Multi-agent simulations are key technologies to know the behavior of complex systems. Each agent in the multi-agent simulations is designed with probabilistic actions to other agents and environment. Hence, the deterministic assessment of complex systems with multi-agent simulations is difficult. Recently, uncertainty quantification attracts much attention to quantitatively assess the uncertainty of simulation results. In this paper, uncertainty quantification of multi-agent based traffic simulation is conducted and its characteristics are discussed. We conducted a series of traffic simulations in Okayama City based on the actually performed traffic census and investigate the relationship between the number of trials of traffic simulations and their reliability.

View full abstract