SEMSim: A Distributed Architecture for Multi-scale Traffic Simulation

With the fast urbanization of our modern society, transportation systems in cities are facing increasing problems such as congestion, collisions, and high levels of emissions. Researchers have been searching for solutions by investigating better urban planning and transportation policies, introducing new technologies such as Intelligent Transportation System (ITS), or introducing more environmentally friendly vehicles such as electric vehicles (EVs). Traffic modeling and simulation is one tool adopted by researchers for more than half a century [1] to help authorities assess new infrastructure design, and new policies without impacting real traffic. City-scale nanoscopic traffic simulation is a challenging problem that requires parallelization and distribution. In this paper, we have given an overview of the architecture for our nanoscopic traffic simulator SEMSim. For efficient parallel simulation, reducing the dependencies between the various LPs is crucial. We have specified a multi-objective optimization problem concerned with the allocation of agents to clusters. In our future work, we will implement a nanoscopic traffic simulation and devise methods to solve this problem dynamically. Given the difficulty of the problem, these methods will have to make use of domain-specific knowledge, such as information about the topology of the road network.