Pooled Autonomous Vehicles

Fully autonomous vehicles are expected to arrive within the next ten years – billions invested by almost all major car makers and many technology companies will ensure that this is not just wishful thinking. With no driver required the concept of owning a vehicle that sits idle 95% of the time however looks increasingly odd. The London and Silicon Valley based think tank RethinkX therefore makes radical predictions about the future of mobility : 90% of all miles driven will be autonomous and shared, as early as 2030. The assumptions in their research are worth studying in greater detail in the future but they appear mostly plausible and interested readers will find plenty of articles and videos that expand on this. In this post I’d like to make a few contributions of my own that focus on pooling and vehicle design in particular, which could make this future even more appealing and affordable.

Benefits of AVs in general

There are so many benefits to autonomous transport as a service, so here is just a quick rundown without going into too much detail:

  • Lower cost, freeing up resources for other spending
  • Freed up space for parking
  • Lives saved from far fewer accidents
  • Longer distance commutes possible, enabling more space for housing and people
  • Empowerment of those without transport, including children and elderly
  • More physical connectedness of people
  • Increased productivity in transport as people work or sleep
  • Parents freed up from transporting children
  • Transport speeds up as road congestion decreases
  • Compared to public transport wait times at interchanges are eliminated
  • Drivers freed up to do more interesting tasks

Given these many advantages it is highly likely that the public will soon prefer transport as a service outside the most rural areas. This is because autonomous vehicles (AVs) will be available more cheaply and more reliably than individually owned vehicles and even some forms of public transport. It is interesting that Elon Musk has explicitly omitted buses in his updated master plan, instead believing that AVs will take their place at similar or better per mile cost.


One key remaining question is whether this transport as a service will be mostly pooled or not. There are two arguments in favour of a largely one party/one vehicle model:

  1. Privacy and comfort: people are used to privacy in their own cars, which is often the reason they currently prefer driving over public transport. Sharing a confined car space with strangers is not very appealing compared to stretching out in your personal space.
  2. Predictability and speed: no detours to pick up or drop off other passengers and no extra wait to be find other riders

By contrast the two main reasons in favour of a more pooled model are:

  1. Cost: the marginal cost of transporting another paying passenger if they share the same pick-up and drop-off points are close to zero
  2. Congestion: road space is limited, especially at peak times and speed could be increased in aggregate if fewer vehicles were on the road

There is a way that these two advantages of unpooled vehicles can be fitted into a pooled solution that satisfies all objectives. Privacy can be created with a new vehicle design while speed comes down to a software solution.

Vehicle design

Based on a large minivan design of approximately 5-6m in length an improved pooling friendly design would look as follows:

  1. The entire length of the vehicle would be taken up by the cabin, similar to buses. The shape may come to resemble a slightly larger version of the VW microbus and its latest prototype. There is no need for an engine compartment at the front, all batteries and engines can be fitted into the floor. Internal combustion engines won’t be competitive by the mid-2020s according to multiple forecasts (UBS and others). As AVs are far safer the need for a front crumple zone (or even a Tesla frunk) disappears. The raised height of the floor also allows less incursion of the wheel arches into the passenger space. The result is a long, almost flat floor plan with a mostly rectangular passenger cabin on top that allows for new interior design ideas. Poor aerodynamics from the boxy shape are a lesser consideration given increasingly cheap energy – like a bus or RV maximizing interior space for a given vehicle size is the most important feature.
  2. To address the privacy and comfort the cabin would be divided into 2×2 compartments, each with its own sliding door. The compartments would be separated by half-height dividers that are raised to full height by default, similar to the sound-proof dividers in current stretch limousines between the driver and passenger compartment. A request from both sides would be needed to lower them for groups travelling together. The windows could similarly be covered, frosted or tinted to enable total privacy.
  3. To further enhance comfort each compartment would contain a seat similar to a fold-flat first class airline seat. This would allow passengers to sit, recline or even lie flat to sleep. It is likely that seat-belt requirements will be removed as AVs will be treated like buses or taxis. Similar to such airline seats there will be fold-out tables and screens that can be lowered from the roof, in case passengers want a surround view of the car or get entertainment from their devices or in-car services onto a bigger screen.
  4. Given the length of over 2m in each compartment there is room to use the space flexibly. A jump seat at the front of the compartment could be included to be used by a second person for shorter trips, similar to the rear-facing jump seats in London taxis. As there is no boot, luggage would be stored in the large compartment itself, either behind the seat or at the front. If more space is required an additional compartment can be booked.

The interior design would address comfort and privacy, despite sharing the overall vehicle. The best analogy here is a first class airline cabin, with the added privacy of partition walls. None of the pictures in the slideshow below quite hit the mark but there is a combination of all of them could be mashed up into a final design.

Scheduling and software

To address the second concern of predictability and speed the solution lies in software and network effects. A passenger arranging transport between two points would book by compartment, not people, depending on the size of the party and luggage requirements. In most urban environments it is likely that there will be many similar trips booked within a short time of each other. In order to achieve the benefits of pooling such trips a miniature hub-and-spoke model would be adopted. The passenger would most likely set off alone in a vehicle until reaching an optimally located spot at which up to three additional feeders would arrive at almost the same time. Passengers would then change to the vehicle that would take them into the area of their shared destinations, before repeating the changeover again near the final destination. No detours for pool passengers! If carefully timed, which is very possible given increasingly sophisticated software and large number of users, such changeovers should not add more than a few minutes to travel time. This in aggregate would be repaid by much fewer vehicles on the road that would allow higher speed. Cities may indeed mandate this.

This scheduling system is in effect a highly personalized form of public transport, gathering traffic along trunk routes, lowering congestion, but at the same time enabling true door-to-door convenience. Network effects would enable the provider having the most vehicles on the road to provide the tightest schedules and lowest costs, creating a market that will most likely become oligopolistic, something that may in time attract the attention of regulators.

As an aside, the traditional way of hailing a taxi may actually make a comeback, e.g. in situations where someone’s phone may not work. Exterior cameras on AVs on the road could pick up a particular hand motion of a person seeking transport, which would then activate facial recognition and acknowledge the request on an exterior display. The service would then send a nearby vehicle to pick up the passenger, who would input the destination on the on-board systems. As this is much less predictable than pre-booked traffic this would be a service with a significantly higher fee.

It should also be mentioned that AVs can drive in tightly packed convoys that greatly increase the number of vehicles that can travel on a given stretch of road at a time. Traffic lights could be prioritized for such highly utilised and tightly packed vehicles to encourage adoption of this mode of transport. Mass rapid transport will likely shrink greatly outside the densest urban environments.

Cost implications

The RethinkX report calculates a very low per mile cost of $0.08-$0.24, which dramatically undercuts all other forms of transport. This translates into an annual cost for a daily 30 mile commute of just $550-$1700, based on 230 working days. However, there are two omissions in these figures: peak demand and deadheading (i.e. non-revenue miles). Peak demand is reflected in Uber surge pricing, a widely disliked but necessary increase in prices of up to 4x to bring supply and demand into equilibrium. Underlying the low cost calculations made in the report are high utilisation rates for all vehicles. To avoid surge pricing without pooling a much larger fleet would be required, which would sit empty much of the day, increasing per mile costs for all. By relying more on pooling during peak hours prices can be kept low and predictable throughout. Conveniently this is also the time where matching riders is easiest. Deadheading is also reduced in a hub-and-spoke model as only the return traffic on the trunk routes may be empty while the feeders could constantly be kept full, as the hubs are themselves moving constantly so could be organized such that empty local trips are minimised.

The result is that pooling may not reduce prices much further as indicated by the report but instead could keep them constant even at peak times.

Long distance transport

Another area that will open up with such pooled AVs is long distance transport. Door-to-door travel, possibly overnight, to a destination a few hundred miles away is suddenly far more comfortable than the alternatives of trains or planes with long interchanges, pre-boarding procedures and transport to and from the hubs. As such trips are most likely planned well in advance it is quite possible to create a small pool of passengers, as only a maximum of four is needed. As this traffic would mostly occur at off-peak hours without much deadheading this should enable even lower per mile cost than local traffic. Given greatly increased safety and zero emission of AVs there should also be no reason to maintain speed limits at the low current levels, allowing for even greater distances. There may be a small subset of the vehicle pool that has larger battery capacities for the longest distances but during the day they’d act like all others.

This form of transport could become the most convenient for anything up to a few hundred miles, beyond which flying is usually the fastest. This may squeeze out high-speed rail in many situations.


To drive even higher utilisation these vehicles could serve as delivery vehicles during off-peak hours. After establishing that the recipient is available they would walk up to the car and after some interaction with their phone the door would open. To ensure only their package would be taken a number of steps could be taken:

  • Large, bulky or valuable items could be transported in their own compartment.
  • Multiple, lower value packages could be put in a single compartment. Cameras would be switched on to both guide the recipient to the correct package (augmented reality) and to ensure only the correct package is taken. Violators could be instantly reported.
  • Alternatively, locked pigeonholes could be installed for delivery hours in one or more compartments. The installation with pre-filled pigeonholes could be done at delivery centers within minutes on top of the lie-flat seats. The recipient would then only get access to their pigeonhole.

This is probably inferior to other ideas laid out in another article describing a van and drone combination   but with near zero marginal cost at off-peak hours this could be a complement to other delivery services. The ability to deliver larger items in the huge luggage space in particular would be a complement in situations where drones would not work. A service similar to Uber Eats could also be viable for the very short local distances required there.

Using these vehicles as both passenger and delivery vehicles, maybe even at the same time will again cut costs for all services and increase utilisation.

Ethics and other uses

In the context of AVs many ethics questions are being raised, such as the behaviour in accident situations, external cameras recording streets and faces. I’d like to address two specific questions here:

  1. Privacy: should cameras be installed in each compartment? Preferably, cameras should only be activated in a few situations: between riders to inspect the state of the cabin to determine whether cleaning service is required, during package delivery or for children travelling on their own so parents can watch over them. Parents not having to do school runs and taxi service to friends and activities will be a major productivity driver!
  2. Does society accept if such highly private spaces on the road are used for activities not primarily related to transport? In many cases we’re fine with that, e.g. food trucks, mobile clinics but what about using a car for taking naps or even as mobile hook-up sites? What may be happening rarely in the few stretch limousines around may become more common in ubiquitous AVs. A free society should judge whether these cases impinge on the freedom of others so we should normally have no business in restricting what people are doing in these new personal spaces.


Given very compelling economics it is very likely that the predicted comprehensive transition from individual ownership to transport as a service will materialise in the near future, whether by 2030 or slightly later. In the inevitable debate of personal mobility vs. increased congestion the partly pooled model outlined in this article is likely to satisfy all concerned. In addition, the proposed single, unified multi-purpose vehicle design may lead to the highest utilisation and lowest cost outcomes. City planners will be pleased that despite the reduction in mass public transport the number of vehicles on the road will remain manageable due to high occupancy. Passengers will appreciate the very low cost, yet first class experience while maintaining personal transport speed, flexibility and privacy.

This slideshow requires JavaScript.


Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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