🐸 ⏹️ 💭 [Forecast] Transport of the future: short-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: the short-term horizon
Part 2: the medium-term horizon
Part 3: the long-term horizon

Unmanned vehicles need to be created as quickly as possible. Teams of engineers and scientists are working hard to bring this project to an end, and the prospects are optimistic.

But there are some obstacles to this. Obstacles generated by society. Especially in the USA, the country with the greatest love for cars on earth. Indeed, despite the fact that most innovations come from the USA, robotic machines may first appear on the roads of other countries (such as India, Japan, China or Germany).

In this text, I will present a list of both existing and hypothetical technologies that, step by step, can help us achieve a world with robotic machines on the road.

What do we have at this moment:

You can not imagine how many technologies of computer assistance while driving are already on the market (or announced).
Anti-lock brakes and computerized traction control have been present in cars for many years
Already now, cars with Drive-by-Wire technology (electronic digital car control system) are popular. Many people predict that there will be more and more such cars.
Lexus LS460L ( ). Citroen C4 Picasso . Ford Lincoln MKS .
5- 7- BMW .
Volvo S80 ( Lexus).
Volvo S60 2010 , , . Volvo , 2020 . XC60 2010 10 , 20 .
Side Assist Audi , . .
Lane Departure Prevention Infiniti M EX , , .
, , . , , , .
Pre-Safe Mercedes . Pre-Collision Lexus , .
Daimler-Benz uniBwM 1994 ( VaMP), 1000 . 1995 Mercedes S- 1600 , 95% , – 158 . .
Opel Insignia GM (, ) .
Audi Travolution , .
Volkswagen , . , , , , .
, . , Dash TeleAtlas.

The features available now are examples of additional technologies that will be marketed as those that make machines safer and more resistant to accidents. Most of the technologies used in robotic machines will hit roads for the first time this way, bypassing military use. Luxury car manufacturers will be able to raise prices and make cars more resistant to collisions, but over time, these technologies will become cheaper and will appear in low-cost cars.

All this is possible, because according to researchAmerican National Highway Traffic Safety Administration, 80% of accidents are caused by driver carelessness. Systems that notice threats that drivers do not pay attention to will warn of danger and will be able to prevent a large number of accidents.

A few years later we will see a virtually “trouble-free” car. This car will be very difficult to break, even if you try to do it on purpose. And this car will not need to interfere with your driving, since it will have to determine the approach to a dangerous situation, from which even a computer cannot get out (despite its excellent perception mechanisms and reaction time). The systems of such cars will prevent human actions that will bring the car closer to the borders of dangerous situations, although most likely it will not work to stop a person who forcibly provokes a dangerous situation. So, for example, when shifting to a danger zone, turning the wheel will be more difficult, but if you put enough effort, the car will obey you. Perhaps for the rest of your life.

The main issue will be how to configure this buffer zone. The closer the driver can get closer to the danger zone without causing the system to operate, the less noticeable is the operation of this system. Ideally, a driver who would never have gotten into an accident would never notice the operation of security systems, because there will be very few situations that a person can resolve and a computer cannot.

Such systems would probably have to ignore oncoming traffic that moves in its lanes. While oncoming traffic can obviously create a dangerous situation (head-on collision) from which no one (neither a person nor a robot) can get out, in our driving habits even the most accurate drivers simply accept this state of affairs. One-way streets and roads with a large separation of sides can solve this problem.

A machine with a similar safety system should be considered fully robotic. In a situation leading to an accident, the driver can simply let go of the control, and the computer will calculate the trajectories of other cars and obstacles, and will do everything possible to make the situation as safe as possible. If the movement is controlled by a person, the computer will obey, but only if the person does not seek to create such a dangerous situation that it will be impossible to get out of it.

The main obstacle for these machines on the way to obtaining the status of fully robotic will be complete ignorance of the basics of traffic rules and navigation. The reasons are short-term factors and disaster avoidance.

It is easy to see how anxious parents pay large sums for such cars for their children, and insurance companies offer owners of such cars big discounts.

At some point, we will have cars that will drive constantly or drive most of the time ourselves, but this will not be fully legalized. These technologies will be allowed to be used as accident prevention systems, because people want a car that can cope with a dangerous situation when the driver falls asleep or faints. With increasing confidence in these technologies, people will increasingly release the steering wheel and give control of the car (it may be illegal).

However, history shows that car safety technologies often simply allow drivers to feel safe and take a greater risk, because of which the accident rate may not decrease or even grow. For such technologies, it is difficult to make any forecasts.

I must note that this technology contradicts one of my statements. I wrote that in the year in which we do not produce unmanned vehicles, human drivers kill 45,000 people in the United States and a million worldwide. Automobiles that prevent accidents can make a difference, because they must reduce the level of mortality, or at least deaths with them can cease to be so.

Some market players may want to stop the development of such vehicles, but almost all the technologies in this market help the development of unmanned vehicles.

By 2020, Volvo promised to release a safe car with an accident prevention system and better protection for passengers so that they are safe in situations where accidents cannot be avoided. They think that interesting findings are in locust behavior .

Difficult navigation

Drivers demand and buy navigation systems that will become prototypes for unmanned vehicle navigation systems. Their functionality includes high-precision A-GPS technology, dynamic processing of traffic data collected from other cars, as well as edits based on monitoring user behavior and reports. Later, such systems will begin to recognize lanes and parking spots, as well as the switching time of traffic signals, congestion, and much more.

There is a whole area called “Intelligent Transport Systems” dedicated to the concepts of networked vehicles with better navigation and traffic control. Most of the work in this area will contribute to the development of robotic machines, since robots are much better at continuously integrating data into the decision-making process.

Simulation of the environment of robotic cars

I have joined people who are proposing to develop an open network platform for unmanned vehicle simulators , which will allow small teams to develop and test software for robotic machines at a low cost, as well as improve open source projects. These software environments can also be used for competitions, the winners of which can receive funding for testing their software on real cars.

Military bot for delivery

Thanks to the DARPA Grand Challenge contests and the Congress’s mandate, to ensure that ⅓ of all military vehicles are unmanned by 2015, military delivery bots will be the first autonomous vehicles . They will be deployed outside the United States, usually in hazardous areas, to move cargo without risk to personnel from enemy attacks. Some goods will be destroyed or stolen, but costs will generally be reduced.

Some goods (such as weapons) are so valuable that they will be guarded as usual. Such convoys will consist of robots carrying weapons, and heavily armored and bomb-resistant escort vehicles.

Military technology will make many other products for civilian use. In addition, a civilian bot for delivery is a concept that will change the world in itself.

Telemetry and networking

Ideally, robotic machines should work completely independently of any other systems. They should not depend on any central management, information dissemination systems, and should not depend on interaction with other machines. However, this does not mean that using this data will not improve the performance of these machines.

Even before unmanned vehicles are on the road, familiar vehicles can benefit from this kind of information. For example, it is useful to obtain data on the change of traffic signals in order to adjust the speed of the ride so as not to get into the red light. Studies show that if the car moves in this way, and therefore does not stop and does not start from scratch, then the city’s mileage will increase by 30% - the hybrid car gets the same benefit from the use of electricity (as noted above, this technology is already used in Audi Travolution).

We are already close to this. 802.11p protocolhas already been set aside for wireless access in the automotive environment. I proposed a gas pedal option that ignores acceleration attempts that will cause you to get under the next traffic light (except when you insist on accelerating).

Information on the dynamics of traffic flows is also useful for planning the optimal route for the trip. No central government should tell the cars where to go - they themselves naturally will choose the fastest and least busy roads. Cars can announce on the network that they plan to move to certain places so that other cars can avoid these places if, according to calculations, traffic congestion is expected in them.

Telemetry from other machines is especially useful for robot machines. If an unmanned vehicle can know that another car in the stream is also controlled by a computer (which means it has a high response speed), then these cars can massively go around obstacles, pedestrians or treacherous cars with a human driver. Of course, a robotic car cannot be completely dependent on such information, but if you decide that you can trust another vehicle, such cars can be more densely placed on the road to increase throughput.

Also useful telemetry may include data such as reports of potholes, information about traction, wet and icy roads. Cars with a human driver can also use this data - this will increase the safety and productivity of the car.

If you want, you can follow the link and learn more about Motornet .

Parking with a parking robot

Lexus with automatic parking function is already available for sale - this car can perform parallel parking in a small area. Imagine the next step - you allow the car to park on its own in a special parking section designed for such operations. A car in which there is a “parking robot” can carry out commands coming from the parking zone, follow signs or (in earlier versions) move along guide wires. The territory for the parking robot will be free from people (that is, non-employees) and will be privately owned, which will make it possible to use such parking lots long before the robot cars are released onto the roads. Cars in such parking lots will be more like robots that are already driving through factory floors and hospitals.

Cars with a robotic parking can park more tightly and free up travel on command. They can park themselves in garages with limited height. The task of autonomous parking is much easier than a full ride on the streets.

Such machines can also be programmed to form a convoy in which many machines will follow each other. This would allow one person parking person to deliver a large group of cars from a car drop-off point to long-term remote parking at the airport - only one person would be needed to transport a dozen cars. Also, the valet can return with another convoy of cars for passengers who are due to arrive in the near future. This trip should not take place at high speed, and therefore it can be quite safe even on public roads. We need to entrust the machine with only following the leader - something that we can achieve today.

People will gladly pay more for a car that they could fit into the entrance for a roboparker near an airport, office or shopping center and go about their business. Having returned, they will only need to send a signal from the phone, and when they come to the parking lot, they will see that their car is ready for departure.

Perhaps the best solution is not to integrate the full-fledged logic of the parking robot into each car. Instead, each auto-driven car will be equipped with a standard interface that will allow a robot parking module to be installed in the car. This module, equipped with the latest technology, will have access to any sensors in the car and belong to the parking lot, and this module will do all the work. In order for the owner to be able to provide the control of his car to the roboparker, a new standard (possibly wireless) will be developed.

You can even imagine a robot parking module that works with old machines that have very little electronics. Almost all cars have cruise control and electronic brake control, and this can be enough - if the roboparker has a small tug that will provide steering, while the parker himself will give the car commands about acceleration and braking. If braking can be controlled on wheels independently, you can only steer with brakes.

In extreme cases, you can build a full-fledged towing module, on which you drive two wheels of the car (as in the case of a tow truck) - this method is suitable for cars that are not equipped to work with the protocol of the robotic parking.

A more advanced version of the robotic parking - a car that itself arrives at the place on demand - may appear later and provide many of the advantages of unmanned vehicles.

We could also see more cars designed for compact automatic parking. An example of such a machine is the MIT City Car project .

This step will soon be marked "Implemented." The robotic team at Stanford modified Volkswagen (a project known as Junior 3) so that the car could find a free parking space and occupy it with a site map in mind.

Personal auto-like car

While in other articles you can see my statements that personal automatic transport (PAT) will become obsolete before it is widely used , the earliest successful projects in this area actually go along the path of unmanned vehicles. The parking shuttle at Heathrow's fifth terminal, called ULTra, is a rubber bus PAT. ULTra moves along a designated lane with small borders - it cannot detect pedestrians and avoid them, and also follow any route other than the designated route. This system was put into test use in 2010, and in the near future it is planned to switch to full-fledged general use.

Abu Dhabi is building a new city called Masdar near the airport. Masdar is planned as a city without cars, so they are building the pedestrian part of the city one level higher. At ground level (which, it seems, is at the basement level of buildings) there are tracks for robotic vehicles for transporting passengers and goods. Instead of driving along guiding borders, these cars move along magnetic paths buried in the ground and do not meet pedestrians or other cars. However, they do have a laser detector that stops them if they still face a person or other obstacle. This system (and the entire city plan) has been reduced to fit the economic center of the city, but it is currently working and has 2 stops for passengers and 3 stops for goods.

Like all PAT systems, such systems require dedicated travel. However, they can be a good demonstration and help people learn more about robotic machines. These cars drive on regular tires and can use the usual infrastructure as they develop, while PATs are tied to their tracks.

Airports are one of the areas where PATs have the greatest chance of success. Heathrow’s example could encourage other airports to take advantage of even more advanced technologies, such as robotic cars that follow paved paths and possess lidars so as not to harm pedestrians.

Teleparker

A possible alternative to a roboparker may be a teleparker, which will include a set of cameras that provide 360-degree visibility from the cab. The image from these cameras will be transmitted to the remote driver sitting at the console and driving the car in the parking lot or along the parking routes - almost like in a video game.

In fact, it is not so difficult for an electric car if the speeds are low and the car is equipped with safety sensors that will stop it if it approaches something (although such strikes can end only with scratches on the body without exposing people to risk).

The teleparker, like the robotic valet parking, provides more efficient parking, especially in large offices and airports.

The military also has a history of remote-controlled robotic vehicle technology. Flying cars (UAVs) are now in fashion, but long-range ground vehicles will also be in demand (the main problem they are facing is a reliable control signal without a reliable radio frequency data network. They can solve this problem with airborne transponders and limited self-government upon loss of signal). Of course, the military would also like to equip such vehicles.

In the event of a network connection failure, the vehicle must be able to maintain a steady state, and then quickly slow down and stop to ensure safety. Today's network technologies are not reliable enough for this, but work is being done on this from the other side.

Continuation 2:

1:
2:
3:

- automotive . 2500 , 650 .

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

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

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