Edge based distributed framework for real time hazard detection and road safety in smart transportation.

The main drawbacks of centralized smart-transportation pipelines are latency, bandwidth, and scalability limitations, which restrict the real-time detection and notification of hazards. Intended to develop and test a distributed edge structure capable of supporting low-latency, efficient hazard detection and disseminated alerts widely and quickly. We combine the edge-based IoT sensing (roadside units, in-vehicle devices, mobile crowdsensing) and local processing at the edge nodes. A combination of ensemble machine-learning (Random Forest, Gradient Boost) with Probabilistic Cellular Automata and Markov Decision Processes to predict hazards based on the speed, density, and environmental conditions of the traffic. An alert distribution layer (V2X, V2V, V2N) is one of the layers of the SUMO simulations, which benchmarks performance on centralized and scheduling benchmarks (RR, LC, FCFS, SJF, Random). The framework has a maximum hazard-identification precision of up to and including 95 percent when simulated, can reduce its alert latency to 0.2 to 0.3 seconds (compared to a baselines minimum of 0.8 to 1.5 seconds), consumes less energy, and balances its loads (lowest 0.04-0.7-sigma edge-node load) and achieves high throughput (27 to 30 tasks/s) with exceptionally good scalability and low drop rates. A V2X distributed edge architecture with V2X and hybrid ML/PCA/MDP analytics can provide accurate, low-latency hazard detection and alerts, and is superior to centralized methods, offering a practical foundation for safer and more resilient transportation in both urban and rural settings.