Governments, the Law and Robocars

I have often been asked by various government officials what they should do to usher in the era of the robocar. Most are interested in being forward thinking, both promoting a life-saving and beneficial technology, and also trying to make their jurisdiction be a world leader in new robotic transportation -- ideally attracting companies and jobs.

The answers to this question are complex, and non-intuitive, at least to those in government who have seen their role as being a provider of infrastructure, as well as transportation services in the form of transit. The idea of a bottom-up private revolution in transportation driven by internet thinking and Moore's law is somewhat foreign to this world.

Auto manufacturing is a large part of the Hungarian economy.

The following are good roles government can play in this revolution

Encourage development and deployment of the technology in your region. Help small innovators and give developers sufficient assurance that they can sell their technology so that they can raise investment.
Plan for new structures of transportation and resulting new styles of urban development and land-use as they become popular.
Make minor changes to infrastructure and the procedures for modifying the roads.
Protect civil rights.
Plan for job disruption as professional drivers are displaced.
Should they, in the future, present a danger to other users of the road, take steps to use regulation or the threat of regulation to reduce that risk and protect the public. Don't code this into law until a problem is understood.

Don't help too much, at least not yet.

Much of this advice will relate to things that should not be done, or things that should be delayed. It is important to realize that while many, including myself, have prognostications on the future implications of the technology, even those of us building the technology are not yet certain what the cars will look like even 5 years from now. We know there will be many changes and surprises.

This means you want to delay certain types of plans as much as you can. Do your planning and make your changes using the knowledge of the future, not the ignorance of the present. If you feel the need to do it now, evaluate if it will really be very difficult to do it later, particularly when things like computer and communications technologies will be better and cheaper in the future. (That's something that I can guarantee.)

Delay writing laws

If writing laws, don't try to define very precisely what the vehicles are, or what they are capable of, or how they will operate in the law until they actually materialize in a usable form. Research vehicles and the vehicles for the public probably are quite different.

If writing laws, be careful you don't ban one type of technology while hoping to enable another one. Creating laws in the future for future technology is much easier than undoing incorrect laws that were made in the past. For example, there have been proposals to write regulations governing testing of vehicles, or operation of vehicles that need a human ready to take control, and some of these proposals inherently make unmanned operations illegal because they don't have that human. If the unmanned vehicles (which are where the true value lies) come sooner than you expect -- and some are already here -- some bold lawmakers have to take the risk to actively reverse the ban, rather than just making new regulations when we learn what the tech actually is. The problem is nobody wants to be the one to be blamed for reversing a ban if something goes wrong, even if what went wrong is clearly not their fault.

You might think you're doing the vehicles a favour if you define special lanes for them -- but such rules could easily lead to them not being allowed in the ordinary lanes.

Instead, write laws that enable things that existing laws forbid. Write laws so that innovators can clear away uncertainty over whether their products will be legal once they get them up to the desired safety levels.

Traditionally, regulators have not regulated technologies before they existed, or defined safety standards before there are examples of safety failures to draw on. Regulation comes only after there are problems, and we understand what they are. Even then, regulation should focus on the true goals, and not trying to regulate the means to those goals. Regulators and legislators are held much less accountable for failing to prevent the bad consequences of future technologies before the technologies arrived and the consequences were proven. To regulate in advance would actually be a departure from past practice.

If safety is the goal -- and that's a good one -- then builders of the technology should be free to do just about anything as long as they can meet the safety goals. Don't tell them how to do their job, certainly not before they have even figured out themselves how they want to do it.

Safety Regulations

I have a detailed article on safety regulations. It advocates a light touch, more what not to do, at least for now.

Don't use laws at all

Normally on the roads, one thinks of using law and regulations to promote safety. When people drive and make bad actions on the road, you need to make a law to discourage those bad actions. This is not the case for robocars, because there will only be a very modest number of teams building robocars, and all their cars will drive the way they program them to. Robocars are not driven by individuals, they are driven by companies and teams. You can get all the developers in a room (or online) and talk to them face to face. If there's a problem, you may be able to just convince them a different approach is needed. They may not even need much convincing, as most companies are quite afraid of liability risks. If they don't agree, you can move to changing laws or regulations if it's necessary, or just threaten to do so. I am not talking secret back-room conspiracies here -- all this should be in the open. What the cars do on the roads will certainly be in the open. While this is very much at odds with traditional practice, it could be a much more flexible and powerful approach.

Understand the true value of unmaned vehicles

I noted above something which deserves repeating. Many of the big effects of this technology on cities, energy, parking, carsharing, delivery and more come about only when the vehicles can operate unamnned to deliver themselves to users, to store themselves and to refuel/recharge themselves. The lifesaving benefit of superior driving with passengers and the timesaving benefit of recovering productive time are great, but are only part of the story. Take special care to assure what you do doesn't inhibit the deployment of safe empty vehicles.

Don't offer lots of money

In some nations, one natural first response is to provide funding to promote local projects. Contrary to many first intuitions, government funding can backfire for a few key reasons. First of all, it has the ability to scare off private funding, and you need private funding for real success here, even if you hope to start with government funding. When money can be had from government grants and credits, private investors are less willing to compete. If the government gives money without taking equity, it might capture a few of the best teams with such a seemingly attractive offer. Private investors are not interested in playing 2nd string.

Government funding can effectively mean the government is picking winners. And the hard truth is you're not very good at picking winners. That's because everybody is not very good at it, even the professional private investors using their own money, who fail all the time. Those investors are much more motivated -- and also, in spite of what you might imagine, more willing to take risk. Private investors plan for some of their investments to fail and expect it. Public investments will also fail, but the mangers of public funds can be both too tolerant of that, or perversely, not tolerant enough of it. It's a hard balance.

Finally, while state funding may seem like easy money, the reality is often different. It often comes with many strings, and the grant-writing process is complex and arcane to the typical entrepreneur. This distracts them from doing the real work they should be doing. (This is not to say that private fund-raising is not also quite distracting.)

Silicon valley doesn't succeed in spite of the lack of government funding for startups. Many people feel it succeeds because of that lack of funding.

If you have money to spend, it may be best to spend on other things that will enable all players in your jurisdiction, rather than just giving it as funding to the ones you think may be winners. Let's also be frank: Whatever intentions you might begin with in setting up funding, in many places the real result is that money goes to the people with the best connections, not the people with the best ideas. Sometimes there is even corruption. Not in your government of course, just all the other ones.

Hold a contest

The robocar revolution really got its kickstart from one very successful government effort -- the DARPA Grand Challenges. These contests, with prizes of one to two million dollars, created the "garage" innovation revolution. Today, with commercial viability much clearer, it is harder to design a contest, but you might find there are elements of the technology which are not yet commercially viable which could be funded by contests. Giving away a 2 million dollar prizes causes much more than 2 million in effort, and private investors are willing to participate fully in a world where government money is coming through contests. Contest winners are picked through objective performance criteria, not subjective or political judgement.

Contests are underway now in China and India.

Help teams recruit good people

What startups need most of all is good people. Help your local startups recruit people to your location, and in particular help them bring in immigrants. This is also against most intuitions, since governments want to encourage industry to build jobs for natives (ie. voters,) not for immigrants. But skilled immigrants pull much more than their own weight, and result in a net win for the area.

The best example is Silicon Valley, where most startups are founded by immigrants and even most of the US citizen workers came from outside California.

The hard truth is you can't build a world-class center of innovation primarily with your own people. California wasn't able to do it, so why do you think you can? So relax immigration rules, at the very least for skilled people, and make it easy for your companies to bring in the best they can find, anywhere in the world.

What laws make sense

Rather than restricting the vehicles before we learn enough about them, we can look for ways to enable them, identifying roadblocks that old law or tradition are putting in their way.

Many initial laws have put a focus on defining the rules for testing. In many locations, testing is probably already legal, if there's a "safety driver" monitoring the system and ready to take control at any time. After all, if it's legal to take your feet off the pedals and activate an adaptive cruise control, it's really not that different to take your hands off the wheel and have the computer steer, as long as the safety driver is watching and carrying out the duties of a driver under the vehicle code.

It's also generally legal to have unskilled student drivers operate vehicles under the supervision of a driving instructor (or even just a licenced adult) who often has a special brake and the ability to grab the wheel. (Sometimes this instructor has special training.)

There was no law that forbade taking your feet off the pedals and activating a cruise control, and no changes in the law were required to test and deploy these. There is also usually no law which forbids taking your hands off the steering wheel. Laws require drivers to be in control of their vehicles, and in robocar systems, they are still in control, through the cruise control or steering system they activate.

Some proposed laws have asked testers to test first on a test track for some number of kilometers. That's OK as long as test tracks are readily available to small players. If not, this is just a barrier to entry, though a government can fix that by making sure test track facilities with all types of road, including highway, are available at a reasonable price.

Other proposals have included a requirement for an insurance bond of as much as $5 million. This can be reasonable, but research should be done to see if such bonds can indeed be purchased by small innovators. Because this is such a new situation, insurance companies may not understand how to insure it, and this might end up being a major burden on the small innovator. Once again, the driving school metaphor makes sense -- how much insurance does a driving school need and how readily can they get it? If insurance is hard to get, the law might take steps to assure it can be bought -- or the government could provide it directly to promote development of the technology.

Don't ban delivery robots!

If you write testing regulations, make sure to carve out the ability to test delivery robots. If not careful, you could ban them. Delivery robots, which are small, super-lightweight vehicles the size of a suitcase are easier to make safe than passenger cars. They can't injure their cargo, and so don't need all the weight of crumple zones and passive safety systems. They don't need windshields and there is no way for a human to ride safely in one.

As such, any testing regulation demanding there be a human in the vehicle for testing bans real testing of the delivery robots. Padded delivery robots that never go very fast and keep to back streets are actually much easier to make safe, and so might arrive earlier if not for such unintended bans.

Laws governing operation

It's too early to write laws governing actual operation of vehicles which operate on public streets and where a human no longer has to watch the road. These are not yet for sale, and won't for at least a couple of years, so there is no pressing need to protect the public from them. The primary purpose for any regulations around these vehicles is to help developers get investment because they can feel confident they will be able to sell the products. Enable that first, then later, if problems develop, enact regulations to protect the public as needed.

It should be noted that Induct (from France) is currently selling a vehicle known as the Navia, which is highly advanced and in fact does not even have a steering wheel. They are selling it for use on private roads (ie. the campus shuttle market) and it only goes 20km/h, so they can assure its safety among pedestrians. However, there is no technical reason why a low-speed vehicle like this could not operate on public streets or in pedestrian malls very soon, as long as its low speed is tolerable to other traffic. Enabling law for this class of technology might allow it to use public roads in a limited way, for example to move between two segments of a campus divided by a short stint on a public street, but not to drive generally on those roads. Eventually it will be ready for general roads.

See below for notes on speed limits and the vehicle code.

Infrastructure improvements and changes

Let's start first with what a city, state or country might do to welcome the cars and make it easier for them to appear first in that place. Then we'll cover how the city should be ready for the big changes this will produce.

It is true that operation can be made easier for robocars with improvements to infrastructure and the practices around maintaining it. At the same time, the first law of the modern robocar revolution is that you don't change the infrastructure, and must not depend on new infrastructure to drive. The contestants in the DARPA grand challenges had no ability to change the roads, and were thus forced to live with them as they are. That constraint generated the revelation that it was possible to do this at all. Prior thinking had required special custom roads or closed-off lanes, even elevated guideways.

If you're itching to do them, there are some simple infrastructure improvements that are good for everybody, and which, over time, will probably provide value.

Simple Improvements

Lines and road markings are nice if they are maintained. However, they should not be too regular -- it turns out irregularity helps today's vehicles know exactly where they are. Applying minor paint markings in certain areas could be helpful -- or even allowing developers to pay for it.
Road signs could be standardized, but far better, there should be a database of all signs and a well-enforced rule that no new sign or change in rules of the road goes into effect without being in that database.
Construction zones should be well marked, and also should never be created without being put into databases that robocar operators can subscribe to. (Making them well marked is primarily a safety backup in case somebody starts construction without logging it in the database on their mobile phone. Belt and suspenders.)
Traffic signals should be clearly visible from a distance, and ideally they should also transmit their state via radio, ideally to a feed over wider data networks. Since most lights don't change without several seconds notice, very low latency is not essential but reliable information, available from a distance, is. Both new traffic signals, and signals known to be hard to see should be the priority.
Construction equipment and emergency vehicles should all make their locations available reliably to assure all accidents, emergency changes and road changes are known in advance. Place cheap transponders (old cell phones could do this) on all pieces of major construction equipment to have them constantly transmit their location, so the system can spot any surprise construction jobs not in the databases. (Belt and suspenders and tape.)

Robocar specific improvements

(Some of these apply only after wider adoption.)

Buildings and streets could feature places for robocars to quickly drop-off and pick up passengers without impeding traffic.
Inexpensive magnets can be placed in new roads which encode precise locations. While cars will be able to drive without these, extra information is always useful. This can be particularly useful in roads that get snow-covered.

Snow is a special issue. Humans can't even find the lanes in a freshly snow-covered road. If your roads get snow, there are a number of simple techniques which can be arranged, such as special markers every so often which will be seen above the snowbanks to mark the edges of the road, or small, inexpensive magnets buried in the road that can be followed by cars.

Track the technlology, though. It is likely other solutions will be found for spotting the road when snow covered, such as ground penetrating radar, different variations of GPS or others not yet seen. You might end up wasting your money.

Accidents and liability

Robocars will cause accidents, but every team out there is committed to producing a system that has a better accident rate than human drivers. If they do this, it's a win for society, but it's not a win for any given person harmed in the smaller number of accidents which do take place.

Car accident liability is the most common and most understood tort in the world. When a human causes an accident we've worked out very well just what their insurance company will pay to make it right. Nobody expects to be immune from liability, but the key issue is whether makers of robocars will be subject to more liability for the same accident than a negligent human driver would.

Today, you might have your leg broken by a drunk, and in the future it might be done by a robocar. If the maker or operator of the robocar has to pay more than the drunk's insurance company, it might cost more to insure against this than it does to insure people today, even with fewer accidents. This will drive people away from the business. If the cost can be understood, and is similar to today's costs, people can deal with it. This means the injured are compensated just as they are today, but not a great deal more.

The first accidents will be expensive because they will involve much more legal work. But over time, builders of robocars have to become confident it will settle down to something similar to the liability systems and costs we have today. If courts decide that because the responsible party is a car company with deep pockets, and so the award must be many times greater, companies can't enter the business. In the end, that means more injuries will happen at the hands of negligent human drivers rather than fewer at the "hands" of negligent programmers or accidentally misoperating cars.

To solve this, governments might consider liability cap rules which would limit extreme damages in cases where there is not some truly novel factor which justifies them.

There is more detailed analysis of accidents and liability here.

The world has many liability systems. New Zealand, for example, has no liability in auto accidents at all. The government pays for all medical care, and it also provides the auto insurance and fixes both cars. As such, there is no financial reason to spend a lot of effort assigning blame and liability. In the opposite direction, the USA allows juries to assign high punitive damages if the accident angers them. The right answer may lie in between, but for teams building the cars, the most important thing will be removing uncertainty. If they are worried they might see extreme, company-destroying levels of liability, they won't get involved, and money will be hard to raise.

Generally, action on this is not needed immediately, but rather can be done after a problem develops. But consider the example of vaccines. While they saved millions of lives, nobody knew they had personally been saved. Those few who were harmed by them convinced juries to award big damages, and vaccine companies started leaving the business until the U.S. Congress passed the National Childhood Vaccine Injury Act and paired it with a reporting system and compensation system -- a win for society.

The Literal and Real Vehicle Codes

I noted above about how in many countries, it is normal to speed and otherwise routinely violate the vehicle laws. In other countries, speed limits are set to a number more in line with the faster drivers and as a result few violate them. In other countries the rules are barely observed, or worse, because everybody violates them, enforcement is based not on following the law but random bad luck or the whims of police.

There is a risk, in societies where trafic code violation is the norm, that a robocar might end up being "crippled" if it conforms strictly to the law. Developers will work hard to follow the law, because they don't want to get in trouble or get their customers in trouble. But if this means the vehicles have to be so timid that they just can't drive in some situations, or drive so differently they become a hazard, it's bad for everybody -- and deprives all of society of the safety benefits of the vehicles.

In some cases, developers might solve this problem by allowing the human occupants or activators of the cars to make their own choices about obeying the law. That leaves that decision up to people, where it should be. Just as a cruise control lets the person pick the speed, regardless of the limit, so can robocars allow that. Cars might let the operator command rolling stops at 4 way stops, or aggressive merging in places where a timid merger never gets into traffic.

This can work, at least for vehicles with a licenced driver inside, but this is not the ideal solution. There should be a real analysis of how driving works, and it's possible the vehicle code might allow the cars to do things humans aren't technically allowed to do -- but generally do in practice -- as long as the developers have certified it will always be done safely, and that it will not attempt to gain the car more privilege than human drivers typically take.

This will be politically contentious of course. If it's "one law for the people and one for the robots" it won't fly. It must be seen as a realization of what the law for the humans really is, even if something more strict is written down. To solve this, a law could also allow human drivers who have the assistance of advanced sensors to gain the same lattitude. Best of all would be to write the law to conform with reality, but politically that's not going to happen in some places.

Here are some elements of the vehicle codes that might need clarification or improvement:

Robocars should be allowed to "stand" almost anywhere, including in front of hydrants and driveways, so long as they can certify they will vacate the spot as needed. (See below.)
Robocars should be allowed to do a rolling stop when they judge it safe. Full stops are in the vehicle code to require humans to take a pause to assess the situation before proceding, but robocars can easily do this safety evaluation at low speed.
Robocars should be allowed to travel at the speed of traffic, even if that exceeds current posted speed limits, if their makers certify them sufficiently safe for travel at that speed.
If a vehicle needs to move briefly into the oncoming lanes in order to avoid hitting something or get around an obstacle (presuming the oncoming lane is clear) it should be clear that this is within the law, particularly in the former case.
In some cities, drivers must have some aggression to drive at all. For example, there are places where if you don't push your way into smaller-than-ideal traffic gaps, expecting the approaching vehicle to slow down slightly, you will never get into or through traffic, and as such human drivers routinely do this. The law should understand the need to take small risks to participate equally on the road.

Street Parking & Standing

A special section of the vehicle code that deserves attention is the one relating to parking, stopping and standing. A full (unmanned capable) robocar will never truly be "parked" because it can always move if need be. An empty robocar in front of a fire hydrant might be illegally parked under your vehicle code, but in fact if it will always move out of the way if it hears or sees a fire engine, or gets a signal from dispatch, then it won't block the fire trucks, and it won't take up precious regular parking. Likewise a car like this which is blocking a lightly used driveway or double parking may not present any problem if it will always clear the way when needed.

In the more distant future, we might see street sides with two complete rows of double parked cars. The ones on the inside can be regular cars, as long as the ones on the outside are robocars. Should a car on the inside put on its turn signal or start moving, a gap could appear, as if by magic, in the line of cars on the outside.

This ability is going to free up huge amounts of parking and land in your cities -- make sure the law doesn't forbid it.

At the municipal level, codes which require buildings to have large amounts of parking might be relaxed as robocars reduce the demand for parking.

Low speed operations

When attempting unmanned operation, the easiest path to safety is to go slowly. That's easy on private roads, but there are speeds that will be seen as too slow on public streets. Such slow vehicles might frustrate human drivers or even obstruct traffic. Some will push to set minimum speeds or protest slow driving. Other arguments suggest it might be valuable to at least temporarily allow this annoyance to encourage safe early deployment of unmanned vehicles for the purposes of delivering vehicles to users, as well as parking and refueling vehicles.

Slow vehicles can follow several strategies to avoid annoying other drivers. They can limit what roads they drive on, to assure they use only low-traffic, low-speed roads and not main channels. They can limit when they drive to use real-time traffic information to assure they use only low-traffic roads. They might only use narrow "half-width" vehicles for 1-2 people which can easily be passed on the roads they use.

They might also take a strategy of actively avoiding blocking traffic. Vehicles on light traffic roads could detect vehicles approaching behind them, and then quickly move to briefly pull off the road. They could pull off into spots not available for parking, such as brief stays in front of private driveways, in front of hydrants or any other spare space. Less likely to be safe, but still possible in theory, would be pulling into the oncoming lane once it is determined it is safe to do so, allowing the vehicle to be passed.

Whatever strategy is chosen, it might be valuable for the law to enable this early deployment opportunity for unmanned vehicles.

Protect civil rights

Governments are charged with both protecting the public and also protecting their civil rights. The temptation to use this technology as a basis for surveillance should be strongly opposed. The privacy of car and road users should be protected. Developers should be encouraged to develop their products and services to preserve the right of anonymous free movement.

The technology people use should also not be easily legally compelled to act against their interests. Police may seek the power to take control of vehicles and redirect them, or report on what users are doing. This impairs the freedom of people to use the technology and make their own decisions. The transition to robocars should not mean we make vehicles that don't obey their owners or riders. Any loss of existing rights or freedoms should be closely examined and resisted.

Help the little guy

When looking to provide assistance to companies and developers in your area, focus on things that help the small entrepreneur. The big companies can usually help themselves. This can include many of the suggestions made above, like immigration reform and assist in recruiting. It can also mean taking on some tasks that are hard for small companies to manage on their own.

A great recent example can be found in Michigan. There, the state is building a test "track" which looks like some downtown streets. They are putting it on the campus of the University of Michigan. Big companies can afford test tracks (and they are expensive even for them) but small innovators can't make something like this on their own. This sort of facility is easy to share and valuable to all. It also eliminates safety concerns about initial tests taking place on public roads. Even for the best teams, the first tests of completely new systems are better done in a closed test facility.

(The U.K.'s 10 million pound purse for a town to develop a pilot project together with a technology company is different. That's picking winners, not encouraging them.)

It's equally important that you don't write any laws which create barriers to entry for small entrepreneurs. Trust me, the big companies will be very happy to support such laws. Usually, when companies are seen to support regulations on themselves, it's because they think those regulations will be even harder on their competitors, particularly the upstarts. Rules that require a lot of testing on private tracks could block many small companies if you don't have a rentable track such as described above. Large insurance bonds may be hard for small companies to get until the technology is established. Complex and expensive testing regimens can be a complete killer of small products -- if they are needed, they should be demanded only before volume sales, not early in the process.

When you work on regulations, pay close attention to the advice given by the larger companies or other parties with entrenched interests. While they won't say it, think about the effects of what they propose on a small company that isn't able to afford teams of lawyers and compliance officers (or even one such person) or other functions which are annoying and costly but just part of doing business for a large company. "Garage innovation" isn't just a silicon valley phrase. Most of the innovation in cars in their early days came from garage tinkerers who turned into big car companies, and most innovation in software comes from these small companies as well.

In fact, don't just imagine how it will affect small innovators -- make sure to ask them. That includes entrepreneurs, academics, angel investors and small businesses.

Open source software development can also be great for all companies. If you encourage creation of tools that can be freely re-used it helps the whole world, but it helps most strongly the place which is the center of development for such tools.

Vehicle Connectivity

Connectivity will be highly useful for robocars. Primarily this will be connectivity over the same mobile data networks phones use. There is a major push underway in the Intelligent Transportation Systems community around the world to create a system for digital radios in cars that do short point-to-point communications, either between cars and local roadside radios (such as radios in traffic lights) and also directly between cars on the same road, known as V2V (Vehicle to Vehicle.)

This technology will be useful for robocars, but only of limited use in regular cars. However, there are parties who have staked careers on it, and they promote it far beyond its utility. Because of the limited utility, it is the correct, prevailing wisdom that such systems would never be deployed voluntarily, and so a legal mandate must be created to force the technology in all new cars.

Such a mandate comes at a cost -- by some estimates $25 billion from new car buyers over the 20 or so years it would take to put the technology in almost all cars. It should be considered what else might be done with that investment. Further there are a number of arguments, which I won't go into here, suggesting the technology won't work particularly well and its value can be delivered other ways. I won't go into all those here but you can read a series of essays on the topic.

It should be noted that there are many applications in these cars for the existing data networks, such as the cellular data network. Improvements in this network are always desirable. It may also make sense to consider a public data broadcast network. This would be a network on an empty digital TV channel, or in an unused slot in an existing digital TV multiplex. It would broadcast a constant stream of useful data for cars, such as traffic information for all roads, timing information for all traffic signals, reports of construction, lane closures and new signs and even reports (received over cellular data) about road debris, ice, stall cars, potholes and more which cars wish to know about.

Relying on COTS (cheap, off the shelf) consumer parts, bought only by cars interested in the data, it could be a way to get travel related data to cars and other mobile parties cheaply and efficiently, without concerns on line-of-sight short range transmissions.

Regulating Safety

There is a large traditional government role in setting safety standards, or at the least defining NCAP "stars," for vehicles made or sold in the country. It is expected that safety agencies will rate the safety of robocars. In addition, as they see particular safety triumphs from the robocar technologies, they may even begin to mandate some of them in cars.

New safety ideas tend to come from the innovators, not from the safety agencies. Tools like seat belts, air bags, anti-lock breaks, crumple zones and electronic stability control were all developed as safety features for high-end cars in the private market. Later, the safety agencies saw their value and mandated them for all new cars, often to great results. (This is not to say the safety agencies never originate good systems, but simply that they more commonly come from industry.)

Newer tools, such as Automatic Cruise control and the various ADAS systems, including those that keep the car in its lane or hit the brakes when approaching another car or pedestrian, were all developed independently by car companies, and even today are generally subject to no safety regulations. This is something many are not clear on -- safety regulations have usually been written years after a technology has been deployed, and often are no more than requirements for NCAP star ratings.

The reality is all makers of robocars are working hard to make them safe because they know they will face serious accident liability costs if they fail. The need to regulate in advance does not exist. The need to regulate after the fact will be minor, and only if it appears vendors are doing something seriously wrong that is causing unacceptable harm.

I say unacceptable harm because no technology is perfect and there will be accidents. What matters if whether those accidents occur at a lower rate than they do for human driving. Regulations which demand perfection, rather than improvement, will block the industry from going forward and saving lives by creating better drivers.

There are a number of unsolved questions regarding robocars, as you might expect. That they are unsolved is not the same as saying they will never be solved. Some imagine this means there is a pressing need for regulatory solution, but this is a (false) intuition, not an established principle.

Recalls

The issue of recalls is an interesting one. Unlike traditional cars, the software in these cars can be updated over the radio, and will be, often. With a regular car, if a safety flaw is found, the maker normally issues a voluntary recall, and customers can bring their cars to the dealer for a fix. Much more rarely, the recall is compulsory, but many voluntary recalls are done on fear of liability or an eventual compulsory recall.

Once the customer is notified, it is up to them to complete the action and bring in the car, unless the problem presents a danger to others.

Over the air update presents a new capability. A recall can be done immediately, and it can even be forced, which is to say the car can refuse to auto-drive or work at all until the recall is done. This is a hugely powerful weapon, but it could spell ruin for companies if they feel they must act aggressively. They will, on a regular basis, discover various flaws in their systems which have safety consequences. Making fully verified and tested fixes for these problems takes time -- in the name of safety. If the companies feel a need to disable all or some self-driving features until they have the fix, they face the danger of completely destroying customer confidence in their product in exchange for very minimal safety increases. (This is particularly true with unmanned products which are shared, and which people are depending on to be picked up for transportation.)

In the case of the unmanned operation, there is not even a customer to notify and get to accept the risk until the car is delivered. As such, if the safety regulations push companies into regular shutdowns, they may not have a business. Safety regulations must only require such shutdowns in cases of more serious risk.

Congestion Management

There's an opportunity -- though it's several years down the road -- to help reduce congestion on the roads with the help of robocars. In the nearer term, to encourage robocars, consider allowing them in managed lanes (HOT lanes) either for free or for a fee that they can manage and report without any infrastructure. (Installing transponder-based HOT lanes is often quite expensive.) As is now becoming understood, carpool lanes on their own can often increase congestion rather than reduce it unless they are fully utilized. Many regions are allowing electric vehicles and toll-payers into carpool lanes to fix that. Robocars could also benefit from this.

Another option would be to allow robocars to exceed the safe speed limits for humans in the off-hours in these lanes, so long as they are efficient and safe at those speeds.

This can be done to encourage the cars, but also because robocars make congested commutes more tolerable to the riders, which is one of their social downsides.

Jobs and disruption -- and the resulting opposition

Robocars will displace workers who make their living driving. This includes taxi drivers and bus drivers, but also eventually delivery drivers and truckers. While the main goals of developers of these vehicles are safety, user convenience and economical transportation, there is no question there will also be undesired employment disruption.

Regulators can prepare society for this disruption, assisting in retraining and finding new work for former professional drivers. For many drivers, it is not a highly-skilled profession, and there is not a huge waste of educational investment, making this move easier.

On the other side of the coin, professional drivers, as well as some of the companies that employ them, will use legal efforts to fight the threat to their livelihoods. This has already been seen in many towns, where taxi lobbies have fought hard against any new technologies which threaten the taxi business. A typical example is Uber and the other smartphone summoned ride companies. In the past, there have been justifications that taxi monopolies be granted and taxi service be highly regulated. In many cities, only a taxi with a medallion from the monopoly or quota system is allowed to pick up passengers. The smartphone based services got around that by letting people summon vehicles on the street using a phone. In doing so they eliminated the primary justification for taxi monopolies -- that customers have no ability to shop among competitive taxi providers when hailing a cab -- and as such the new technology upends the economics of that industry.

Expect to see taxi drivers, non-robocar taxi companies and others use their legal weight to fight adoption of technology that competes with them. They will couch it in other terms -- they will attack the safety and whatever else is vulnerable, but consider the motives of all who lobby. (Including the makers of robocars of course, though their motives are typically more overt.)

This force has led to somewhat irrational results. Consider Las Vegas, where hotels are not allowed to send their own shuttles to the very-nearby airport to pick up guests, and a privately build monorail runs to the edge of the airport, but not to the terminal -- all because of a strong taxi lobby.

Somewhat later, as delivery robots become popular, you will see another disruption in retail. Small retailers will be hit further, and even big box retailers may fall to online retailers with 30 minute delivery.

Special Values

Robocars can provide more than just safety, convenience and superior transport for the mainstream. Here are a few other areas of government involvement in transportation you may want to consider:

The Disabled & Paratransit

The disabled do not have the freedom of mobility that cars give to others, and they even have trouble using public transit. Robocars change all that, once they are able to do unmanned delivery of vehicles to the customer. The simple provision of door to door transportation that the disabled person need not drive is a revolution for them, but it's also possible to send specialty vehicles for people with special needs, such as vehicles that make it easy to roll-inside on a wheelchair, or which are adapted for the blind, infirm or even those with cognitive problems.

In the USA and several other countries, there are active "paratransit" laws which require the city to provide suitable alternate transportation for the disabled. These services come at high cost to the cities, and are not that convenient for the passengers either. Robocars will change all that, saving huge amounts of money required for both running paratransit and modifying transportation facilities for wheelchairs. This is an opportunity to watch out for.

Transit Planning

Robocars have dramatic potential effects on public transit. It is my prediction they will actually replace public transit outside of a few rush hour services in all but the most dense cities. Public transit is efficient at rush hour, but loses that efficiency off-peak when robocars would do the job. It is also possible to consider self-driving vans and minibuses as an alternative to many transit modes once a driver is not needed.

For more details, see essays on the end of transit and my newest essay with a radical vision for the future of mass transit.

City Planning

The robocar city has quite different needs from the traditional one. It's a city where people can get from place to place quickly and greenly. It's a city that no longer needs most of the parking lots which cover so much of its land. It's a city where living 2 miles from an interesting retail street is like living close to it. It's a polycentric city where there are man CBDs. It's a city with better traffic flow and less pollution.

Urban planners and transportation planners need to be thinking about these potential changes. I have an article about Urban Planning Considerations and a more speculative one on Robocar Oriented Development.

Drunk Driving

A real dilemma exists around impaired driving. On the one hand, somebody drunk riding in a robocar can't be counted on safely to do many of the things you would like the rider to be able to do, like take over control in a strange situation. Because of that, common thinking is that these cars won't be able to carry drunks until they are ready for full speed unmanned operation.

This is challenging because nobody wants to be responsible for facilitating a drunk-driving related accident. Current draft regulations have been written to say that cars must have a licenced driver in them, and that this person, even if not handling the wheel, is "the driver" under the vehicle code. This means that if they are impaired, and they are the driver, they would not be allowed to operate the robocar.

At the same time, impairment is the cause of a very large fraction of accidents and fatalities. It could be argued that getting drunks away from the wheel is one of the best things robocars could do for public safety.

In reality, drunks are bad drivers -- and they should not be on the road -- but most are only moderately impaired. Studies have suggested that texting is as dangerous as driving drunk, which is both a statement about the risks of texting and about the risks of DUI. Huge numbers of people drive drunk with no consequences, even though too many drive drunk with terrible consequences.

We'll face the hard choice when a drunk gets into a robocar. Let them ride, with the enhanced risk of an accident if they have to take control, or turn off self-driving functions and have them foolishly decide to drive the whole route manually. (Having the car refuse to drive in manual or self-driving mode seems like an obvious answer but refusing manual drive politically has only taken place with on-the-job drivers and people with DUI convictions.)

I don't have an answer to this question, other than the suggestion that either extreme is probably not the best answer.

What do do once deployment is in full swing...

The advice here covers issues in a country where robocars are just starting to arrive, and when they begin to reach modest deployment. Once they get major penetration and become a larger share of the cars of the road -- which is probably around 2030 -- the rules change yet again, and there is an ability to eliminate road congestion and even start thinking about infrastructure changes to assist the robocars. More on that, later.

Comments can be left at the blog post.