Bio-Inspired Multi-Agent Manufacturing Control Systems with Social Behaviour (Biologisch geïnspireerd multi-agent fabrieksbesturing met sociaal gedrag)
Bio-Inspired Multi-Agent Manufacturing Control Systems with Social Behaviour
Hadeli, .; M0120817
Manufacturing nowadays is facing two major challenges. The firstchallenge is how a manufacturing system can cope with current andfuture trends such as more customised products, shorter product lifecycles, quality concern, etc. The second challenge is how amanufacturing system can address changes and disturbances thatpersistently happen on the shop floor, for example machine failure, rush orders, late arrival of materials, etc.To address these challenges, manufacturing systems need to consistentlyadapt themselves regarding their technology, products, organisationalstructures, etc. This creates the need for novel manufacturing controlsystems that are robust in handling such requirements both effectively and efficiently.The research result that is presented in this dissertation is meantto answer these challenges. This dissertation describes and discussestwo topics. Firstly, it describes bio-inspired multi-agentmanufacturing control systems that answer the need for robustmanufacturing control. The bio-inspired multi-agent manufacturingcontrol system is based on multi-agent technology. Agents in thissystem are structured according to the Product-Resource-Order-Staff(PROSA) reference architecture. Furthermore, the general coordinationbetween agents is based on the Ant-Colony Engineering mechanism. Ant-ColonyEngineering is a coordination mechanism that is inspired by the wayants in nature forage for food. Food foraging ants provide aninspiration to spread information and make global information availablelocally.A second issue in the bio-inspired multi-agent manufacturingcontrol system is the system nervousness. The previously discussedmanufacturing control system is able to generate short-term forecaststhat predict both resource loads and order routings. These forecastsbecome known throughout the multi-agent system with some time delay. Ifthe agents make their decisions based on these forecasts, propermeasures need to be taken into account for these delays, especiallywhen disturbances (rush orders, machine breakdowns) occur. If agentsreact too eagerly and swiftly, the forecasts become unreliable. Thisdissertation clarifies this issue and describes the measures in thecontrol system design that can be used to address the problem. Bytaking the nervousness issue into account, the agents behave in asocially acceptable manner that reconciles adaptation to changedcircumstances with predictability.