Probabilistic Reactive Disassembly Planning

Abstract

A paradigm for reactive disassembly planning is developed which combines prior information about the disassembly task encoded in a predictive plan as well as measurements collected during the process in order to select the current most promising disassembly operation. Planning relies on a probabilistic inference mechanism which employs a Bayesian Network. It integrates perception and action and uses the dependencies among various parts of a product to propagate uncertainty regarding their condition as sensed during the disassembly process. Our methodology allows a computerized assessment of an End-Of-Life value of a product which will enable manufacturers to combine design principles that take into account environmental considerations at minimum cost during the design cycle. The approach is demonstrated by a case study.