HLSR: Hybrid Live Forecast Selective Dynamic Vehicle Rerouting for Realistic Urban Traffic

Summary

Researchers Xiao Wang, Shun Ren Yang, and Hui Nien Hung propose HLSR, a selective hybrid live-forecast vehicle rerouting framework designed to reduce urban traffic congestion more realistically than network-wide rerouting approaches. Traditional live travel-time shortest-path rerouting assumes nearly every on-road vehicle is replanned every decision period, which is unrealistic in practice. HLSR instead operates under a limited intervention scope, fusing live edge speeds with short-horizon forecasts. The framework combines dual-threshold congestion detection, calibrated upstream selection, and driver-tailored travel-time prediction. It further introduces approaching-vehicle expansion, travel-time-weighted k-shortest-path generation, and a horizon-dependent hybrid live-forecast segment speed used for multi-cost route allocation. By selectively intervening only where and when needed and blending current observed speeds with forecasts at different horizons, HLSR aims to balance effectiveness with realistic deployment constraints. The paper is available on arXiv as 2608.18056 and falls within the machine learning research domain.