Large-Scale Dynamic Adaptive Traffic Lights Scheduling Algorithm Based on Genetic Programming

Intersections are the basic structure of traffic networks and the main locations of traffic congestion in cities. How to maximize the throughput of intersections and prevent vehicle spillback has always been a challenge in traffic control research. This paper establishes a traffic flow attribute statistical model based on roads, proposes a traffic light control strategy based on genetic programming algorithms, and establishes a relationship link between statistical data and traffic lights through decision trees. The optimization goals are to reduce average travel time, reduce average waiting time, and increase historical average speed, achieving intelligent traffic lights scheduling algorithm that can adapt in real time to dynamic traffic conditions. Under large-scale traffic flow conditions on the Singapore map, using actuated signals as the baseline, the coordinated passage rate increased by 115.05%, average travel time decreased by 33.61%, historical average speed increased by 163.78%, average number of stops decreased by 57.48%, and average waiting time decreased by 50.47%.