RACER: Real-Time Adaptive Congestion-Aware Emergency Routing in Urban Vehicular Networks

Abstract

The smooth and efficient movement of emergency vehicles in congested areas has always been a problem in intelligent transportation systems (ITS). The conventional approach to routing, typically through static shortest path calculations, often fails to respond to dynamically changing traffic conditions, leading to avoidable delays in critical situations. In this article, we propose a new approach to routing, termed RACER (Real-time Adaptive Congestion-aware Emergency Routing), which dynamically responds to changing traffic conditions without requiring additional traffic-signal-control infrastructure, relying instead on congestion information obtained through standard vehicle-to-infrastructure (V2I) telemetry such as roadside units or cellular reporting, which we model in SUMO via its Traffic Control Interface (TraCI). This is implemented through a combination of a proactive multi-edge look-ahead approach, a congestion-aware cost function, and a controlled approach to rerouting, ensuring stability during navigation. The proposed approach is evaluated using the SUMO microscopic traffic simulator on two large-scale real urban road networks (Bhubaneswar and Visakhapatnam), across five source and destination pairs, three congestion levels (light, moderate, severe), and 20 random seeds, for a total of 1,200 controlled runs. Because emergency-vehicle travel times are heavily right-skewed, we report the median as the primary metric alongside the mean. RACER attains the lowest median travel time across all routes and congestion levels, improving on the strongest baseline in every route, and its median travel time remains essentially flat as congestion increases (372/395/385 s for light/moderate/severe), in contrast to static Dijkstra, which degrades sharply (993/1993/3321 s). Paired statistical testing confirms the improvements over all baselines are significant ( $p\lt 10^{-29}$ ), and measured wall-clock runtime confirms the method operates in real time. We further make explicit the vehicle-to-infrastructure communication architecture on which the method operates, and show that its bounded, cooldown-gated rerouting keeps the control-plane signaling overhead low (on average fewer than three route updates per trip), making congestion-aware routing feasible over capacity-limited vehicular networks. These results demonstrate the effectiveness of incorporating congestion awareness into routing decisions, leading to faster and more reliable emergency response in such congested areas.