Brad Templeton Home
Downsides to robocars
Robot cars (robocars) are an overwhelmingly positive technology. However, nothing is perfect. Here are some issues we'll see in a world that moves to robocars.
Robocars will take their passengers quite literally door-to-door. Indeed, if they are electric and are given remote door opener codes, they will take their passengers inside buildings.
(In the movie "Minority Report," they actually climbed the walls of buildings to act as personal elevators -- that's still science fiction for now.)
Deliverbots can do the same thing -- bring things right inside your house, even from a store around the corner.
This has the danger of creating a class of urban dwellers who barely walk at all, which is bad for overall health, and bad for the character of urban spaces.
We might help this a bit with the "bikebot" -- a small robot that clamps to a bicycle and can quickly bring it anywhere, so those who wish exercise can use a bike when conditions are right. Robocars also easily facilitate one-way trips which combine walking and riding -- if people have the will to do it. (For example, have your robocar drop you a mile from work on your commute, but pick you up right there when in a hurry.)
Deliberately walking one way for part of a trip and returning by robocar becomes very practical too.
There may be more alcohol abuse if we remove the constraints on drinking that "needing to drive home" provides. However, all-transit/taxi large cities seem to survive this problem.
People rarely do what they should do. As car engines have gotten a fair bit more efficient over the last 25 years, the result has mostly not been cars with better mileage. It's been heavier cars with the same mileage. While this problem does not prevent a robocar future, it reduces the benefit. People may move to bigger, more luxurious cars if they can afford them.
Indeed, some segment of the population might decide to always travel in a living-room sized RV style vehicle. Today that's not practical because of the difficulty of driving such vehicles and the extreme difficulty of parking them, as well as their high fuel cost. That fuel cost, however, is only triple that of a small car and well within the means of many.
It could be far worse. Some will decide to always travel with a fleet of RV like vehicles which dock and form a house everwhere they go, at least if there is room.
Encouraging longer commutes and sprawl
The robocar commute will be very comfortable, and passengers will be able to work, read, watch TV, use the computer or sleep while riding. This could encourage much longer commutes by making them more tolerable.
This is not environmentally a good thing, even if the vehicles are themselves much more efficient. Long-range trips will still be mostly in fuel-burning vehicles until we get better battery technology, even considering the robocar's ability to take itself to a charging station.
This may encourage more suburban sprawl, and more isolated 'burbclaves -- remote gated communities. While many people like to live in a walkable core neighbourhood, more remote areas will offer more land, and a peaceful safe environment. The car itself spawned the suburb, it is uncertain what the robocar will spawn.
(Robocars will make cities nicer though, with much less pollution and noise. Today's transit is remarkably noisy.)
Sleeper cars might encourage vacation homes that are further away (and cheaper) due to the 8 hour trip. A slow sleeper car (doing a trip that's 4 hours at 70mph in 8 hours at 35mph) would be much greener for such trips, but people will also elect vacation properties that are 8 hours away at 70mph if they wanted to.
A sleeper car trip will still be greener than a plane flight, however, for trips on the order of the San Francisco to L.A. run.
It is not just commutes which may become longer. People may be more willing to send their children to schools that are further away. Today there has been a trend of parents always driving their children to school (instead of walking) and constraints from that will be removed. Desegregation busing takes on a new dimensions as well.
It's even possible to imagine that robocars could re-destroy the "downtown" concept. Most downtowns thrive because they are transportation hubs. One can get there, and one can move from place to place within a downtown with short trips. With robocars -- and no transit -- offices, cinemas and many other "destination" locations may see no need to pay downtown rents.
With traffic congestion a thing of the past, moving from destination to destination becomes much easier in a dispersed city, and the moves will be work/play time. The desire to have walkable "cool" areas could preserve the downtown concept, but cities might become many urban clusters rather than a single downtown. There are limits -- one will want robocar trips to other desired locations to be short, say 5 minutes at 30mph or 2.5 miles -- but the rules will change.
(The most efficient cars will be lightweight urban cars designed for street speed. Highway speed cars will be heavier and less economical.)
I certainly don't have the training in urban planning to predict what may happen here -- I am not sure if anybody does. We don't yet have enough information about what quick, pleasant, on-demand direct transportation would do to cities and suburbs.
Abusing robocar collision-avoidance
Because robocars will avoid collisions so well, some people will both abuse that and come to depend on it. Human drivers may enjoy the fact that they can just plow through lanes of traffic at high speed, and the robocars will get out of the way if there is any room. Such driving should deserve a ticket, but enforcement without Orwellian devices is a challenge. Imagine the typical Boston driver truly able to not care about the other cars. A Robocar will always lose a game of chicken.
Teen-agers will regularly test how good the robocars are. They will jump out in traffic among them. They will throw things into the road. They will veer and disrupt because it is fun. They will play street hockey without fear, perhaps letting robocars zoom through their game. They will enjoy causing robocars to suddenly brake, disrupting the ride for their passengers, and then run away.
While it should be safe for a child to wander into the street, parents may start to not watch their children as much around traffic. Pedestrians may feel they can routinely cross the street anywhere without concern. (They presumably can, but it will disrupt things for passengers.) Children growing up in an all-robocar neighbourhood will act unsafely when they first come to streets with human drivers.
Where does the electricity come from?
If robocars make a shift to electric cars real, there isn't nearly enough power in the grid to support them. That means creating lots of new power generation and delivery infrastructure. Currently half of U.S. electricity comes from coal -- not a very good alternative, though our robocars will use less overall energy.
There is debate about whether nuclear is currently economical even if it gets public support. There are promising technologies in thermo-solar, cheap PV and geothermal, but they are just promising technologies. While our goal is to get off fossil fuels, it may make sense to stay with them until alternatives arise, but with lower usage.
While hydro is low emission, it destroys valleys and almost all the available hydro sites in the USA are already developed. On the plus side, electric car batteries charged at night may allow much more efficient use of the grid, nuclear plants in particular.
Deliverbots (and empty car-moves) may suffer a piracy problem. While they will of course have telemetry so they can send video of the pirate attack, in the end sophisticated pirates will be able to grab them and take them and their cargo. Deliverbot operators may simply decide not to route deliverbots through areas known to have pirates, and may not offer deliveries into those areas.
Both deliverbots and robocars will no doubt be designed with anti-theft systems which render them inoperable if stolen. For example, they could be programmed to require a return to a home base every few days. Without that return, they would become incapable of any trip but the return. Of course thieves could still steal them for parts and metal.
We will also see vandalism, particularly from teens, particularly of vacant vehicles but even of occupied ones.
Care must be taken to protect privacy of robotaxi and robocar users. There will be temptations in government with all high-tech, computer networked vehicles -- not just robocars -- to make records of where and when they go places. Even robocars that aren't explicitly tracked will still be trackable because of their network queries and network roaming.
Indeed, thanks to licence plate scanners, this is already happening.
Robotaxis face a more serious problem, especially private cars hired out as robotaxis. These will probably not be anonymously hailed on the streets. Owners will want riders to have accounts, and want some way to find them if the interior of the vehicle is damaged or dirtied during a rental.
A picture of the interior of the robotaxi can be taken before and after the rental, without doing surveillance of the rider during the ride.
To protect privacy, records of trips should be destroyed after it has been confirmed the vehicle is still in good condition by the next customer. We may need laws to demand such data destruction, but expect those to be hard to win against police craving new tools. It could also be possible to use an intermediate company which protects your identity from the robotaxi company, and offers a bond to pay for any misuse of the vehicle. The owner doesn't need to know who you are, just that they can make you pay if you abuse the vehicle.
It's important to realize that the technology does not demand having a central traffic control, or recording of where people go. But there are a few features to be gained from this and there will be some pressure to produce such systems. Care must be taken to resist this central control architecture in design.
The most compelling case for central authority involves highly congested areas at peak times. There, it is desireable to be able to ask for a reservation for a slot in the congested area, and for the controller to hand out a limited number of slots to reduce congestion. This can be built without privacy-invasion, but it's more work to build it that way.
You may want to read this sidebar on privacy issues and how to deal with some of them for more information.
There will be strong temptations by governments to ask for remote control of our robocars. Robocars, at least as shipped by the manufacturers, won't be able to break traffic laws. While this also eventually eliminates the need for traffic cops to police them, police will still seek power. (In an all-robocar city, traffic laws become mostly moot, though. Recently small european towns have started removing their traffic signs and replacing them with a simple policy of courteous driving. That's not out of the question for robocars.)
Another interesting fictional treatment is the Hugo award winning (best SF novel of 2006) Rainbow's End which you can also purchase via his site. Robocars are a background item in Vinge's "one foot in the future" novel. In various scenes, cars refuse to obey their occupants to break traffic laws or go onto unapproved dirt roads.
The police will demand some level of control of robocars, in particular citing anti-terrorism as the reason. This might justify a kill switch which could be used locally to disable a robocar believed to be on a malicious mission, but a remote recall (against the passenger's will) will need to be fought. There will be different battles for a personally owned robocar and a hired robotaxi. The hired robotaxi will surely remain somewhat under its true owner's command, though a set of rights for passengers needs defining.
There will also be debate about how much people can modify their robocars systems. Cars will presumably be licenced -- as they are today -- with only licenced and tested systems allowed on the road with other cars or people. This will also be used as a mechanism of public control unless great care is taken. Safety is important but it will be very easy for the licencing system to stifle innovation and block small companies from entering the market -- in fact the big companies will lobby for more licencing requirements for just that reason.
I personally believe that we can design the other robocars so they are not in danger from a less safe, personally modified robocar. Only the willing occupants of the experimental car would be taking the added risk. Each revision of robocar software would earn a trust reputation, and brand new and unknown versions would be given a wider berth by the others.
Police and military will want ways to block robocars from entry to "locked down" areas. Concepts like Presidential motorcades will become easy, perhaps too easy.
We must decide who our machines will serve -- their owners, or the government. Government control is not assured, above the licencing required today, but there will be call for it.
Central control of robocars also opens the potential for abuse of that central control by both corrupt or repressive governments. Perhaps worse, it offers a potential weapon to enemies of a country, including small groups and terrorists, if they can manage to take over or compromise the central control. They might even be able to suborn the central control remotely through computer attack, or through the presence of a small number of inside agents combined with remote computer attack. The potential catastrophe from such an attack is so high that central control should be avoided at almost any cost.
One counter-note: Today, the fact that all human drivers routinely make small errors is used by the police to find an excuse to pull over anybody they wish to pull over. In theory, a robocar's activity should never present a police officer with probable cause to pull over the vehicle, unless the occupant has ordered it to do something unusual and illegal. Police may find other excuses, but this will become harder for them.
Note: An expanded section on freedom and robocars is planned.
I, for one, welcome our new robot overlords
The level of AI required for driving is, as I have written, comparatively minor. Almost all animals can navigate in their own form of traffic, and not hit things. Even insects. So the cars won't be our robot overlords.
However, as robocars become an industry, this will result in more money being applied to AI research than has ever been applied before. The results in machine vision, prediction and low level thinking will have further application towards stronger, more aware AI.
This will bring many positive results, of course. But there are also compelling arguments that there are considerable risks in doing this, and in the much longer term view, they should be evaluated.
The military wants robocars for safe delivery. But let's not pretend, they also want killbots. They want to send a robot to a target and have it destroy the target. Usually under the remote control of a soldier tele-operator, but sometimes autonomously like a land-based cruise missile.
You want killbots for your side in a war, perhaps, but not for the other side. But the other side will have them, eventually. This will change the nature of war, and surely not always in a good way.
This question is discussed in more depth in the roadblocks section. There will be bugs in robocar systems, certainly initially, but they will continue to pop up from time to time. And there will be some occasions, hopefully very few, where the bugs cause injuries and even death to passengers and even pedestrians. Society will have to balance this against the millions killed by human drivers.
If there is too much "monoculture" in the deployment of robocar software, we will also face the possibility of a bug being discovered that is so serious that all cars with that software must shut down or revert to known safe versions.
It is possible for the computer systems of robocars, and the systems transmitting information used by robocars for navigation, could be compromised by attackers, either as pranks, maliciously, or by foreigners in an act of war.
Indeed, the designers of robocar systems should expect that the outer ring of software in the car (such as the built in computer workstation) will be compromised from time to time. The driving system must be kept isolated from too many outside inputs, and those inputs that are necessary must be highly scrutinized.
The driving system itself will need fail-safe "watcher" modules that are not connected to the outside world, highly scrutinized and rarely changed. These watcher modules will be akin to the systems developed to stop human drivers from doing dangerous things with their cars -- they will also stop the robot driver, and bring the car to a safe halt if they detect too much of a problem.
Here is where software monoculture -- most cars running the same software -- must be avoided. Monoculture allows an attacker to discover a single flaw and suddenly attack millions of cars at once. On the other hand, software multi-culture means that various systems will have differing levels of quality and security, which presents its own issues.
Designers of these systems, especially the driving systems and the fail-safe systems, will need to follow the absolutely best principles of safe software design. These include using multiple independent systems which all make driving decisions on their own and then "vote" on what to do. In an emergency, the majority wins. In a non-emergency, if any system disagrees, it's a sign of trouble, and may call for going to a safe mode.
No, they won't be running these cars on Microsoft Windows, or even linux.
Rewriting Car culture
In the USA, car ownership and driving play a large role in culture. Some of that role will vanish, some will undergo dramatic change.
This is not insurmountable -- in fact most of the people of the world don't own cars, and there are many cities, including New York, where most people don't own them or express their identity through them. In places like Hong Kong, for example, where very few own cars, people express themselves in clothing and wearable devices.
Still, many will lament the passing of elements of car culture. They will also lament not having "their stuff" in their car, always available or always loaded and ready to go. While there are a variety of ways to help with that, particularly as other areas of robotics advance, I doubt that anything can truly replace this. Largely, this loss will be contrasted with all the lives saved.
Sport driving -- taking a roadster up a windy mountain road or a convertible down the coast -- should still exist for quite some time to come. For these markets, a sports-car with robocar functionality will allow the car to autopilot the passenger to the fun, rural roads, and let them become a driver when they want to.
Social travel upside
Some elements of car culture might get better. The family road trip can involve the whole family interacting, since neither parent need drive, and some fraction will even accommodate a vehicle where the seats face one another. Of course some kids might also hate this. Today, people rent stretch limousines with a chauffeur and hold a party in the back. With robocars this becomes easy and inexpensive. Every group car trip can be a social event.
This might eventually generate demand for larger "social" vans which allow people to stand up or move more freely, or sit more comfortably. While high vehicles have more wind drag, the sharing of the vehicle should keep this green, in particular because people will hire the right sized vehicle for their social trip.
Empty vehicles and extra trips
A possibly non-green consequence or robocars could be lots of empty vehicles on the roads. For example, a couple who owns a robocar and does not like to use/provide robotaxis might have one person commute into work with the car, then have the car return itself home for the other person's use (or staggered commute), and then repeat the process at the end of the day.
Commuting, which its huge number of one-way trips, all at the same time, presents a problem for all forms of transportation. It either needs parking and lots of vehicles (for the traditional single user car commute) or mostly empty vehicles that return in the anti-commute direction.
There is some argument that since a large part of the energy "use" of a vehicle is the energy used to manufacture it, in some circumstances having vehicles make empty trips can be greener than having multiple vehicles.
Vacation trips (driving out to the lake for the weekend) could also present a problem, and may just require a high level of vehicle ownership.
Fortunately, ultralightweight robocars will be very energy efficient when empty, as they will carry little more than the weight of their own batteries. (In the extreme, some might even reconfigure for a lower drag profile.)
We could also find ourselves addicted to deliverbots that can just bring us anything on short notice. While they will also be extremely efficient -- small, light, electric and low to the ground -- it's not hard to see us overusing them. They will always be better than a person going personally to the store to get something, but we might send out for a lot more somethings.
Vacant robocars/whistlecars and deliverbots will have many reasons to go more slowly than occupied cars if they aren't in a hurry to get somewhere to serve a human. It's simpler and safer, and depending on vehicle design, more fuel efficient. Sleeper cars may also want to deliberately go slowly.
Energy costs in a car are based on engine efficiency at a given speed, rolling resistance and drag. Typical calculations for today's cars place the most efficient speed in the 40-60mph range. More for streamlined light cars, less for heavy cars. Because energy lost goes up with the square of speed, it's very difficult to be most efficient at the sort of speeds humans like to go on the highway. This chart has some examples.
Note that engines can be redesigned to alter this equation (for example to be more efficient even slower.) Electric vehicles with much more efficient engines will be most efficient at slower speeds since drag will be their main issue.
These slower cars will get in the way of human occupied vehicles which are in a hurry. On multi-lane highways this is probably tolerable, and the law might even require the slow vehicles to group or space so as to impede traffic the least. Indeed, they could be programmed to obey requests from human vehicles to speed up and slow down temporarily, and even pull off the road, to make passing more seamless.
Human drivers will find them very frustrating, even if they can issue commands for them to move out of the way.
Robocars will also go slowly on city streets when approaching traffic lights when they know that going faster will just mean hitting the red light, while the slower speed assures a green. While this should not frustrate rational human drivers, it still will. And it will be a legitimate frustration for vehicles that plan to turn before or at the light, and thus don't mind hitting the red.
If history is our guide, there may be too much pressure to make robocars too safe, to the point that this causes problems far worse than the added safety would justify. While erring on the side of safety is not a bad principle, our legal system and our design systems often go far beyond this. This could go beyond vehicles moving too slowly.
Will vehicles all try to stop the moment a pedestrian's foot steps onto the street (as California law actually requires) or will we generate a system of safe jaywalking on low-volume streets?
Designers might be pressured to leave larger margins in traffic decisions even than humans need. They probably will also avoid having margins so close that they scare people, even though the computer can execute them with 100% safety.
Vehicles will probably be programmed to not execute various "unsafe" actions, even if ordered to by their owner. In the real world, we bend the law constantly in areas of human discretion. Strict adherance to the law would not necessarily provide the desired result. Humans are constantly doing things that have real, but low probability of causing an accident, some of which have become part of normal traffic flow. Defensive driving is good, but paranoid driving can ruin traffic flow.
Car industry disruption
Robocars will cause significant disruption in the automobile industry, which is a downside for older players, though possibly an upside for nimble companies.
Chances are total car purchases will fall, simple because people can share vehicles so readily, either by hiring out their own cars, or using taxi fleets. Most cars do sit idle the vast majority of the time, and this inefficiency will be improved by robocars. Of course, cars that run all the time will wear out more quickly than mostly idle cars, but high-use cars can be designed to fit new usage cycles -- as taxis and other fleet vehicles already are.
If people stop deriving a lot of their identity from their car, they may buy even fewer cars. In many cities of the world very few people own cars. I expect it will be rare to give people a ride in your car, since it is simpler for group trips to summon a minivan robotaxi of the size needed. The "soccer moms" of the world may still decide to own vans if they regularly haul around children who must be supervised and thus can't be put into a child robotaxi.
People will still like to impress with the vehicle in which they roll up to social occasions, and will own or hire fancier and newer cars if they can afford them. Those who bought a Rolls Royce didn't do so for economy or efficiency, and they may still insist on always, or almost always using a "status" car, or one customized to their personality.
Some will view cars as commodities, and not care much about how they look. That's how most people treat taxis in taxi-heavy cities -- only a very few try to express status with their taxi by getting a special limo. (In part that can be that limos require advanced arrangement while taxis can be hailed.)
A move to electric cars allows much more flexibility in car design. We may see the availability of standard modular components for cars which allow small entrepreneurs to build highly specialized cars for specialized markets. With electric drive, you can start with a basic chassis and small power-train and have many choices about vehicle shape, placement of seats and amenities and other features. People might even find they can design cars to order. Cars might be designed cheaply on CAD systems and fed into computer controlled fabrication systems to be unique to individuals who wish it.
Car manufacturers might decide to only hire out their cars rather than sell them. Thus it's possible to imagine customers "subscribing" to a premium car line, giving them exclusive access to that type of car on demand, so that they almost always are seen in it.
However, whatever new ideas automotive entrepreneurs come up with, there will be major disruption in the market, and in jobs within the industry.
Robocars will mean a loss of jobs for professional drivers. Taxi drivers and chauffeurs of course, but bus and truck drivers as well. (The risk of piracy may make trucks move in robo-convoys with people along as security rather than as drivers.)
Because of this, the Teamsters union may oppose robocars. Taxi drivers themselves are not a powerful political force, though taxi companies are. Taxi companies who oppose the entry of private individuals into the robotaxi market might use jobs as a political excuse. (Though I suspect taxi companies would end up being middlemen who broker between passengers and private car owners who want to hire out their robocars.)
It is common for new technologies to have a job impact. While people should have concern for such disruption, it is very difficult to advocate slowing a beneficial new technology, particularly one that can save lives, in order to keep people working.
Let's consider when this can really happen.