Increasing Robustness of Agents’ Decision-Making in Production Automation using Sanctioning
Abstract
Industry 4.0 requires high reconfigurability and flexibility of cyber-physical production systems (CPPS). Agent-based approaches are introduced to realize decentralized decision-making and flexibility within CPPS. Agents negotiate with each other regarding task allocation in production systems to achieve a global goal together. In non-deterministic systems, agents’ decision-making can become inaccurate due to misalignment between the agents’ beliefs and the actual state of the physical system they represent. (Un)Intentional misestimations can lead to non-optimal task allocation regarding the global system’s goal. Additionally, decisions that benefit individual agents’ goals, such as ‘get all tasks’, can contradict the global system’s goal. In this paper, a sanctioning approach known from socio-technical systems is integrated into a non-deterministic production plant consisting of a process part and a logistics part to increase the robustness of agents’ decision-making.