Revising bus routes to improve access for the transport disadvantaged: A reinforcement learning approach

Traditional bus route design tools mainly aim to minimise a cost-based objective such as passenger travel times and/or operator costs. These approaches favour high-demand areas in distributing transit services and benefit densely populated areas. Public transport must provide viable transport services to socially disadvantaged people to allow them to undertake essential social and economic activities. This study uses a transit accessibility measure in the route design objective and demonstrates its impacts and implications through a case study in Penrith, an area in Sydney, Australia. The results reveal a limitation of using accessibility as the sole design objective. An equity measure called the transit service need index is incorporated in the objective to revise and redistribute bus services to improve accessibility in areas with more transport disadvantaged people. Reinforcement learning is used for bus route optimisation. The route design tool could be useful for transit planners and operators in route planning and evaluation. The study demonstrates some different approaches of route design objective formulations to meet increasingly diverse needs and expectations for public transport.