MSF-PSO: A Multi-Strategy Particle Swarm Optimization Framework for Dedicated Highway Traffic Control of Small Passenger Vehicles

Abstract

This work aims to improve the passage efficiency of small passenger vehicles in highway environments. It proposes a Multi-Strategy Fusion Particle Swarm Optimization (MSF-PSO) algorithm to optimize travel time, lane utilization, lane-change frequency, and driving stability. The model architecture utilizes adaptive inertia weight adjustment to balance global and local search capabilities while implementing dynamic learning factor optimization to enhance individual and swarm learning capacities of particles. Also, it incorporates K-means clustering for population diversity maintenance and employs a Cauchy disturbance mechanism to facilitate particle escape from local optima. These components work synergistically to enable the algorithm to demonstrate faster convergence speed and superior global search capability in complex traffic environments. The algorithm undergoes validation on the SUMO traffic simulation platform using high-precision trajectory data from the HighD dataset. Experimental results demonstrate that compared to baseline algorithms, including Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Adaptive Particle Swarm Optimization (APSO), MSF-PSO significantly improves traffic efficiency and driving stability. Specifically, MSF-PSO reduces the average travel time to 243.7 seconds after 100 iterations, demonstrating remarkable improvements over baseline algorithms. This performance represents reductions of 8.6% compared to Haris & Nam (266.6 seconds), 11.4% versus APSO (275.4 seconds), 18.5% relative to PSO (299.2 seconds), and 19.8% compared to GA (303.8 seconds). Additionally, MSF-PSO achieves higher lane utilization (88.8%), lower lane-changing frequency (2.3 times/vehicle), and reduced speed fluctuation variance (7.9 km²/h²). Therefore, this work provides support for optimization in intelligent transportation.