Parallel agent-based modeling for improving traffic flow simulation

The growing complexity and dynamism of modern traffic systems, driven by connected vehicles, autonomous mobility, and logistics operations, have significantly increased the computational demands placed on traffic simulation tools. Traditional simulators such as SUMO (Simulation of Urban MObility), which rely on sequential execution, often fail to scale in large or real-time scenarios involving thousands of vehicles and extensive road networks. This paper addresses these limitations by reducing the runtime of SUMO simulations to enable faster and more responsive analysis. To achieve this, we introduce a task-level parallelization strategy that accelerates simulation execution without requiring any modification to SUMO’s core engine.The proposed framework parallelizes vehicle- and segment-level computations across multiple worker processes, allowing tasks such as emission estimation, dynamic routing, and agent state evaluation to be executed concurrently. This parallel execution substantially reduces simulation latency and facilitates real-time analysis and optimization, enhancing the ability of traffic managers to make informed decisions. The framework is evaluated over two real-world scenarios, a medium-sized interurban corridor and a major highway network, achieving speedups of up to 4×compared to standard sequential execution. These results demonstrate that parallel computing plays a crucial role in enabling scalable, responsive, and realistic traffic simulations, especially in the context of connected and autonomous vehicles, where large amounts of data must be analyzed and processed.