A first-principles technical analysis of what actually works in vehicular routing at city scale.
"Conventional vehicular routing is broken. Build for the architecture that exists in 5 years, not the one that existed 5 years ago."
Modern cities regularly push arterial roads beyond 400–500 vehicles per kilometre. Traditional traffic management, designed for sparse and mostly unconnected fleets, collapses under this load. The result is familiar: gridlock, delayed safety messages, and fuel wasted at scale.
This work takes a first-principles view of routing in vehicular networks (VANETs): what fails, what scales, and what is realistically deployable in dense urban environments. The focus is on behaviour under real constraints—mobility, latency, density, and infrastructure—not on idealised simulations.
Routing choices are grouped into four families, each with a clear operating envelope:
| Protocol Family | Strengths | Weaknesses | Verdict |
|---|---|---|---|
| Topology-Based (AODV, DSDV) | Simple, well-studied | Control overhead, collapses > 50 veh/km | Legacy / fallback |
| Geographic (GPSR, GyTAR) | Low latency, scalable | Needs maps, struggles in sparse networks | Current best option |
| Broadcast / Geocast | Rapid safety dissemination | Broadcast storms if unmanaged | Safety-specific |
| Learning-Based (Deep RL, etc.) | Adapts to live conditions | Training, sim-to-real gap, compute cost | Emerging / near-term |
- Geographic, road-aware routing is the practical default for dense urban VANETs.
- Topology-based protocols belong at the edges: stable backhaul, RSU meshes, and controlled corridors.
- Broadcast and geocast must be engineered with storm-suppression; naïve flooding is untenable.
- Learning-based approaches are promising, but bounded today by sim-to-real gaps and embedded inference limits.
- Travel time reductions on coordinated “green wave” corridors in the range of 20–35%.
- Safety-critical alerts delivered in sub‑100 ms windows when routing is designed for latency first.
- City-scale freight and fleet optimisation with double‑digit percentage gains in utilisation and fuel efficiency.
- Communication: DSRC (IEEE 802.11p), C‑V2X, 5G NR‑V2X
- Network & simulation: VANET, SUMO, OMNeT++ / Veins
- Protocols: GPSR, AODV, GyTAR, VADD, geocast variants
- Intelligence: Deep reinforcement learning, federated learning for routing policy adaptation
A. Jeswin Karunya Benedict
Department of Computer Science and Engineering
Vellore Institute of Technology, Andhra Pradesh, India
