Bridging the Gap Between Semantics and Control for Industry 4.0 and Autonomous Production

Small lot size production requires decoupled specification of the product and the production system to allow for flexible manufacturing, which leads to autonomous systems. Their key are algorithms, that are able to sequence and parametrize skills in a meaningful way such that a given task is fulfilled. This planning process requires suitable models of all relevant aspects of the production which includes the system's state and estimates of its behavior. In this paper we propose formal models for declarative and procedural knowledge, which unify the symbolic and sub-symbolic representations. Currently, the curse of dimensionality prevents state of the art techniques to solve non-trivial real-world tasks that pose mixed discrete and continuous states. Our models are designed hierarchically from the very beginning and thus allow an automatic decomposition of planning problems. This allows us to compute solutions online for complex tasks that are only defined by the goal configuration of the product. We validate our approach with an assembly use-case that illustrates end-to-end autonomous robotic production in a real-world setting. The separate definition of product and hardware, as well as the number of steps required to fulfill this task demands a multilayered planning and execution for online control.