assembly parts. To reduce real-time computations, the procedures are simplified to a minimum set by using reduction rules of polyhedral convex cones.
