Research on intelligent dispatching strategies for cluster-type urban traffic from a data-driven perspective

Abstract

In 2024, the Shenzhen-Zhongshan Corridor officially opened, further enhancing Zhongshan City's status as a hub city for integrated transportation in the west of the Pearl River Delta. This growth, however, has put strain on the existing road network as well. Traditional traffic control models using empirical judgment and fixed signal timing are unable to effectively solve the 'supply-demand contradiction' caused by the superposition of cross-regional tidal traffic and through traffic in the city, especially in light of its usual cluster layout in space. Based on the above, this paper tries to build a digital intelligent dispatching system for important traffic congestion and cross-cluster bus route system with Zhongshan City from the perspective of multi-source data fusion. The first part of the research involves the creation of a basic dataset of internet traffic data, along with simulated traffic flow.The preliminary step in the research involves the development of a basic dataset from internet traffic data that is cleaned, along with simulated traffic flow. Then, to handle the nonlinear problems in the short-term prediction problem, a Long Short-Term Memory (LSTM) network is presented, which is further connected with a deep reinforcement learning algorithm for signal timing optimization to realize dynamic optimization of short-term traffic flow prediction. Analytic experiments are implemented in the Shalang section of NH-1 and bus line K01, and the results show that, with the help of the improved strategy, the average travel speed of the improved strategy in the morning peak period exceeds that of the traditional strategy by 23.2% at the key nodes, and it performs well at guaranteeing the bus load factors from the drastic fluctuation. The conclusions of this research can serve as theoretical guidance and practical reference to help guide the traffic digitalization of medium-sized cities.