Integrating Autonomous Vehicles and Public Transit

There are three broad strategies to integrate autonomous vehicles (AVs) with public transport:

1. Expand first-mile access to transit
2. Replace low-productivity bus routes
3. Re-optimize Transit network

I use Singapore as the case study of the first strategy. To get this right, all three conditions must be met: Better service for passengers; No increase in road congestion ; A viable business model for AVs.

The Context: First-Mile Frictions in Singapore

Singapore—a city-state with one of the world’s most efficient public transport systems. The MRT and LRT systems are dense, fast, and well-used.

How can AVs help solve the first-mile problem—getting people from their homes to transit stations efficiently, affordably, and comfortably?

The map below shows MRT ridership and first-mile access patterns during the morning peak hours (7–9 am), revealing the crucial role of buses in connecting commuters to the train network.

Tampines: 52% rely on buses ; Woodlands: 55%; Bedok: 61%

A pressing need to strengthen first-mile connectivity.

Tampines

Tampines is a high-density residential area. We conducted a detailed mapping of bus boardings and alightings around the MRT station during peak hours, giving us a granular view of travel behavior.

Not all routes are created equal.

A Hybrid Strategy to Augment—Not Replace—Buses

In Tampines, we proposed a hybrid strategy:

• Preserve the 16 busiest bus routes that serve 90% of demand
• Repurpose the remaining 11 low-ridership routes (10% of demand) with on-demand AV services

We didn’t eliminate those 11 routes. Instead, we strategically rerouted them—allowing buses to focus on high-demand corridors, while AVs serve scattered, low-density areas.

Routes like Bus 3_2 and Bus 21_1 were used as examples—showing how small redesigns can unlock AV integration.

The Simulation: Behind the Scenes

We built an agent-based simulation to model three key actors:

• Passengers choosing travel modes based on time and cost
• AV operators managing fleet size, pricing, and dispatching
• Bus agencies adjusting routes and schedules

We measured three outcomes:

1. Out-of-Vehicle Time (OVT) Walking + waiting time—ensuring service quality for passengers
2. Vehicle Kilometers Traveled (VKT) A proxy for traffic impact—ensuring the system doesn’t worsen congestion
3. Profitability Making sure AV operators can survive financially

The Narrow Window: Triple Win Requires Precision

To make this work, we must hit all three goals:

• Better service for passengers
• No increase in road congestion
• Viable business model for AVs

The window for an integrated AV-PT system exists – but is narrow.

Integration itself does not guarantee success. It requires intentional design.

[P.S. The narrowness may reflect that the fact that the existing bus network in Singapore is efficient to start with. ]

The Future: Trains, Buses, and AVs—Each Doing What They Do Best

Imagine a system where: Trains serve trunk corridors. Buses serve high-density feeders. AVs fill in the rest—on-demand, shared, and responsive.

— Jinhua

Methodology Paper: Integrating Shared Autonomous Vehicle in Public Transportation System: A Supply-Side Simulation of the First-Mile Service in Singapore, Transportation Research Part A (full paper)