On the Path to Biointelligent Production: Developing a Control System for Biointelligent Logistics

Abstract

Increasing demand volatility, resource constraints, and failures of production resources increase the need for production logistics systems that can adapt without requiring a central scheduler to determine every operation assignment. However, the operational transfer of biological control mechanisms into executable production-logistics models remains insufficiently specified. This paper therefore develops, implements, and evaluates a bio-inspired control framework for decentralised and self-organising production logistics. Following the inspiration pathway of biological transformation, the biological mechanisms of self-organisation, stigmergic coordination, and checkpoint-based regulation are abstracted and transferred to production logistics control mechanisms. Production units are represented as autonomous agents with heterogeneous capabilities, operational parameters and emission rates, while orders and transport resources are modelled as dynamic simulation objects within a shared event-driven environment. Temporary task-specific structures are formed through state-dependent assignments between orders, their next required operations, and available and capable production units. Failure and repair events support the interruption and subsequent reassignment, thereby allowing the production logistics network to reconfigure in response to resource unavailability. The resulting agent-based model library and simulation module provide an executable representation of decentralised production logistics control and are applied to cooperative multi-agent reinforcement learning considering time- and emission-related objectives. In the investigated scenario, the learned policy outperforms the considered rule-based dispatching strategies in terms of team reward and achieves lower mean makespan and total CO₂ emissions across the investigated failure settings.