In this paper, we explore the potential of multi-agent reinforcement learning (MARL) for managing the driving behavior of autonomous guided vehicles (AGVs) in production logistics environments with single-lane tracks, where deadlocks pose a significant challenge. We build upon previous work and adopt a MARL approach using the Proximal Policy Optimization (PPO) algorithm. We conduct a thorough hyperparameter search and investigate the impact of varying numbers of agents on the performance of the AGVs. Our results demonstrate the effectiveness of the MARL approach in addressing deadlocks and coordinating AGV behavior, as well as the scalability of the learned policy to different numbers of agents. The Bayesian optimization process and increased iteration count contribute to improved performance and more stable learning curves.

Decision-making in supply chains is challenged by high complexity, a combination of continuous and discrete processes, integrated and interdependent operations, dynamics, and adaptability. The rapidly increasing data availability, computing power and intelligent algorithms unveil new potentials in adaptive data-driven decision-making. Reinforcement Learning, a class of machine learning algorithms, is one of the data-driven methods. This semi-systematic literature review explores the current state of the art of reinforcement learning in supply chain management (SCM) and proposes a classification framework. The framework classifies academic papers based on supply chain drivers, algorithms, data sources, and industrial sectors. The conducted review revealed a few critical insights. First, the classic Q-learning algorithm is still the most popular one. Second, inventory management is the most common application of reinforcement learning in supply chains, as it is a pivotal element of supply chain synchronisation. Last, most reviewed papers address toy-like SCM problems driven by artificial data. Therefore, shifting to industry-scale problems will be a crucial challenge in the next years. If this shift is successful, the vision of data-driven decision-making in real-time could become a reality.

Automated guided vehicles (AGVs) form a large and important part of logistic systems to improve productivity and reduce costs. When multiple AGVs are running in limited and uncertain environments, lots of issues can occur, such as collisions and deadlocks, which need to be addressed. This paper presents a flexible simulation model for a warehouse with various AGVs. We implemented all three typical strategies to handle deadlocks (prevention, avoidance and detection and resolution). The results show that there is no dominant strategy and that the results strongly depend on the individual case and the input parameters.