Toward Self-Organizing Production Logistics: A Multi-Agent Approach

TL;DR

This paper proposes a conceptual framework and initial multi-agent architecture for self-organizing production logistics to better handle variability and disturbances in circular production systems.

Contribution

It introduces the concept of SOPL, identifies key drivers, and presents a multi-agent architecture and demonstration roadmap for adaptive, distributed production logistics systems.

Findings

Initial laboratory demonstrator developed for disturbance handling

Multi-agent architecture combining embodied and non-embodied agents

Roadmap for progressive implementation of SOPL systems

Abstract

Production logistics (PL) is increasingly exposed to variability, dynamic interdependencies, and operational disturbances that challenge conventional centralized planning and control. These characteristics are particularly pronounced in circular production systems, but are increasingly relevant across PL more generally. This paper addresses this challenge through the concept of Self-Organizing Production Logistics (SOPL) using the Design Science Research Methodology (DSRM) as a structuring framework. The paper identifies key technological and systemic drivers motivating SOPL, including autonomous logistics resources, distributed AI-based decision-making, and increasing operational uncertainty in circular production. Based on these drivers, system-level objectives and design requirements for SOPL are derived. Building on these requirements, an initial multi-agent architecture is proposed that combines embodied and non-embodied agents, event-driven coordination, semantic knowledge structures, and digital twins. In addition, a three-phase demonstration roadmap is presented, ranging from an initial laboratory demonstrator toward increasingly distributed and adaptive SOPL systems. The Phase I demonstrator serves as an experimental setup for investigating disturbance handling, human involvement, and supervisory coordination in an order-driven kitting and supply scenario. Overall, the paper contributes a conceptual foundation for the design, implementation, and experimental evaluation of SOPL systems.