Dynamic machine-level planning for Plug & Produce manufacturing

Abstract Manufacturing industries face persistent challenges due to a shortage of skilled labor and growing demand for customized products, while established automation methods are not cost-effective for high-mix, low-volume manufacturing. An intelligent automation concept capable of daily reconfiguration to accommodate new manufacturing conditions is required. Manufacturers request digital configuration over detailed programming of operational sequences and process data. Sub-operations, such as material transport and resource switching, are preferably automatically planned. This paper introduces an ontology to facilitate the digital configuration of agent-defined transfer operations for materials, parts, and resources. Automatic planning has previously been applied to higher-level path planning, such as the flow of materials and products through buffers and workstations. This study extends these algorithms by incorporating recursive planning for strategic routing and real-time reconfiguration of parts, materials, and resources at the machine level. The automatic planners are incorporated and modeled in a configurable multi-agent execution system. The planning problem is decomposed into simpler subproblems via a recursive, automated problem-generation process, while the number of edges is reduced by a proposed transport-edge mechanism. Validations are conducted at a Plug & Produce robot station equipped with tools, toolholders, parts, and associated materials. Planning Domain Definition Language (PDDL) planners are compared with Satisfiability Modulo Theory (SMT) solvers, where the latter performed best for the automatic planning of robotic tool-change operations and material transfer in a Plug & Produce automated manufacturing context.