This paper presents a live coupling between the agent-based travel demand and simulation modeling frameworks mobiTopp and MATSim to better capture feedbacks between on-demand ride-pooling services and travel demand. The approach embeds MATSim's demand-responsive transit module into mobiTopp's short-term simulation, so that each mode choice decision can respond to the contemporaneous fleet state, including vehicle availability, waiting times, detours and system-side rejections. This enables a consistent distinction between supply-side rejections due to limited capacity and passenger-side rejections of unattractive offers. We implement the coupling for a ride-pooling service in a weekly model of Hamburg, Germany, with about 187,000 agents and 3.8 million trips, and compare multiple scenarios that vary fleet size and vehicle capacity, as well as an uncoupled benchmark in which ride-pooling is always available with static service attributes. The results show that larger fleets increase the number of served trips, reduce waiting times and raise pooling rates, but also exhibit diminishing returns and lower vehicle utilisation at very high fleet sizes. We conclude that live coupling of demand and assignment models substantially improves the behavioural realism of ride-pooling simulations and supports more robust policy analysis, while highlighting the possibilities of using this approach for other applications. Furthermore, for applications with very large fleets there is a need for more selective request generation and computational enhancements.
Wilkes et al. (2026) studied this question.