Abstract
A machine assembly / disassembly planning system based on agent cooperation has been developed. It can create an optimal assembly and / or disassembly sequence for any goal, such as disassembling / assembling all parts or replacing a broken-down part. The planning system is reactive, i.e., it can cope with unpredictable incidents, such as changing goals while the system is under operation. We define twenty six class agents for machine assembly / disassembly tasks and use an improved algorithm whose original one was proposed by Maes. The framework of the proposed planning system and detail definition of the planning algorithm are introduced and the results of experiments are discussed.
