Passenger centred Transfer Coordination in disturbed Public Transport using Reinforcement Learning

Kurzfassung

Public transport has proven as an efficient and ecological solution to provide mobility to a large part of the society. Disturbances can impede public transport operation and reduce user acceptance. Current disturbance management strategies primary focus on minimising impacts on operational planning deeming passengers’ needs less important. As a development step towards passenger-centred intermodal disturbance management system, we present a minimal reinforcement learning example for transfer coordination. The environment builds up on the microscopic traffic sim- ulator SUMO. While we consider reinforcement learning as a valuable approach to find novel solutions to passenger- centred disturbance management in larger setups, the goal of this work is to demonstrate the applicability of this method. To this end, it is applied in a small public transport network, demonstrating comparable performance to that of a deterministic transfer coordination strategy.