🀄️ 👩🏿‍✈️ 🤠 [Forecast] Transport of the future: long-term horizon 🚦 🍰 📜

About the Author: Brad Templeton is a software engineer, an evangelist of robotic cars since 2007, and worked on Google in his early years. Founder of ClariNet , honorary chairman of the Electronic Frontier Foundation and director of the Foresight Institute , founder of the faculty at Singularity University .

Part 1: short-term horizon
Part 2: medium-term horizon
Part 3: long-term horizon

Long term technology horizon

Fish School Testing

To convince skeptics of the safety and vitality of robotic cars, I believe that they need to go through something like what I call the fish school test. If you ever swam next to a school of fish and tried to touch one of them, then you know that this is almost impossible.

Our cars should be as smart as these fish.

A swarm of robotic machines will move on the test track. The skeptic will be given the keys to an ordinary car - a sports car or Hammer (as he prefers) and will be offered to crash into one of the cars. Robot machines will pass this test if the driver cannot crash into any of them.

Then the demonstrators will show that you yourself can ride inside this swarm of robotic machines, and not one of them will crash into you. In fact, you can’t even touch them, no matter how you accelerate, slow down, cut them, and so on.

How will these machines handle this? They will need to identify all vehicles, people and objects on the road. They will also have to predict where this person or vehicle may shift in normal and abnormal behavior.

A skeptic's Hammer will be identified, after which all possible trajectories along which he can move will be considered. A robotic car will never put itself in a position in which it (even with its excellent reaction time) cannot escape the situation created by the strange behavior of a human driver.

Given a very good reaction time, this is not so difficult. A robotic car that can notice that the vehicle in front of it is slowing down (by brake lights or not) can react within a few milliseconds. A person needs about 750 ms.

If the car can react within 20 milliseconds (1/50 of a second) and can brake just like the car in front, then after full braking it can save a gap of 1 meter from the car in front. I would not want to figure it out so accurately, but this should give you an idea of what we are talking about. The same goes for acceleration.

Difficulties arise in the case of heavy traffic. In such circumstances, you may know that you have the opportunity to brake or turn off if any vehicle makes a sudden turn, but if there is another vehicle blocking your path, you may be blocked if you are not sure that it too will do the maneuver. To help in such situations, a reliable protocol will be used, with the help of which cars can make requests to other cars and find out whether they will turn together, or to observe a tight location on the road. Please note that the vehicle will not need to use this protocol to ask another car to turn - the goal is to know that your car can count on turning another car if it seeswhat you turn (not that it was harmful to talk about it on the radio).

Of course, all this is exactly what schools of fish do. Each fish knows that everyone else will turn with it. These are their instincts, and trust comes from the simple knowledge that your fish brothers are with you. In the case of robotic machines, trust (if necessary) can be established using the reputation and certificates of suppliers. If the car near you does not have an agreement on trust and reliable management, move away from it. With pedestrians, everything is more complicated. Theoretically, a pedestrian can enter the road at any time (in California, crossings in the wrong place are illegal, and technically, if a pedestrian goes on the road - not just at the crosswalk - all cars are required by law to stop. In fact, this law is never respected and never applied )

However, by the standards of computers, pedestrians move slowly. And as long as we don’t have a very tight arrangement of cars on the road, cars, like a school of fish, will leave significant gaps so that they can go around pedestrians like fish around a diver.

Then skeptics can agree with such cars on the road.

I wrote a more detailed analysis of the fish school test .

Supervision of human drivers

As the technology of robotic cars improves, they will be installed in cars that are still driven by humans. In tens, or maybe hundreds of thousands of such cars (at least, technologies related to sensors will be actively involved, some of them are already used to prevent accidents).

Driving information and sensor data will be recorded to create an ever-expanding set of test data for robotic vehicle systems. The brain of a robot car will reproduce data from sensors and make decisions, and these decisions will be compared with what human drivers did. Situations in which robotic machines will make bad decisions will be corrected, and if the machines do better, human data will be corrected.

Soon, every robotic car system can be tested as if it had traveled billions of miles in the real world, including many unusual situations, crashes, all kinds of weather and transportation problems. Test sets will grow and grow (since it will be wrong to just create a system that will pass a fixed set of tests, such systems should always be tested on new sets). Fortunately, human drivers will provide a huge amount of new test data every day about their actions and real-world situations.

Driver surveillance

As confidence in technology improves, roles change. The brain of a robot machine will gain control of the machine, but a person will have to follow and be ready to correct system errors. All this data will be added to the test suites.

Soon we will be able to evaluate the systems of robotic machines and find out that they can travel billions of miles without errors, faced with the conditions of the real world. This will not mean that these systems are ideal, but we can trust them to manage without constant monitoring.

Over time, the test suite will become larger and larger. When any of the systems of robotic cars makes a mistake noticed by a person, all systems of such cars should be able to check how they cope with the situation that caused the error, so each error is made only once. Each new updated version of the software will experience travel across billions of recorded miles in the real world.

Special lanes for robotic cars

Although I do not believe that it is practical (except as a temporary measure), we could see the development of special lanes for robotic cars on highways. These lanes will be similar to the bus lanes that we see today. Indeed, such a test strip was created in San Diego.

Some expect that these lanes will have special markers for robotic machines, or wires buried in the ground along which the machines can navigate. I do not think it is necessary. Conventional stripes with clearly distinguishable markings will be enough for modern technology.

Rather, these lanes could prohibit the passage of ordinary cars with live drivers while they have robotic cars. This can allow the formation of convoys that are almost as good as carpooling when it comes to fuel efficiency.

Similar to lanes for cars with several passengers (for example, there are such in Los Angeles), they can be physically isolated from normal lanes, just to make people feel comfortable in the early days. Of course, if such roads were built, wealthier people would rush for cars that will drive faster, and in which you can read or work while traveling. The main obstacle is the issue of the price of creating dedicated bands.

Admission of robotic machines to all lanes of the highway

As soon as people get used to the safe operation of robotic machines in designated lanes, the doors of other lanes with human drivers will be open for such machines. Driving on the highway is a pretty easy task compared to driving around the city. As with cruise control, people will drive to the highway on their own and turn on the “autopilot” mode. Robot cars are likely to look for other cars similar to them, and go by convoy, and if there are no such cars nearby, then new "trains" will be formed.

Sleep machines

Another type of first robotic machine could be a "sleep machine." It would be a car with beds instead of seats, designed for long trips on night roads at low speeds. As noted earlier, low speeds provide much better fuel economy.

Users of such machines can climb in the family with them as a family in the evening and wake up on Saturday morning at their dacha. Instead of leaving at 4 p.m. on Sunday, they will leave at 11 p.m. and arrive home by 7 a.m. the next day, without losing a full night of sleep.

Such cars should be quiet and ride very gently, and not everyone could sleep in them. Personally, I could never sleep well in transport. But some do it very well. Since it is difficult to realize belts for a lying position, other safety systems will be required (the walls of a car in which windows are not really needed can be inflated airbags).

Theoretically, sleeping machines can be used to travel to work, which allows those who want to sleep less at home, leave for work earlier and fill up on the road.

Street Robot Access

Finally, a great day will come in some city, somewhere in a world where there is no fighting. Robot cars will be allowed on some city streets. First, they will ride under the supervision of a human driver who is ready to take control at any time. But over time, if the engineers do their job right, people will find that they don’t need to take control. Then the future of urban robot cars can begin.

This can only happen with a limited set of streets, like a pilot project. Or robotic cars may first gain access to a carriageway set up separately for express bus or street cars. Perhaps such cars will only be allowed on special lanes or slower streets. People may have to alternate between driving on autopilot and self-driving. First, such cars will have the usual controls and will require a driver with permissions.

For some time.

Ultra lightweight

As the safety performance of human-powered robotic machines and cars that use accident prevention technologies grows, people and security agencies will need to get used to much lighter vehicles. They will no longer feel that they need a giant steel cage in order to be safe.

All this opens up many interesting possibilities in the design of vehicles and, of course, this means more energy-efficient vehicles. One-person cars may look more like an electric fiberglass tricycle rather than a familiar car.

Such vehicles can be cheap and super-efficient, not only because they are lighter, but also because they do not need a large range of travel, acceleration or speed - they are simply used for short city trips (that is, for most trips).

In the picture you see Aero-Rider, an electric tricycle that you can buy today. It produces about 300 BTUs per person mile - 10 times better than a car and 3-4 times better than the most efficient transit systems.

Please note that modern ultra-lightweight designs are not stable enough at higher speeds, although lowering the passenger’s position, lowering the battery’s weight and controlling a computer that can move the battery’s weight can solve this problem.

Robotaxi

Robotaxis, which I talked about a lot in the main article, can lead to a real revolution. They can go on the road almost immediately, but only for passengers with a driver’s license, since I expect that these will be the requirements of traffic rules for several years.

As soon as robotic cars can transport passengers who cannot drive, the robotax revolution can begin in earnest.

Claims against drivers

Around the same time, I expect that we will begin to see statements of claim (in court cases related to car accidents caused by humans) filed because the drivers decided to drive the car on their own rather than using autopilot.

All this may be a continuation of previous lawsuits that will concern car accident prevention systems. The cause of some accidents will be called the lack of these systems. Courts will have to decide what that means.

Streets for robotic cars only

As the number of robotic cars increases, and their ability to organize efficient transportation with less traffic congestion becomes apparent, we will see the impetus for some streets to be allocated only for the passage of robotic machines, at least at certain times of the day (for example, during mass travel to work).

The rails?

Some are wondering if robotic machines will ever dominate the world, can we return to the rails so that there will be vehicles that can move along both ordinary and railroads? If such rails are much cheaper to manufacture (including the additional cost of operating vehicles capable of using these rails), you can see new roads (especially overground or underground) that will be used for quiet and inexpensive trips. In this case, the rails will not be used to avoid the use of complex computer systems, they will be used because they are cheaper to stack than the roadway.

Zones for Robot Machines Only

Just as many urban areas consider or impose congestion taxes to limit traffic during the day, we are likely to see movement to allow traffic in certain business districts and other areas prone to traffic jams only for robotic machines. Living drivers will need special permission to enter these areas. People can reach the borders of such zones and call robots.

Cities for robotic cars only

After a number of years, I expect entire cities to ban human driving, or at least not to encourage it. Human driving will still be allowed on rural roads and some tracks in order to satisfy those who like sports driving for some time. Also for motorsport and sports driving, private tracks will have to exist.

- , . , , -.
- – .
, – , . .
– . , .
- - , , .
( ) - . .
- , 10 100 .
, , , . ( , , ) .
- . . - ( ). .
, - . , .

1:
2:
3:

- automotive . 2500 , 650 .

, , . ( 30, ), -, -, - (DSP-) .

, . , , , . , automotive. , , .

[Forecast] Transport of the future: long-term horizon