The Minibus Way: How Autonomous Vehicles can Replace Trains

Autonomous vehicles (AVs) stand to revolutionize transportation and the structures of cities. They will make better use of road infrastructure, reduce collisions, facilitate longer-distance commuting, and reduce the need for parking. They will also likely draw riders away from traditional transit systems, which have been slow to introduce driverless technology and cover a more limited range of routes and destinations than AV networks can. However, AV technology offers cities the opportunity to reshape their transportation networks by replacing their rail lines and subways with corridors for autonomous minibuses, corridors I dub “minibusways.”

The claim that autonomous vehicles could replace subways, buses, and rail lines has been the subject of heated debate. However, this debate has often revolved around the assumption that these autonomous vehicles would be personal cars with one or few occupants. It is clear that driverless cars would not be able to deliver enough peak capacity to carry as many riders as a subway line; it’s geometrically impossible. 3,900 autonomous vehicles per lane per hour with 1.5 people per vehicle is 5,850 people. Those 5,850 people could fit into four New York City Subway trains with room to spare.

The solution lies in shared autonomous minibuses, which offer the flexibility of AVs and the high capacity of trains. On busy lines, the interior of these minibuses would resemble a subway car, maximizing the number of people who can fit inside, but they’d provide much better service than trains can. Replacing rail lines and subways with minibusways would lead to greater flexibility, better service, higher capacity, faster journeys, and lower costs.

Easymile Driverless Bus
Minibusways would feature a lot of vehicles that look like this one, though not always this exact size. They’d also have much higher top speeds (60 mph+).

This piece will discuss how minibusways would work in a world where AV technology is  advanced and reliable. At present, there is still a ways to go, but ruminating on how we could best use this technology to improve cities can help us prepare for the future as well as understand current transportation challenges.

As I will detail below, minibusways would work best when minibuses can pass each other at stations, but they do produce a number of benefits regardless. In this piece, I’ll look at scenarios where there are two passing lanes, one passing lane, and no passing lanes. For a bit more on why I chose these layouts, see Notes and Appendices.

1. Minibusways with Two Passing Lanes at Stations

The Basics

Whenever possible, minibusways should have separate stopping and passing lanes at stations for traffic in each direction. This would allow for much quicker journeys and dramatically higher throughput because it lets minibuses pass each other. The basic idea is that the through lanes can function like highway lanes and sustain a consistent flow of vehicles at high speed.

mbw 2 passing lanes
Minibusway station with two passing lanes

Selecting a Destination 

To use a minibusway, a traveler would use a mobile app or a touchscreen kiosk at the minibus station to select their destination before they get on a minibus. The system would direct the traveler onto a specific minibus based on the destination they selected. This concept is used in some transportation systems today, such as West Virginia University’s Morgantown Personal Rapid Transit, which is detailed below in a video by Tom Scott. It allows people with the same destination to cluster into the same vehicle, which would then skip most or all other stops on its way to that destination. (Three key differences between PRT and the minibusway concept are that PRT vehicles move much slower, are typically smaller, and work only on special guideways, as opposed to just normal paved roads.)





When someone inputs their destination, their screen would then indicate which minibus to board. The screen might say, “Board minibus B3, which is due to arrive in 2 minutes,” and the traveler would then wait for a minibus displaying “B3” on an LED screen to arrive at the platform. This is similar to to the “destination dispatch” system used for elevators in some buildings:



As long as there is a decent level of demand on a minibusway, each vehicle will function as an express service, only having to stop at a few stops over the course of an entire line. There would not be separate “express” and “local” stations as there are on four-track subways and train lines; every station would have express service, i.e. service with few or no intermediate stops. This would speed up journeys significantly. Every stop skipped would speed up a journey by about 40 seconds.

For example, if 20 people board a 20-person minibus in Tuckahoe, NY and are all commuting to Grand Central Terminal in Manhattan, that minibus would head straight to Grand Central without making any stops. Currently, travelers from that town would take a train that makes 3-11 intermediate stops, so the shift to a non-stop journey would save 2-7 minutes.

Minibuses would also come much more frequently than trains. Since a single minibus is much smaller than a train, minibuses would stop at stations much more frequently to pick up the same number of passengers. Instead of a train every 10 minutes, a given station might see at least one minibus stopping every minute, each one routed to get passengers to their destinations with as few stops as possible. High frequency service makes journey times much more reliable, as it eliminates the risk that passengers will have to wait a long time just to start moving.

Autonomous minibuses would sync with AVs on local streets to help people make complete journeys using one navigation platform. One could imagine a platform like today’s ride sharing apps, where one enters their start and end points and the app then tells them which AVs to catch to complete the journey. An AV might drop a passenger at the minibusway station, at which point a minibus would be queued to pick them up there and take them further along their way. This would make transferring simple and quick, and it would make the minibusway system a lot more accessible to people who do not live within walking distance of a station. It’s similar to taking a bus to the train station, with the key differences that passengers would experience far fewer intermediate stops (in both vehicles) and that service would bring them closer to where they need to go.

mbw transfer from street
Example of a journey using surface streets and a transfer to a vehicle on the minibusway

Alternatively, where practical, minibusways could have ramps that allow vehicles to join from local streets (just like highways do). Where exactly to have ramps as opposed to stations is a judgement call based on the relative cost, effects on journey times, local impact, etc. However, people still might sometimes change vehicles before going on the minibusway in order to get on a higher capacity vehicle than the one used on the feeder service. Some locations could even feature a hybrid design with both a station and ramps (see below).

mbw hybrid station ramps
Minibusway station with stopping lanes at street level

Estimating the Capacity of a Minibusway with Passing Lanes

So, what would the capacity of a four-lane minibusway look like? The book Autonomous Driving: Technical, Legal, and Social Impacts, notes, “Compared to today’s observed capacity values of a lane of 2200 veh./h, an increase of traffic volume to about 3900 veh./h would thus be possible with purely autonomous traffic” (p.321). The 3900 figure is referring to one lane of a highway. For now, let’s assume that there are two lanes between stations and two additional stopping lanes for minibuses to pull over at stations. Potentially, about 3900 AVs could travel in each direction per hour. However, the autonomous minibuses would be longer and heavier than an autonomous car, so the capacity wouldn’t be quite that high. The same book estimates a capacity of 2,500 autonomous semi trucks per hour, so as a (very) rough estimate, let’s say that road capacity for 30-person autonomous minibuses lies right between that of cars and trucks: 3200 per hour in each direction. In this case, with every vehicle full, 96,000 people per hour in each direction could move across any given point on a minibusway. By comparison, a theoretical maximum of 93,432 people can pass in each direction through any given point on New York City’s 4-track Lexington Avenue Line (4/5/6 trains).

Matching Vehicles with Demand 

Most minibusways would not need to move 96,000 people per hour. For relatively low-demand routes, having a greater number of smaller capacity vehicles would do a lot to ensure that rides are frequent and flexible. For example, if only 60 people per hour want to travel along a given corridor, they’d be better served by ten 6-seat minibuses than by four 15-seat minibuses.

On any transit corridor, demand will fluctuate throughout the day. Transit companies end up running mostly-empty trains and buses during off-peak hours because they have a mandate to maintain a certain frequency of service and design vehicles to be as big as they need to be to handle rush hour crowds. Autonomous minibuses are much smaller vehicles and do not require human drivers, so during off-peak hours, transit organizations can run them more often and with fewer empty seats.

Minibusways could work on a congestion-pricing basis, where the toll is the minimum amount necessary to ensure that the road does not get overwhelmed with vehicles. Companies would compete for riders with vehicles of different sizes and designs, and the best vehicles for each route and time would emerge as dominant. If demand on a minibusway is low, the toll will drop, and people will tend to request more comfortable vehicles and/or private rides, as those will be affordable. If it’s rush hour on a busy minibusway, people will flock toward less comfortable and larger-capacity vehicles, as then they are splitting the toll with more people.

Vehicles would just have to meet some standards for factors such as acceleration and the time it takes them to discharge and load passengers at stations. Minibusways could even include allow low-capacity private AVs, provided their owners are willing to pay the entire toll themselves. 

2. Minibusways with One Passing Lane at Stations

Some candidates for minibusway conversion, such as three-track subway lines, would have room for only one passing lane at stations. The good news is that a minibusway with just one passing lane can also maintain free-flow traffic in both directions. The trick is that the flow of traffic in the center lane can be reversed many times per hour, so as long as minibuses are stopped in one lane at a time, there are always two lanes open for free-flow traffic.

MBW 3 Lanes States

Take a station with one reversible through lane (Lane 2) and two side platforms servicing the stopping lanes: one for southbound traffic (Lane 1), one for northbound (Lane 3). Between stations, minibuses typically travel in lanes 1 and 3. When no minibuses are stopped at either platform, traffic would continue on freely through the station using Lanes 1 and 3. Now say that a few minibuses stop in Lane 3 to discharge passengers onto the northbound platform. Before reaching the station, the other northbound minibuses would then switch to Lane 2 (the center lane) and pass the stopped minibuses. Southbound traffic flows freely using Lane 1. Once the northbound minibuses have left the station, the reverse scenario can occur: some southbound minibuses stop in Lane 1, the other southbound traffic passes them using Lane 2, and northbound minibuses all flow freely through Lane 3.

The traffic control system could achieve this by designating when vehicles in each direction of traffic can stop at these stations. For example, from 8:00-8:02, northbound minibuses can stop, from 8:02-8:04, southbound minibuses can stop, 8:04-8:06, northbound minibuses can stop, and so on. Traffic in the middle lane (Lane 2) is always headed in the same direction as the stopping minibuses. In reality, more passengers are probably travelling in one direction than the other at a given time, so it might be ideal to allocate more stopping time to that direction of traffic, e.g. 6 minutes northbound stopping per 2 minutes of southbound stopping. When the system is unable to ensure that stopping minibuses are due to reach their station during the designated time, it would have them pause in Lane 2 somewhere in between stations (where that lane would be empty) and merge back into traffic once they can stop at the following station.

3. Minibusways without Passing Lanes 

In minibusways without passing lanes, minibuses would travel in caravans, functioning as trains but with cars that aren’t physically connected. A minibusway conversion can still dramatically improve capacity by allowing vehicles to run closer together. Because trains are large, heavy, and run on steel wheels, they cannot stop as quickly as a car or minibus. As a result, subway and rail lines require signalling systems that keep the trains a good distance apart, so that they can safely break to avoid collisions. The need for these gaps constrains how many trains can run on a line at a given time.

Unlike trains, caravans of autonomous vehicles can run extremely close together. The minibuses can break much quicker than trains, and since they are driven by computers, each minibus can react extremely quickly if the vehicle in front of it brakes suddenly. The capacity of these minibusways is therefore constrained not by the signalling system, but by how long the minibuses spend stopped at the busiest station (their “dwell time.”) If minibuses at this station have to stop at the platform for 40 seconds, then 90 minibus caravans, the equivalent of 90 trains, can run per hour (3600 seconds/40 seconds) in each direction. By comparison, London’s Victoria Line can run as many as 36 trains per hour, the most frequent service in the entire Underground system.

Minibusways can speed up service on two-lane ROWs because they make skip-stop service practical. Skip-stop service is where some stations are marked A, some marked B and some marked A-B (or “all-stops”). Trains (or in this case, minibus caravans) either skip either A stops or the B stops, which allows them to travel faster on the line, saving about 40 seconds for every stop they skip on metros and electric rail lines and about 80 seconds on diesel lines.

Skip Stop Service
1961 map showing skip-stop service on Chicago’s ‘L’ (modern day Red and Green Lines). Skip-stop service ended in the 1990s due to declining ridership and long wait times at A and B stations.

The main reason skip-stop service is not more widespread is that A and B stops get half as many trains, so passengers going to/from those stations end up with longer waits. If trains run every 5 minutes, then A and B stations only get service once every 10 minutes. However, minibusways allow for high enough service frequency that no one would have to wait long for service. A and B stops would each see as many as 45 caravans per hour, so the impact on average wait times would be minimal. Moreover, during off-peak hours, minibuses can run frequently in small caravans. For example, when demand is low, 60 individual minibuses, “caravans of one,” could run per hour, with a minibus stopping at A and B stations every two minutes.

Even without room for passing lanes, minibusways still make infrastructure maintenance less costly because they run on pavement instead of rails, allow for vehicles to go to and from surface streets, speed up journeys, and provide a better match between vehicle size and passenger demand.

In Conclusion,

A minibusway offers higher capacity than a train line while providing a better network, faster journeys, lower operational and capital costs, and greater flexibility. Rather than reinventing the wheel, it  combines a number of advantages of existing transportation technologies:

  • It allows people to cluster into vehicles based on their destination like WVU’s PRT system does.
  • It responds to demand like rideshare platforms do.
  • It can support a continuous flow of traffic like grade-separated highways do.
  • It uses high capacity vehicles that can run on normal streets like buses do.
  • It has very high capacity like trains have.
  • When paired with a network of AVs on surface streets, it can ensure that everyone is dropped off within a short walk of their destination like taxis do.

To function well, dense cities don’t need trains. They don’t need traditional buses. They do need ways for a lot of people to move quickly and safely through a limited amount of space. Trains and buses do that job well in some cities, but if and when the technology is ready, autonomous minibuses will do it better.

One Last Thing: Some boosters of autonomous vehicles are so enthusiastic about the technology that they wheel around televisions showing groups of people footage of these vehicles in action. They call themselves the AV club. 


One thought on “The Minibus Way: How Autonomous Vehicles can Replace Trains

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s