Multi-Level Optimization for a Demand-Responsive Connector

Local public transport is a more sustainable alternative to motorized private transport and can contribute both to reducing the traffic volume and to reducing emissions. However, in order to attract more users, public transport needs to become more attractive, especially during off-peak hours and in city outskirts. Currently, there is often a lack of flexibility due to low frequency and a lack of connections. This can be remedied by so-called flexible transit services, which combine conventional fixed-route transit with demand-responsive transit.

This dissertation addresses the anticipatory planning of a demand-responsive connector, where fixed-route and demand-responsive transit are connected via fixed transfer points. We aim to combine the tactical and operational planning of such a system in a novel way using mathematical optimization models. In a first step, robust solutions are generated on a tactical level based on a static stochastic model. In a second step, dynamic stochastic planning is performed on the operational level. Finally, the tactical plans are integrated into the planning at the operational level using reinforcement learning, resulting in a multi-level optimization.

Researcher: Alexander Bosse