Abstract This paper examines the potential of autonomous vehicle (AV) technology for enhancing first and last mile services for a light-rail station. We use an event- and agent-based simulation model to compare the performance of fixed and demand-responsive routing services. The routing of on-demand services is based on a matching algorithm in which incoming passenger requests are prioritized and assigned to vehicles under capacity constraints. Our findings indicate that, for a high-frequency light-rail feeder system, fixed routing is the preferred option, even with the assumed reduction in operational costs due to driver-less operations. However, we also observe that demand-responsive services can be as effective as fixed routing in off-peak hours, provided the heuristics for matching passengers to vehicles are effective. This implies that a combination of the two services could be beneficial in certain contexts. In addition, our results demonstrate that urban sprawl has an impact on the performance of the system, with the demand-responsive services becoming relatively better when urban sprawl increases, while the fixed routing remains superior across most key-performance indicators. To assess the performance of the different services, we employ cost–benefit analysis.

Highlights • Use of agent-based simulation to model first and last mile autonomous feeder services for a light-rail system. • In a high-frequency light rail station feeder context fixed-routing is superior to demand responsive services across all scenarios. • On-demand services perform better in off-peak periods and are on par with fixed-routing if hard rejection policies are applied. • Increased urban sprawl led to better bus utilization for demand responsive-systems but flexible routing is still superior. • First and last mile planning is context specific and depends on demand density as well as the routing design.