Given the limited cross-border collaboration, Germany and the Netherlands are unable to fully utilize the capacities of their EMS systems in their border regions. To address this challenge, our study tackles cross-border EMS in the region by creating and solving strategic, tactical, and operational-level problems. At the strategic level, we develop a stochastic programming model to identify the optimal ambulance bases under uncertainty. The model integrates demand forecasts developed using machine learning methods while incorporating fairness, legal cross-border agreements, and hospital capacity constraints. At the tactical level, we use Reinforcement Learning (RL) and rule-based heuristics to have real-time redeployment. RL agents obtain adaptive redeployment policies, whereas heuristics offer efficient benchmarks for repositioning in time-sensitive situations. The RL algorithm uses predicted demand patterns as input to improve proactive relocation decisions. At the operational level, we aim to optimize the dispatching decisions for incoming calls by implementing Bayesian Neural Networks (BNNs) in the RL algorithms, which incorporate variability in demand, travel durations, and hospital capacity. The outcome can lead to improved response times and fair job distribution in cross-border collaborations. This model combines proactive optimization, heuristic flexibility, and adaptive learning to provide a framework for fast, fair, and sustainable cross-border EMS collaboration.
