Emergent effects in multi-agent simulations of road pricing

Road pricing is debated as an option of transportation policy. Especially in metropolitan areas congestion pricing is promising to reduce congestion and to protect the environment. In order to reach the promised results the choice and design of a policy is very important, especially in a ”second-best ” context. Therefore it is worth to attempt detailed predictions of the effects and implications of the planned pricing scheme. Most if not all state-of-the-practice methodologies forecasting those effects are • aggregate and in consequence do not consider social and economic characteristics of individual travelers. • static in time and in consequence do not consider temporal effects such as toll avoidance In order to bridge this gap, multi-agent microsimulations can be used. Our large-scale multi-agent traffic simulation is capable to simulate a complete day-plan of up to seven million individuals (agents). In contrast to other approaches, our simulation truly traces the synthetic travelers through their day, thus enabling us (at least in principle) to model emergent effects such as complex re-scheduling across the whole day. This paper describes the implementation of a toll-scheme for the bigger Zurich area and presents the results of the simulation. We point out how agents (population) react to changed prices of transportation by modifying their consumption patterns. The analysis of the policy is based on the performance of simulated day-plans of the agents. This performance is directly given by a utility function, which is used to measure gains and losses of different groups of inhabitants in the research area. Based on these measurements we provide an economic interpretation of the policy and highlight emergent phenomena like changes in route choice and time reactions