Agent-based simulations of urban mobility typically model how people choose travel modes using discrete choice models, machine learning fitted on structured travel survey data, or rule-based systems. These methods have known limitations in capturing the adaptive, context-dependent behaviour of real travellers.
The new preprint, "Towards Adaptive and Real-Survey-Aligned Simulated Urban Mobility Behaviour of Generative Agents," outlines a direction for using generative agents instead. The goal is to make simulated agents both adaptive and aligned with real survey data, potentially addressing the inflexibility of existing approaches.
The significance lies in the potential for more realistic simulations of multimodal urban mobility, which could support better planning and policy decisions. However, the available abstract only states the problem and the proposed direction; details of the method and validation are not yet described in the source.