Crowdsourced Smartphone Sensing and AI-Driven Edge Computing for Real-Time Urban Traffic Incident Detection and Response

Abstract

Rapid urbanization and rising traffic density have increased the frequency and impact of roadway incidents, while conventional infrastructure-based monitoring systems remain constrained by limited coverage, high deployment cost, and delayed centralized processing. This paper presents a crowdsourced smartphone–edge framework for real-time urban traffic incident detection and response, in which commuter smartphones act as opportunistic sensing nodes and nearby edge units perform local corroboration before forwarding validated events to a central traffic management layer. The proposed framework combines on-device sensing using accelerometer and GPS cues, privacy-aware event abstraction, edge-based spatial-temporal clustering, and centralized response actions such as signal adaptation, rerouting support, and emergency notification. To evaluate the framework, a city-scale simulation environment was developed using SUMO and Python-based edge emulation over a representative urban road network with 25,000 virtual vehicles and 100 overlapping edge-service zones. Experimental analysis under multiple operating conditions, including peak-hour congestion, noisy sensing, varying participation rates, and simultaneous incidents, showed that the proposed framework reduced average incident detection latency from 40.6 s to 14.2 s, improved true positive detection from 72% to 92%, lowered average waiting time in impacted zones from 6.8 min to 4.7 min, and reduced network data usage by 48% relative to a baseline centralized system. These findings indicate that privacy-aware mobile crowdsensing, when coupled with edge-side validation, can provide a responsive, scalable, and communication-efficient basis for next-generation urban incident monitoring and traffic management.