Position: Collaboration Between the City and the Machine Learning Community is Crucial to Efficient Autonomous Vehicles Routing

Autonomous vehicles (AVs) are operating on public roads in several cities. Assuming they use Multi-Agent Reinforcement Learning (MARL) for simultaneous route optimization, higher AV penetration rates may degrade traffic networks’ system-wide performance. We study AV routing decisions in a traffic environment shared with human drivers. Our experiments with standard MARL algorithms reveal that, both in simplified and complex networks, policies often fail to converge to an optimal solution or require long training iterations. This convergence issue is amplified by the fact that we cannot rely entirely on simulated training, as there are no accurate models of human routing behavior. In addition, real-world training in cities risks destabilizing urban traffic systems, increasing externalities, such as $CO_2$ emissions, and introducing non-stationarity as human drivers will adapt unpredictably to AV behaviors. **In this position paper, we argue that city authorities must collaborate with the ML community to monitor and critically evaluate the routing algorithms proposed by car companies, ensuring fair, system-efficient algorithms that maintain, or even improve, the performance of urban traffic networks.**