Urban Traffic Big Data Intelligent Scheduling Model Combining Optimal Transport Theory and Sinkhorn Algorithm

Abstract

In urban traffic big data scheduling, this paper proposes an intelligent scheduling model combining Optimal Transport theory and the Sinkhorn algorithm to address insufficient flow optimization, with consideration of traffic resource allocation in connected transportation environments supported by wireless sensing, road-side communication, and electromagnetic information infrastructure. In the specific implementation, starting from the OD matrix, traffic demand is modeled as a probability distribution, and the optimal transport plan for traffic allocation is obtained by solving the Wasserstein distance. Then, entropy regularization is introduced, and the Sinkhorn algorithm is employed to obtain a fast-converging transport solution under large-scale data conditions. Next, a real-time traffic data input mechanism is integrated to dynamically update the transportation cost matrix and recalculate the scheduling plan. Finally, a multilevel scheduling strategy is applied based on global optimization, and different priorities and local constraints are set in combination with regional functions to improve the model’s adaptability to complex traffic scenarios. Experimental results show that the scheduling time of the proposed method under high traffic density is 132.5 ms ± 5.6 ms, the CPU utilization rate is 52.9% ± 3.1%, and the model has high scheduling efficiency. Under the condition of more than 2,100 vehicles/hour, the traffic flow balance is maintained at 0.7, and the congestion index is only 1.32, indicating an obvious traffic optimization effect.