Sleeper cars and the unexpected efficiency of solo transport


My article last week on Sleeper cars generated lots of interest, and Volvo's concept car generated lots of press too. Some of it was quite negative, including a article provocatively calling them the dystopian future we must prevent. I thought I would revisit some of the things about such vehicles that run quite counter to the intuitions of those used to 20th century transportation and transit.

Dominic Wilcox built this prototype. The stained glass is silly but the shape is roughly right.
Needs luggage room, different hinge, etc.

The first is the comparison with airplanes. The average domestic airline gets about 50 passenger-miles per gallon. Short trips are actually worse, because more of the fuel is spent in taxiing and climbing. You probably have noticed that the best hybrid cars also get about 50 vehicle miles per gallon, and with the average of 1.5 people on board have already beaten the plane.

I suggest the solo sleeper car might be the most efficient vehicle on the highway, as well as the cheapest -- once sensors start getting made by the millions. Compare the solo sleeper car with any feature list for any modern car and you will be shocked at all the things it won't have, which save money and weight. It will be a frame with a mattress, covered by a hinged fiberglass shell with no or few windows. The only unusual things about it, aside from the robocar sensors and computers, will be the active suspension system and special airbags for a horizontal passenger.

From an efficiency standpoint, it will be light and small. Much, much lighter and smaller than that Prius. Above your legs it may be only 2' high. Above your upper body perhaps 3.5'. And it will only be 3' wide, except at the wheels.

It won't need much engine. If it's liquid fuel powered, 100mpg is a reasonable target. If it's electric, 150 wh/mile is also reasonable. Gallons are not well defined there, but that's a worst case of 75mpg for pure fossil fuel electricity, but much better miles per pound of emissions in cleaner grids. To be electric, the car can pause every 150 miles or so for rapid charging to keep the battery small and light.

It doesn't need much engine because there is zero need for rapid acceleration or power. A 3 cylinder hybrid powertrain might do the job perfectly well unless there are steep mountains.

Platooning is possible for about a 20% saving on energy, in theory, but probably not in the first generation -- better to avoid risk early on and instead prefer as empty a segment of road as is possible late at night, which is pretty empty.

If the distance is under 350 miles, our vehicle can also go 50mph or less, which adds more efficiency. Sleeping passengers are the only ones who don't mind if the trip is slower.

So it easily beats the airplane -- but what about the train? According to the DoE, Amtrak is at about 52 passenger/miles per gallon, the same as the plane. The only difference is that trains run full far less often than planes, so they have room to take more passengers with low additional energy use. This leads many transportation writers to engage in wishful thinking -- if more people would just take the train, just look how efficient it could be! But you can't wish people onto the train.

For couple and family travel, it will be a group vehicle, but of course the family will find it particularly nice to load and bed down from their own home, and their travel will be extremely efficient -- much better than the plane.

The paradox of group transit

Earlier, I identified a paradox for 21st century transit, that smaller vehicles with less sharing are actually more efficient. Nowhere is this more true than for the idea of the sleeper car.

Compare the two options for an overnight trip, namely a solo sleeper car and a shared sleeper service on board a bus or train. The shared service suffers from several major downsides to the rider:

  1. You must travel on its schedule, not your own.
  2. You just travel to (and from) the station, which may not be in the direction you are going.
  3. Since it departs at a specific time, you must depart for the station with enough time to go through transfer, boarding, changing of clothes and a "buffer" to make sure you will never be too late. This is made worse if berths are not pre-assigned or trips are oversold.

It is quite common for people to depart for an overnight trip an hour or more before its scheduled departure time. Worse if one hopes to use public transit to get to the station for the overnight trip. Sometimes that departure is late, as well. In addition, there will be travel time from the destination station.

Compare that to the private solo ride. There, the passenger will spend a full night at home with family and prepare for bed as they always do, brushing teeth and putting on PJs etc. Then, at the exact time of their own choosing, they will step out the door and into the waiting sleeper car. They can dawdle reading their child a bedtime story if they feel like it. Zero travel time, and zero uncertainty of time.

If we imagine one of those lucky people who falls asleep quickly and soundly, they are asleep within 5 minutes of bedding down in the solo sleeper or the group sleeper. And thus the huge difference -- 5 minutes of awake travel vs. 65. It's a staggering difference. Normally when people debate between transit and private car, they might find the transit takes 50% longer, but offers advantages like being cheaper and greener and not having to drive or park. When it takes 13 times longer there will be no competition at all.

The only way the group sleeper can compete is if it's a lot cheaper and greener, and a lot less inconvenient. The only way to do that is to have departures very frequently from lots and lots of places. If there is a sleeper departing every 10 minutes, from somewhere that's 10 minutes away from you, then you don't have to worry about precise timing. You might leave for that only 20 minutes before it goes. That would require a lot of shared sleepers plying the route -- several hundreds per night. That means they have to be small -- 4 to 8 passengers. Not trains or buses. And on top of that, they can't run at full capacity, since if you want the flexible ability to just leave when you want, there has to be spare berths on almost all of them.

This is the paradox. The bigger you make the shared vehicle, the more the riders must be inconvenienced in order to travel together. The more that inconvenience, the more people are pushed to pay the extra money and energy for private transport, which in turn reduces the load factor and efficiency of the group vehicle. And in the human-driven world, that has pushed transit into larger more infrequent service and a spiral of decline.

This is a common error of transportation planners. They seek the maximum theoretical efficiency, and see that as large shared vehicles. The optimum efficiency is the maximum practical efficiency based on people's real use patterns, which are based on comfort, convenience and many other factors, and how they compete with alternatives.

The group vehicle can do a little better if the early passengers are very sound sleepers who will not be disturbed by stops to pick up other passengers. Then you can line them up along a route. But this is difficult.

You still don't get the tremendous convenience of leaving when you want to, though. And the privacy that comes with a solo vehicle. People care more about privacy when sleeping than when commuting. Reading reviews of Cabin you see mention of lack of privacy, conversations by others or even the driver and other interesting complaints that are inherent to group transport.

For awake passengers, group transport is easy, because we can, with minimal inconvenience, have stops for passengers to come and go, transferring to and from solo "last mile" vehicles. Most people are fine with a few short stops that don't take them off their desired route if it saves they money or makes them greener.

While you can't wish people onto the train, you can make policy to push them there. By far the best way to do that is to price all externalities and costs into transport, and remove all subsidies. If that were done, there would be more group transport during rush hour. But it would not universally help transit. Poor transit with poor load factors that is actually inefficient would actually go out of business, and sleeper cars are a likely place that would happen.

Use of roads

Some transport planners dislike solo cars because of road capacity issues. That's definitely a problem at rush hour, and so the use of solo sleepers for the morning commute is a different issue. Fortunately commuters are much more able to find convenient group transport if the vehicles are small enough to offer very frequent and flexible timing. Since they are not planning a full night's sleep, commuters will sleep in their day clothes and may in fact switch from seated to sleeping. At rush hour, everybody is rushing, so it's easy to find people to pool with at the time you want, all living not too far from you. That's not the case for overnight travel.

Overnight travel, of course, makes use of the roads at their most empty. The car heading out of town at 11pm is a car that's not going to the airport in the next morning's rush hour. Intercity sleeper cars will not congest traffic, unless they want to arrive at their destination downtown from 7am-9am. For those wishing to do that, the same principles of road access metering should apply, so that we never allow more cars on a road than it has capacity to handle, almost eliminating congestion and the negative effects of induced demand. Robotic sleeper cars can be expected -- required -- to follow such road metering rules.

A solo robotic sleeper need only be about 4' wide, allowing you to fit 2 per lane like motorcycles. In addition, if sleeper cars at night replace taxis to the airport during the busy times of the day, they would actually reduce congestion by distributing the load on the road.

The destination

What happens at the destination depends on where that destination is. In today's polycentric cities, it will only be the downtown some of the time, but most group vehicles (other than planes) try to head there.

On outbound trips, most people might like to end their sleep at a depot which offers private changing areas and showers. From there, they could catch any form of regular transportation, from robotaxis to group vehicles or combinations of those. These depots do not need to be in the downtown, particularly if many final destinations are not in the downtown. At these outskirts depots, arrival after 7pm should not be a problem in terms of congestion. Group sleepers will also need such depots (which really are simply existing hotels that offer this as a service) but will need to centralize somewhat in selecting them.

Inbound trips have a bit of a difference. There, the group and solo sleeper are on a more even footing if the passenger is not in a convenient location to prepare for travel. Even so, the solo service will allow them to leave at their own preferred time, and from any of a thousand places that might offer a preparation room for a small fee. The main difference will come at the home destination, where the solo sleeper will go directly to the passenger's home, and wait silently for them to awaken. The group sleeper will go to some station, and may or may not require awakening there.

Other notes

Some people wondered if it wasn't wasteful to have a car that was used only at night, and whether it might make more sense to use the robotaxis that work during the day and are idle in the night.

A sleeper car doing the SF-LA run would go 400 miles per night. That's actually more than your typical urban taxi will drive. Manhattan taxis do about 180 miles/day. Suburban taxis will do a bit more. If it works 312 nights a year that's 125,000 miles per year. That's only 2 years of operation before it hits the lifetime of today's typical cars (which is actually 190,000 miles.) I suspect this car can actually last much longer because it does only highway driving. Change the mattress every so often and make it easy to clean and it might last much longer. But it's still a high duty cycle with no need to increase it.

You do want to keep it reliable, though if it suffers a breakdown along the way, another car will show up shortly if the operator keeps such spares scattered. Or worst case, a regular car with a seat (and more than a refund for the inconvenience) should come to you quickly.

You could make cars that convert between day and night use, but at extra cost. If electric, they would unfortunately arrive fully discharged for the start of rush hour, which is not great. They could charge up and do mid-day or evening rush hour, then recharge for the overnight run. But it's simpler to have sleeper cars for sleeping and seated cars for awake riding.


I think the point that people won't want to travel to the station is fair. But what about loading the sleeper car into a train? That is, why not board a sleeper car at your doorstep, and let it drive and load itself onto a train? To you it makes no difference, but a train full of sleeper cars might still be more energy efficient than than a fleet of individual sleeper cars?

The problem is I don't think that's very efficient. Takes up a lot of room on the train. I think it's a challenge (though not impossible) to do this in a way that will not disturb the occupant. It might be acceptable to wake them briefly.

Trains can do well for drag (good ratio of length to frontal area) but they tend to be super heavy, losing the advantage the low rolling resistance of steel on steel has.

This has some problems, though. Everybody still has to be on the same schedule (though they can leave from different places.) Unless you can have trains every 20-30 minutes in the late evening. And can they get out of the vehicle to go to the bathroom? No stopping for a pee if loaded on a car ferry.

Frankly, unless you can load the vehicle on the ferry very silently, it may be better just to have a sleeper take you to the train, then wake up and walk to a berth on the train.

Another option, if the track is clear, would be to put rail wheels on the cars, and have them able to go onto the track, and to platoon with other cars, possibly even physically dock. You can also platoon on the highway but that has challenges. You could also get power from the rails in that case.

I really like the idea of cars with wheels that can be used on train rails. On/off stations could be much more frequent than train stations are now, and that addresses the issue of needing a bathroom part way through the trip. I think a key advantage to using a car instead of a plane or train is that you have your car with you at the end of the journey, and don't have to rent one. You also don't have to transfer luggage from the plane or train to a car, bus, or taxi. I expect that cars designed for this application will be much more comfortable and convenient, so that even when you aren't sleeping, you can enjoy yourself or get work done. The experience will be much better than a coach ticket on an airplane. Thinking further, robocars that are making deliveries can also take advantage of rails for longer distance shipments. The key is designing a rail control network so that cars can enter and leave the rail line with little impact on the rest of the rail traffic.

These are possible, but there is an expense to that of course. As the trains become less useful, one can more easily pave the tracks so all cars can use them. You need to pave the exits and entrances (with rails still embedded) if you don't.

Luggage transfer is no issue for two vehicles that are stopped next to one another, trunk next to trunk. There is no "car rental" in the robotaxi world, you just summon cars as you need them, so there is no problem of "don't have to rent one."

Add new comment