Design and Implementation of an Optimization System for Networked Rescue Mobility in Urban Traffic

Abstract

Efficient and safe emergency response is a critical requirement in urban mobility systems. Emergency response drivers have to strike a balance between arriving at the emergency site quickly and avoiding dangerous traffic situations. This thesis addresses the challenge of forming timely and secure emergency corridors in urban environments by using Vehicle-to-Everything (V2X) technology. As part of the Gaia-X4 Advanced Mobility Services initiative, the RGS (Rettungsgassenservice, Emergency Corridor Service) controller computes optimal paths for emergency vehicles and communicates them to other road users. The thesis formulates the corridor generation as a dynamic simulation optimization problem, where the objective is to minimize emergency vehicle response time while ensuring safety and stability of traffic behavior. The controller uses the SUMO traffic simulator to predict the outcomes of traffic situations. It evaluates three optimization strategies: the Downhill-Simplex algorithm, a memetic algorithm, and a deep reinforcement learning agent, in both synthetic and real-world traffic scenarios. Additionally, the influence of non-cooperative vehicles on optimization performance is studied. Results demonstrate the ability of the controller to support emergency corridor formation in various traffic situations. While the reinforcement learning agent most effectively solves simple scenarios, the memetic algorithm excels at the most complex real-world scenario in the evaluation suite.