An improved framework for large-scale multi-agent simulations of travel behavior

We describe a framework for running large-scale multi-agent sim-ulations of travel behavior. The framework represents each trav-eler as an individual “Agent ” that makes independent decisions about its desired use of the transportation system during a typical day. An Agent keeps a record of its decisions in a “Plan. ” A Plan contains the Agent’s schedule of activities it wants to perform dur-ing the day, including times and locations, along with the travel modes and routes it intends to utilize to travel between activities. An Agent database gives every Agent a memory where it can store several possible Plans, as well as performance information it uses to compare how well different Plans meet its needs. Agents score a Plan’s performance based on the output of the micro-simulator. The Agent database also allows Agents to periodically generate new Plans by connecting them to behavioral Modules that model the different kinds of decisions that affect an Agent’s Plan. For example, one Module chooses routes, another chooses activity du-rations. This paper describes the design and our current imple-mentation of this framework, plus the results of some verification scenarios.