Why Don’t You Have A Self-Driving Car Yet? Part Two Outlines Some Social Problems

Many people are disappointed they don’t have or ride in a self-driving car yet, expecting it sooner. In part-one of this two part series (with videos) some of the key issues standing in the way were explored. Here in part two, some other issues, many of which are logistic and social, are laid out.

Road Citizenship

Making vehicles safe, and proving it, is one thing. Another important skill might be called “road citizenship,” namely being a productive driver on the road, not getting in other people’s way, not impeding traffic, not acting unpredictably. In the early days, all robocars tend to be programmed to drive conservatively in order to stay safe. There is almost a dial you can turn between being conservative and safer and being assertive and not slowing traffic. The natural instinct is to turn that dial to “conservative” when you begin, but it can’t stay there forever.

You might make a very safe car but it can’t deploy if it’s always getting honked at or blocking lanes while it waits for the safest moment. A notorious problem for many robocars, and also for humans, is the unprotected left turn. There’s a lot to worry about, including oncoming traffic, traffic behind you, pedestrians entering crosswalks and more. Waymo has been famously criticized for trying hard to avoid even doing these turns, to the point that it will pick a longer route with 3 right turns to avoid that unprotected left, even when it’s not a good strategy. Cruise just had their first injury accident in an unprotected left when another car was speeding. Senior citizens have higher accident rates as they age, and a lot of those new accidents are during unprotected lefts.

Because others want to make the turn behind you, you can’t dawdle here, nor can you in many other places. There is speculation that some of the accidents where robocars have been rear-ended were partly caused by the car pausing too long in a way that was unexpected.

In some cities, you won’t get anywhere if you’re not assertive. Waymo found over 10 years ago that you had to be assertive at 4-way stops or you would not get through them. Tesla got in trouble with regulators when they allowed their cars to do slow “rolling stops” at empty 4-way stops, even though that’s not dangerous and most humans do it. MobilEye has built an entire planning methodology it calls Responsibility-Sensitive Safety, or RSS, which defines the driving options a car has that keep it legal and responsible while allowing it to be more aggressive.

And in many countries, we know that almost all human drivers exceed the speed limit, and those who don’t become a barrier to traffic. Companies are very reluctant to deliberately program cars to break the law, even if breaking the law is needed for good road citizenship — and that’s a problem.

Sorry, you live in the wrong town

While these problems sit in the way of many teams, it should be noted that robotaxi service has been deployed now in several towns. Waymo has had full-time robotaxi service, with nobody in the vehicle, deployed in one suburb of Phoenix for several years, and they just expanded to the downtown and to the non-downtown parts of San Francisco. Cruise operates with no safety driver at night in the same area of San Francisco, and will soon open in Austin and Phoenix. Several Chinese companies have service (though there is an employee not in the driver’s seat) in Shenzhen, Beijing, Shanghai, Guangzhou and other Chinese cities. Baidu just removed the safety driver in Wuhan and Chongqing. Several other companies have had pilot services (with human safety drivers) in a variety of towns, and more services are planned for the near future in places like Dubai, Miami, Las Vegas, Munich, Tel Aviv and more.

These are pilots, but even after all these companies go into commercial deployment, which they hope to do within just a couple of years at most, you won’t see service everywhere. It’s actually expensive to deploy a robotaxi fleet in a town. Not only do you have to buy or build thousands of cars, but you must test and certify that you can operate well in that town, and make nice with local officials, and maybe do some marketing. It takes money and management time. Even companies like Google, Apple and GM, who have more money than several small countries, can’t deploy everywhere in the USA, and certainly not the world, at once.

It will be done a few cities at a time. How quickly it comes to your city will depend on how easy it is to operate there, how friendly the government is, what sort of weather you have, and how good a business serving your town will be. If you live in the country, a robotaxi service isn’t coming until the 2030s, probably.

You can ride one, but can’t buy one

Robotaxis may exist in some towns, and come to more, but you can only ride in them, you can’t buy them. In fact, all the leading self-driving teams are trying to make robotaxis, not a car that you can buy. Tesla hopes to offer a self-driving function for people who buy their cars, but they are very far behind the leading teams – they have bet on hoping for a unpredictable breakthrough that drives on any road and can do it with just the cameras they put in cars back in 2016.

Cruise also announced they hope to sell cars to consumers with some function in 2025.

Driving everywhere is really hard, so hard that the smart teams don’t think it’s the goal to try for first. So hard that nobody is even remotely close to doing it. A robotaxi drives a limited area that you can test and verify. It will never drive down a street you’ve never seen before. It comes home to the depot every night and gets updates if it needs them. It’s much easier to make things work in that environment than to make a car you sell to a customer, never to see again.

This is a problem for traditional automakers who sell cars to customers. They can’t make a car that just drives itself in a few cities. Nobody will by the “Chevy Tahoe” if it only works at Lake Tahoe. No one city or state is a big enough market for a car model from a big car company.

Pick up and Drop Off

Most teams focus on driving, but cars have to also handle picking up and dropping off passengers, which is not just the work of Pikov Andropov, the Russian chauffeur on “Car Talk” but an essential part of giving rides. I call it P-U-D-O or PuDo. While doing this is not rocket science, it involves a lot of detail work, understanding the curb and every spot on it, as well as driveways and parking lots. The road is actually simple compared to those, and each bit of curb has its own rules, and each private lot can have its own signs. Even human drivers have trouble with it. Robocar teams have all started by first handling driving, leaving this to later.

Cruise launched their service in San Francisco without doing PuDo. Operating only at night, on quiet streets, they just stopped in the lane to pick up and drop off passengers. They aren’t the only ones, it’s not unusual to see cab and Uber drivers do the same thing. The city didn’t like it though. They want companies to solve this before they deploy. Some companies solve it by only doing PuDo in a limited set of spots, you may have to walk a short distance to get your ride. For full deployment all this detail work has to be done.

Business Model

Pilot robotaxis are on the road, but before they spread around the world, companies have to settle on a first draft of a business model. All of them have started with Uber or taxi style service, pay a fare, mostly per mile, to take a ride. That’s a start,and can even make money, but they didn’t invest tens of billions to be a cheaper Uber. In fact, those that are charging money for rides today aren’t doing that to get revenues – the money charged is a drop in the bucket of their current costs. They only do it as a dress rehearsal, to see how customers use the product when they have to pay for it.

The final business model might be a cheaper version of Uber, but it’s more likely they will try other approaches like subscriptions or mixed fees in order to get it right and to convince customers to give up owning at least one of their cars and replace it with robotaxi service. That’s where the real revenue is.

This isn’t a real blocker, though. They can get the cars out there without the right business model, they just won’t make money the way they hope to.

An app and other logistics

Making a usable robocar involves some UI in the car, but it also requires a mobile app, so people can summon cars, give them new instructions and pay for them. That’s not a great challenge, it’s similar to the work of creating the Uber system, but it’s not nothing either, and you need this before you can use it.

Making it too safe

Making the vehicle safe was the first challenge teams went after, and some have achieved that. Proving that it’s safe was the next challenge which is still not complete for most teams. But one of the barriers is they want to prove it’s reached a level of safety that is actually too high. By shooting for the moon, they are slowing down deployment. You may ask, how can a level of safety be too high? Perfect safety isn’t attainable, and aiming for it is a fool’s errand, but there is much debate about what’s enough. Everybody says “safety first” but in fact without functionality and economy, there is no product to create the safety.

Developers are afraid of causing harm, both because they are generally good people, and because causing accidents could easily derail or even destroy their project. Lawsuits over robocar accidents will be far more costly than ordinary car accidents in most cases – so much more expensive that they could erase the benefits that come from having fewer accidents. There is a giant insurance industry that has streamlined and reduced the cost paid in accidents. Some would say it has done this too well, but accidents caused by robots owned by deep pocketed corporations will not get any of that reduction.

Public and Government Acceptance

In some places things are being held up by both legal issues and public acceptance. You can’t deploy unless you’re confident it’s legal, and even then, you’ll have problems if you think the public will reject it. In many states in the USA, states were eager, even proactive, to declare that at least testing of vehicles was legal. (It started off legal, because of course nobody thought to write “no robots” in the vehicle codes.) Since then there’s been a wide variety of approaches by states and countries to the legality of deploying real services. Most people know that robocars will change the world, and for the better, and they don’t want to be late to the game, so for now governments are trying to move quickly to define it as legal. This will happen more quickly in the USA and China than it does in Europe, though.

A surprising opponent may be transit agencies, who usually see all transportation through the lens of public transit and view transit as a goal, rather than a means. Some see robocars as competition, as Uber also has been, and rather than reacting to the competition by improving and out-competing, as government agencies they may fall to the temptation of impeding things with regulations. This happened to companies that made vanpool services that took riders away from transit, then were pushed out of business.

The other risk is pre-regulation, which is to say attempts to write rules for robocars before they are even deployed on the road. The normal history of new automotive safety technologies was to allow them to be deployed for many years before regulations were written, and the eventual regulations mostly just required all the laggard automakers to start including these great new things like seatbelts, anti-lock brakes and collision avoidance technology. Even the best teams don’t know the final form of their product yet, and the regulators certainly don’t either.

The public has to accept you too. Some of the public are eager and some are afraid. In Waymo’s early territory, there have been a few cases of the public being more than verbally resistant, but that might change if teams aren’t careful about their public image. Even so, this does not seem a problem. The public is amazingly accepting of new technology like this, trusting it even before they should.

Non-blockers

There are some tasks that need to be done but they are not blockers to early deployment. People are working on them, but they don’t have to complete them before you get to ride in a robocar.

One area, which appears up a lot in press demos because it shows well, is the design of special interiors and functions to do while you ride. People like to dream of a car of the future that looks quite different from 20th century cars, but you don’t actually need this. Teams are trying hard to figure out good user interfaces, and several companies are even custom designing new future vehicles from the ground up. We’ll like these things in the future, but for now people will be content with the standard comfort of the back seat of a taxi today, just staring at their phone. That’s what everybody does now in the back of Ubers and on any transit vehicle, and it gets the job done.

It’s also not needed to make a car that can drive everywhere. You might want that in a car sold to end users, but a robotaxi service only needs to serve a commercially viable set of routes, and it need not even be that viable during the growth phase. That’s why teams are putting focus on places without snow, or subsets of towns. They can go live without going everywhere, and get more places later.

So, when?

The answer to when you’ll ride is June 23rd, 2023 at 4:14 pm Pacific Time. Well, no, and anybody who names a date is foolish. The real date is different in every city, and depends on these blockers.

All this said, for many of you, it many not be too long before you ride in your robotaxi. In the world’s lucrative warmer cities, you might see that now or by the mid-2020s. You’ll certainly see it on a trip before long. If you have snow, don’t worry, many are eager to solve that problem soon after as there is too much of a market there.

You will also soon see regular cars you can buy that can handle all the main highways. That’s a useful, if not world-changing product, and would give a lot of time back to commuters and intercity voyagers. And who knows, maybe Tesla or somebody else will get that breakthrough they are willing to let loose on a road they never tested on before.

These barriers all sit in the way of the teams working to get you in a robocar. However, they largely don’t require hard breakthroughs (other than those seeking to use only computer vision.) The investment money is there and it makes success probable. You should expect to see more pilot projects arise, leading to a ‘”land rush” in the mid-2020s as more cities are claimed by different teams for robotaxi service. Consumer cars should get the ability to handle freeways and arteries, though one that can do all the streets of every town in the USA is some distance away. Before long, you will ride.