Game-based Synthesis for Distributed Control of Industrial Assembly Lines

The manufacturing industry nowadays uses a high degree of automation that contributes significantly to its commercial success. Set-up of an automation system is usually very expensive and only pays off by amortization of costs over its life span. Shorter product life cycles and increasing personalization of products, which are the results of the so-called fourth industrial revolution, endanger this concept: it becomes more and more important that not only set-up, but also adaptation of automation systems to changing production requirements is cost-efficient. Modification of the control software plays an important role in this context, because programming of such systems often takes place at a low level and is hence restricted to experts who are aware of the consequences of their design decisions. This work presents a model-driven workflow that aims at reducing the costs of adapting the control software in industrial automation systems. Its goal is to replace the tedious and error-prone manual adaptation by an automatic approach which is based on synthesis. This has two major benefits: first, the possibility for non-experts to adapt the control software allows experts in industrial automation to focus on other tasks, reducing overall costs. Second, programming errors are avoided, because re-synthesis is triggered with the push of a button and the synthesized control programs are correct by construction. In order to achieve these benefits, this work introduces formalisms for describing automation systems and the production tasks to accomplish. The sequence of primitive control actions to execute is then automatically synthesized by a solver that is based on game theory. The result is transformed into executable code, which allows both offline simulation as well as online execution within the real automation system. For distributed control systems, the approach is capable of generating multiple control programs that cooperate in order to achieve the production goal. The presented workflow integrates nicely with existing concepts from industrial automation, such as programmable logic controller (PLC) programming languages as defined in IEC 61131-3. The contributions of this work are the definition of a suitable domain specific modeling language for precise description of automation systems and automation tasks, the transformation of the model into an input for the game-based solver as well as a software library for simulation and execution of the generated control programs on centralized and distributed target control units. In order to show the operability of the approach, the development workflow is applied to various industrial-grade automation systems. Programmable logic controllers, personal computers and microcontrollers are used as target platforms. For the scenarios examined in this work, synthesis time is in the range of seconds to minutes and hence acceptable in the target domain. Evaluation also includes experiments with cost-based optimization and parallel execution of individual production steps.