Neema Nassir, Andrés Fielbaum, Sapan Tiwari, Avishai (Avi) Ceder
Abstract
Modular Autonomous Vehicles (MAVs) enable public transport services in which small pods couple into platoons on trunk segments and decouple to serve stops or diverging branches. This supports in-motion transfers (IMT) between intersecting platoons while passengers remain onboard, creating transferless connectivity in many-to-many origin-destination networks. Such flexibility can deliver substantial passenger benefits by combining high-frequency trunk performance with expanded coverage on branches. However, it also introduces operational challenges not present in conventional systems, including the need to coordinate pod exchanges at junctions, model decouple-dwell-recouple behaviour at stops, and maintain feasible platoon sizes across space-time.We develop a network flow mathematical formulation and network transformation for vehicle scheduling and fleet sizing of MAV network services. We adopt a discrete time-space representation and construct a time-expanded network with two new transformations: (i) junction-coupling links that synchronise left, through, and right pod swaps when concurrent platoons meet, and (ii) stop transformations that represent decouple-dwell-recouple movements enabling stop-less operations while the through-platoon continues uninterrupted. Embedding these transformations in the time-expanded network allows all operational requirements to be represented as link flows with integer bounds. The resulting optimisation model is a minimum-cost circulation problem that minimises fleet size (primary objective) and non-revenue movements (secondary objective), subject to timetable feasibility, platoon-size bounds, and depot and junction capacities. Repositioning is represented explicitly through deadhead links with non-zero costs. The proposed formulation is linear, totally-unimodular (due to integer constraints), and highly scalable.The proposed framework supports evaluation of capital and operating requirements for alternative service designs. We demonstrate its effectiveness on a small public-transport network with modular vehicles. We extend the case study network in Khan and Menéndez, 2024 to accommodate asymmetric service, where different routes or directions operate at unequal frequencies. This generalisation enables the evaluation of heterogeneous corridor patterns and directional peak services. Through a set of case study scenarios, we demonstrate how the model quantifies service performance and cost-effectiveness metrics, including revenue hours (or kilometres) per vehicle hours (kilometres), and the resulting fleet requirements under alternative asymmetric service patterns.