A Novel Multi-agent Evolutionary Algorithm for Assembly Sequence Planning

Many evolutionary algorithms for assembly sequence planning (ASP) have been researched. But those algorithms have lots of blind searching because individuals have little consideration about geometry and assembly process information of product in the evolutionary process. To improve individuals' intelligence and decrease blind searching, motivated by the self-assembly computing and multi-agent evolutionary algorithm, a novel multi-agent evolutionary algorithm for assembly sequence planning (NMAEA-ASP) is presented. In the algorithm, learning, competition and mutation are designed for each agent. Learning, competition and mutation are realized by assembly and disassembly. Some notions such as assemblyunit, power about assembly are also introduced. Experimental results show that NMAEA-ASP can find an approximate solution faster than other evolutionary algorithms.