How to make a successful Automated Transit Network

An economist's perspective

By Roman Zakharenko

An automated transit network (ATN henceforth) is a conceptual transportation system that features fully autonomous on-demand vehicles using a network of guideways, and that is fully isolated from other traffic. The ATN concept offers great benefits over existing land transportation modes: safety of a railroad, convenience of taxi, low operating cost. Separation from ground-level traffic provides both higher speed for ATN vehicles and safety for pedestrians. High energy efficiency is also easily achieved.

The ATN concept was contemplated since 1950s, and individual small-scale projects have been successfully implemented since 1970s. Yet, despite its obvious advantages and despite improvements in complementary communication and computing technology, ATN have so far been unable to compete with other, more traditional modes of ground transportation. The objective of this note is to present an economist's perspective on the factors of success of this technology.

Network effects

Economic theory operates with a concept of network effects. These effects exist when a consumer's intrinsic value of a good or service increases with the number of other consumers using this good/service. For example, one prefers to join the social network to which most of his/her friends have already joined, to use the online map with most users (as such a map is likely to offer the most detailed and up-to-date information), and to use the most popular ride-hailing app (as such app partners with more drivers and thus offers shorter waiting times). Markets with network effects are typically dominated by a single product, usually the one that was more successful in the early days of the respective market. Products with smaller market share usually lose their customers and are forced to exit.

Network effects are very strong in ground transportation. The Wikipedia article "Track gauge in the United States" lists fifteen different railroad track gauges used in the US; most of these were introduced at the dawn of the railroad era, and were replaced by the standard gauge of 1435mm by late 19-th century. The demand for standardization was driven by network effects: every town wanted to be part of the largest network, which allowed trade with the largest number of other towns.

Another example of network effects in transportation is competition between freight trains and road trucks in the mid-twentieth century United States. Trains are superior to trucks is almost every respect. Transportation by rail is more energy efficient, safer, less labor-intensive, and less dependent on weather conditions. Yet, railroads have lost competition to trucks due to a single disadvantage: network size. A truck could be driven to every home or office, and a train could not. The power of network effects was sufficient to drive an otherwise superior technology out of business.

This evidence allows us to formulate a necessary condition for success of a new technology like ATN: a new transportation technology can succeed only if it is capable to grow its network at high rate, matching that of the competing technology (the automobile) within reasonable time.

Further sections elaborate on the factors of rapid network growth.

Optimal guideways

Every transportation system consists of two key elements: guideways and vehicles. It is the guideways that exhibit network effects: more guideways make the network more attractive to customers, by allowing them to reach more destinations. The vehicles do not exhibit network effects: a more congested network certainly does not attract customers.

This observation allows us to formulate a corollary of the previous necessary condition of technology success: a new transportation technology should minimize the guideway cost in its early stages; the cost of vehicles is less important for initial success.

This recommendation was certainly followed by pioneers of the two most successful ground transportation technologies, the train and the automobile. The early vehicles were quite expensive, but the initial guideways were not. Inventors of the steam locomotive did not invent the railroad track, but only slightly modified the existing wagonways. Only after the steam locomotive has proven itself, the problem of better rail design received attention.

Likewise, the first producers of automobiles did not invent their own road, but used a vast existing network of horse roads. By the time the first mile of road was paved (Detroit, 1909), 75 thousand cars were made annually in the United States alone.[1]

Because the concept of ATN implies the use of guideways fully isolated from other traffic, which currently do not exist, these guideways should be designed in a way that minimizes their cost and strips them off all elements unnecessary for safe transportation at automobile speed. It is up to engineers to define the "cheapest possible" guideway; perhaps it is simply a pair of rails attached to some kind of (elevated or underground) surface. The power rail can be added in urban environments to supply electric power to the vehicles. In rural areas with lower traffic and longer distances, electric power provision may be too costly; long-range vehicle batteries are also not an option as their high weight would make the guideway support more expensive. The best solution for low-density areas could be the use of a gasoline- or hydrogen-powered generator, externally attached to the vehicle.

One-abreast vehicle seating is better than two or three: not only it allows narrower track, but also reduces the load per axle, making the guideway lighter. Narrow gauge also simplifies access to indoor stations via existing windows. Most vehicles today carry only one occupant anyway; automation, by allowing travel without a driver's license, will only increase the share of single-occupant vehicles.

Supported-vehicle systems are better than suspended-vehicle systems, because the former allow shorter support columns and thus lower costs.

Performance-enhancing but costly guideway elements, such as linear motor or magnetic levitation elements, or solar panels, can be introduced at later stages of development on busiest segments of the network.

Organizational structure

In virtually all ATN projects, vehicles and guideways are owned or leased by a single operator. This organizational structure is not consistent with the practice of the two successful transportation modes, railroads and automobiles. Each of these is a decentralized system of independent operators; in case of automobiles, vehicle operators are also independent from road operators. Easy access of new individuals or businesses into network operation is essential for its rapid growth. Therefore, successful ATN concepts should allow new independent guideway operators to join the existing network, and also allow addition of new public or private stations. They should also allow existence of independent vehicle operators, with access to the entire network (except private stations). Allowing vehicle design by independent firms is also a possibility to consider.


Freight transportation

Virtually all ATN proposals have focused on passenger transportation. For example, a 240-page review of ATN technology[2] uses the word "freight" only 6 times, typically to mention to a secondary supplement to a passenger ATN project. The reasons for such lack of interest in freight transportation are unclear; perhaps ATN developers view freight vehicles too large for their compact guideways. To better understand the future of freight transportation within ATN concept, it is therefore useful to discuss the general theory of vehicle size.

The size of individual passenger vehicles, quite obviously, is designed to match the dimensions of human body. Understanding the size of freight vehicles is slightly more involved. Theoretically, the size of a commercial vehicle carrying divisible goods is determined by non-scalable costs, i.e. costs that cannot be scaled up or down in proportion with vehicle size. To reduce such costs per unit of cargo, larger vehicles are used. In case of trucks, the non-scalable cost is that of the driver's labor: while the size of every part of the truck can be cut in half, the size of the driver (and, importantly, of the driver's wage) cannot be downscaled. But if the driver is the bottleneck that determines the truck size, then automation should lead to a dramatic reduction of that size. Without the need to pay the driver's wage, nothing prevents transportation companies from splitting a 48-ft semi-trailer into two 24-ft trailers, and then further split these into four 12-ft trailers.[3]

How small can freight vehicles get? That depends on the next non-scalable element of vehicle capital- or operating cost. The possible candidates are expensive indivisible equipment, such as a lidar, or the requirement to monitor individual vehicles by human operators (e.g. some projects envision 10 trucks per operator). It is my understanding, however, that ATN developers to do not intend to use lidars; nor they plan monitoring individual vehicles by humans. If that is the case, there is virtually no limit to downsizing freight vehicles; they can become as small as an individual consumer good that they carry. Imagine a bottle of soda or a box of pizza, with four wheels and some motor attached: this may be the image of the future ATN freight vehicle.

Such transportation technology will likely cause a revolution in retail industry, removing many intermediate steps between the producer and the final consumer. It can also revolutionize waste management: if every box of pizza, after being used, can return on its own to the restaurant, then pizza boxes (and other packaging items) will soon become reusable. Numerous other innovations will become feasible, e.g. automated trash pickup from public areas, or airline baggage check-in/reclaim near the passenger's home.

This implies that freight transportation can actually gain far more from introduction of ATN than passenger transportation. Therefore, an automated transportation technology should emphasize and possibly prioritize freight transportation. Launching unmanned freight traffic before passengers may also reduce initial guideway operating costs (e.g. by removing the need for emergency response crews), and will allow to test and demonstrate the safety of the technology.

ATN developers seem to be confident that their networks will never be congested. But if every consumer item can travel individually as described above, such forecast may be far off the mark. In fact, without proper regulation, ATN may become the most congested transportation system in human history.

Most economists recommend to curb congestion by imposing per-vehicle tolls. Such tolls create incentives for "cargo-pooling", i.e. bundling several consumer items into one vehicle. Such bundling also implies that individual parcels may have to change vehicles and make connections during their trip, much like mass-transit passengers do today. Therefore, some sorting stations, where parcels can change vehicles, may be needed.

The prospect of congestion and the need to control it is actually good news for ATN developers. In the absence of natural non-scalable costs of vehicle operation, tolls become an artificial non-scalable cost, reducing the number of freight vehicles and increasing their size. The toll revenues will increase profitability of the network. Congestion will also increase incentives to build additional lines, further increasing the network size.

My own earlier research[4] proposes to replace fixed tolls by tolls that increase with self-reported value of time (VoT), and to give priority to those with higher declared VoT. It is shown that the proposed mechanism benefits all members of the traffic: those with high VoT gain from receiving priority, while those with low VoT gain from having to pay less (compared to fixed-toll scenario). Higher customer satisfaction will also increase ridership and thus revenue of transportation providers. Such "priority-for-sale" will also increase incentives for passenger-pooling and cargo-pooling, as two passengers (two parcel senders) using one vehicle can not only save on tolls but also buy more priority.


This note outlines an economist's view of the factors of success of the "automated transit network" concept. The recommendations include making the guideways as cheap as possible (even if that requires more expensive vehicles), making organizational provisions to allow new independent operators to join the network, and emphasizing development of small freight vehicles. These strategies maximize the rate of initial network growth, which is essential for success of a new transportation technology.

Roman Zakharenko, PhD

Associate professor, HSE University (Moscow, Russia)

December 2, 2021

[1] Wikipedia article "U.S. Automobile Production Figures"


[3] Surprisingly, numerous designers of self-driving trucks seem to ignore this simple economics of vehicle size. All projects I could find, except low-speed delivery robots, copy the dimensions of existing trucks.

[4] Zakharenko, Roman: "Traffic Priority Mechanisms". Transportation Science (2020)