Exploiting graph-theorectic tools for matching and partitioning of agent population in an agent-based model for traffic and transportation applications

Abstract

In this position paper, we exploit the tools from the realm of graph theory to matching and portioning problems of agent population in an agent-based model for traffic and transportation applications. We take the agent-based carpooling application as an example scenario. The first problem is matching, which concerns finding the optimal pairing among agents. The second problem is partitioning, which is crucial for achieving scalability and for other problems that can be parallelized by separating the passenger population to subpopulations such that the interaction between different sub-populations is minimal. Since in real-life applications the agent population, as well as their preferences, very often change, we also discuss incremental solutions to these problems.