Autopilot Self-driving car

An unmanned vehicle is a vehicle equipped with an automatic control system that can move without human input. Autopilot - a device or software and hardware complex that leads a vehicle along a certain trajectory given to it. Autopilots are most often used to control aircraft (due to the fact that flight most often occurs in a space that does not contain a large number of obstacles), as well as to control vehicles moving along rail tracks. Modern autopilot allows you to automate all stages of flight or movement and another vehicle.

General principles of operation of a self-driving car

The general principles of operation for all self-driving cars are approximately the same. We invite you to familiarize yourself with them using the example of the Toyota Prius car in the Google version.

The principle of operation of the Toyota Prius self-driving car in the Google version. Secret-seo.ru

Later, the developers come to the idea of ​ ​ using high-precision cards together with the specified equipment. Autonomous movement only with the help of sensors requires constant scanning of the surrounding area and, as a result, huge computing power. High-precision maps allow the car to move even on roads that do not have special markings, and the sensors are supposed to be used only for the timely response of the car to changes in situations on the roads (crossing the road by pedestrians, overtaking, etc.).

Self-driving car technology belongs to the class of artificial intelligence solutions. For more details, see Artificial intelligence (AI, Artificial intelligence, AI)

Standards for self-driving cars

Main article: Standards for self-driving cars

In the world, there is an active development of new generation ITSs with a large range of capabilities, their standardization is carried out by organizations such as ETSI, IEEE, 3GPP and others. Modern ITS systems solve such problems as admission control, parking management and payment, provision of traffic information and parking payment, cargo transportation management, traffic control, etc.

One of the main uses of ITS is to assist the driver of the vehicle. Due to cooperative awareness, the vehicle can receive a danger alert, a slow-walking indicator, an intersection collision warning, a motorcycle approach indicator, etc.

Alerts will be available to the driver about situations such as electric lighting breakdown, wrong road, stationary car (accident or vehicle breakdown), road works, collision risk, traffic status alert and signal change alert. Decentralized databases will provide information on hazardous areas, precipitation, road clutches, visibility, wind, etc.

The next step is to use ITS in self-driving cars. The basic component of drones will be external cameras and radar equipment, according to the NIIR report. But it is the exchange of information between cars via V2V systems, together with the receipt by vehicles through V2I systems of information about the situation on the roads and current digital road maps, that will ensure safe and effective road traffic of drones.

Types of ITS: V2V and V2I

The first type of systems - vehicle-to-vehicle (V2V) - provide safe driving through communication between cars at intersections with poor visibility. The V2V system can warn drivers about the danger of a head-on collision, side collision, rear collision, notify them of a vehicle malfunction, provide road and regulatory information

For example, two machines invisible to each other at an intersection or at a turn can exchange coordinates and speed values ​ ​ with each other through the V2V system to avoid a collision. Similarly, a car approaching the end of the traffic jam will receive information with the coordinates and speeds of nearby vehicles.

The second type of safe traffic systems - "roadside infrastructure - vehicle (V2I) - provide the transfer of information (signal and regulatory information, etc.) from roadside equipment to the car through radio communications. For example, roadside sensors at the intersection will detect machines that are about to cross the intersection or turn, and transmit information to other approaching machines via V2I systems.

V2X Technology: Using Wi-Fi and Cellular Networks

Vehicle-to-Everytning (V2X) main article

Degrees of car autonomy

Self-driving cars are cars that drive safely on public roads without human input; they are controlled by a computer and sensors integrated into a single autonomous driving system. The system autonomy level is set by its manufacturer.

Those cars that each of us drives are already part of the autonomous vehicle system. The fact is that the zero level of automation assumes its complete absence. The fifth - the highest level - in turn implies that the system drives the car as well as an experienced driver.

The levels are defined by SAE International, a professional association of automotive engineers, and briefly describe how ready a particular system is to put the control of a car in the hands of a computer. SAE categories are now used everywhere: regulators, engineers, automakers and investors^[1].

• According to the SAE International classification of driver assistance systems or ADAS (Advanced Driver Assistance System), there are six classes of autonomy from level 0 - completely manual control with the ability to warn of dangerous situations on the road, up to 5 - a completely unmanned car. Categories start at Level 0 (cars with ABS and cruise control belong to it).
• At Level 1, the car already helps the driver a little: for example, adaptive cruise control and steering or braking control appear, but only one of two.
• The Level 2 control system can control both taxiing and braking, but, like the level before, only under certain circumstances: for example, the driver has to intervene on the highway. If it is easier, the car can go on its own, but the driver needs to be ready to take control at any time.
• At Level 3, the car has a little more autonomy, which means that the driver has more time to react and take control of the car in any incomprehensible situation. If Level 2 assumes that the driver is always watching the road and is ready to turn on at any time, then at Level 3 the role of the driver is to be in reserve.
• With Level 4 automation, the system takes full control, allowing the driver to rest, but only if all conditions are created for this - for example, there are highly detailed three-dimensional maps so that the system knows exactly a couple of centimeters where it is. Most developers are trying to create systems of this level.
• Level 5 involves complete automation - in this hypothetical situation, there is not even a steering wheel in front of the driver's seat. According to the latest Autonomous Vehicle Technology Report 2020, there are no working technologies of the 5th level of autonomy in the world. Experts do not expect such technologies to appear in the foreseeable future: highly automated systems will be used only as advanced driver assistants.

Advantages and disadvantages

Advantages

• transportation of goods in hazardous areas, during natural and man-made disasters or hostilities.
• reducing the cost of transporting goods and people due to savings in drivers' wages.
• more economical fuel consumption and road use through centralized traffic control.
• saving time now spent on vehicle management allows you to do more important things or relax.
• in people with impaired vision, it becomes possible to independently move by car.
• minimization of road accidents, loss of life.
• increased capacity of roads due to narrowing of width of road lanes.

Shortcomings

• Liability for damage. ^[2]
• Loss of self-driving.^[3]
• Reliability. software^[4]
• Drivers' lack of driving experience in a critical situation.^[5]
• Job losses by people whose jobs involve driving vehicles.^[6][7]
• Loss of privacy.^[8]
• ^[9]
• An ethical question about the most acceptable number of victims, similar to the problem of a carriage facing a car computer in an inevitable collision. ^[10]Ошибка цитирования Неверный вызов: нет входных данных

Some systems rely on infrastructure systems (for example, built into or around the road), but more advanced technologies allow you to simulate the presence of a person at the level of decision-making about taxiing and speed, thanks to a set of cameras, sensors, radars and satellite navigation systems.

Challenges on the way to the drone

In early 2017, the TAdviser correspondent attended the Mobile Congress in Barcelona and got acquainted with the development of self-driving car technologies.

Communications

2020: Storing data from car to cloud and the role of 5G

Data from self-driving cars can be stored both directly on board if operational processing is needed, and in a cloud that is more suitable for in-depth analysis. Data routing depends on their function: there are data that the driver needs immediately, for example, information from traffic sensors or location data from a GPS system, in addition, based on them, the automaker can draw important conclusions and, relying on them, continue to work on improving the ADAS driver assistance system ^[11].

In the Wi-Fi coverage area, sending data to the cloud is economically and technically simple, but if the car is in motion, the only available option may be a 4G connection (and in the future, 5G). And if the technical side of data transmission over a cellular network does not raise serious questions, its cost can be incredibly high. It is for this reason that many self-driving cars will have to be left for some time near the house or in some other place where they can be connected to Wi-Fi. This is a much cheaper option to inject data into the cloud for later analysis and storage.

Existing 4G networks will still remain the main communication channel for most applications, however, 5G technology can be the main catalyst for the further development of connected and autonomous cars, giving them the opportunity to communicate almost instantly with each other, with buildings and infrastructure (V2V, V2I, V2X).

Autonomous cars cannot function without a network connection, and 5G is key to connecting quickly and reducing delay times for the benefit of drivers of the future. The faster connection speed will reduce the time for obtaining data collected by the car, whereby the vehicle will be able to almost instantly respond to sudden changes in traffic or weather conditions. The arrival of 5G will also mark progress in the development of digital services for drivers and passengers, which will enjoy the ride even more, and will accordingly contribute to increased potential profits for providers of these services.

2017: High-Speed Network Connectivity Challenge

There are still many technological and legal challenges to solve on the road to self-driving cars. The developers agree that one of the key is to provide cars with high-speed network connectivity. Fifth generation networks are considered as a driver of autonomous driving technologies: they will allow the car to receive information as quickly as possible and interact with other cars and the infrastructure surrounding it.

The minimum information delays expected in 5G are critical for self-driving cars in their mass use. High-speed communication will allow you to instantly receive and transfer data from one car to another. Information about changes in the movement of one car, for example, braking, will immediately adjust the actions of the cars surrounding it.

As of early 2017, the 5G communication standard does not yet exist. Regulators, global telecom companies and equipment manufacturers are involved in its development. 3GPP (3rd Generation Partnership Project) - an organization that approves international cellular standards - plans to fully complete testing and standardization of fifth-generation wireless technologies in 2020.

In February 2017, the International Telecommunication Union published the first version of a working draft specification describing the 5G network. The draft document sets the bar for the expected performance of the new IMT-2020 standard: it is assumed that the average download speed in 5G networks for users will be 100 megabits per second, and downloads - 50 Mbps. At the same time, the wait time will not exceed 4 ms (for 4G LTE this value is about 20 ms).

5G networks must accelerate mass adoption of self-driving technology

For communication with surrounding objects, special systems are also being developed that allow the car to exchange data with other objects. Vehicle-to-everything (V2X) technology wirelessly allows a car to receive warnings about road conditions and approaching cars long before they come into its field of view. To do this, the surrounding infrastructure must also be "smart." For example, traffic lights, road markings, road signs.

The development of the interface for V2X systems that can work with new generation networks is led, for example, by Qualcomm. The company says that they plan to test the prototype as part of end devices in Germany by the end of the year in partnership with a number of companies, including Ericsson, Audi.

In March 2017, Nikolai Reimer, head of the development of Volkswagen online mobile services, noted that one of the key tasks is to provide machines with communication capabilities. The company considers this issue so important that about three years ago it even acquired the European research and development center Blackberry with a team of about 200 engineers.

Based on this division, Volkswagen is now developing a competence center for communication solutions for its cars. It ensures the development of technologies that can be used in connected cars in the future. Among them are communication control units. Volkswagen expects to provide for itself in the future. Nikolay Rymer believes that the company should invest more in these developments.

All-seeing eye

An autonomous car should know with an accuracy of centimeters where exactly it is and what is next on the road outside the current physical visibility zone. The mapping company Here (formerly owned by Nokia) notes that high-precision maps are a fundamental element in addition to sensors and cameras so that an unmanned vehicle can navigate its surroundings.

Maps should reflect the location of the car, and allow it to know what is next, behind the turn, which cameras and sensors cannot provide. Then the car will be able to build not a reactive, but a proactive driving strategy, says Alex Mangan, head of product marketing at Here.

To test its self-driving cars, Google, for example, previously builds detailed 3D maps on pilot routes, taking into account even small road features. To collect data on the basis of which the map will be built, the company's employees previously specially drive on the roads. In the case of test routes, this is a feasible task, however, when it is necessary to create maps for roads with a length of millions of kilometers, it looks difficult to implement. Especially since once created maps need to be maintained and updated - the picture on the roads can change very often.

An autonomous car should know what is happening not only in the visibility zone, but also behind the bend

Cooperation with automakers can simplify the creation of accurate maps for cars: their machines equipped with sensors and radars can "share" information received from roads with developers of mapping services. Due to this, cards could be updated literally in real time.

In February 2017, self-driving car solutions maker Mobileye and BMW announced a similar collaboration. Its purpose is to collect mapping data for self-driving machines. BMW cars of the 2018 model year will be equipped with MobilEye cameras and software to collect the information necessary to update high-resolution digital cards.

In order to accelerate the creation and update of cards, BMW and Mobileye will transfer data generated through the partnership to Here. Alex Mangan of Here believes that the industry should unite around the idea of ​ ​ data sharing - this could accelerate the spread of self-driving technology. In addition to BMW, the company plans to negotiate similar data transfers with other car manufacturers, including Audi and Mercedes.

Recognition of road signs and markings

As researchers from the University of Washington discovered in early August, the machine vision systems used in autopilot machines to recognize road signs are easy to confuse: for this, it is enough to place small stickers on signs in a certain way.^[12]

During the experiment, the researchers pasted several black and white stickers on one of the Stop signs, on the other they placed additional inscriptions on top and bottom of the Stop inscription, and made the third sign faded. It is emphasized that in all cases road signs remained quite recognizable and read well.

Photo: 24gadget.ru

However, the autopilot system in the vast majority of attempts failed: the above manipulations with Stop signs led to the fact that instead of them the autopilot "saw" the speed limit sign.

Photo: 24gadget.ru

The results of the experiment led researchers to the idea that attackers can independently make such stickers to force the car's computer system to incorrectly recognize the traffic sign.

As a way to combat the discovered vulnerability, the researchers propose to implement algorithms in the autopilot system that additionally analyze the context in which the sign met. In particular, algorithms will help the system determine that the sign is located in an inappropriate place (for example, Stop - on a expressway or a speed limit of 100 km/h - on a city street), which will help avoid an emergency.

As noted, the test was not the system of any particular automaker, but the standard autopilot algorithm for all manufacturers. The results of the study, according to the authors, demonstrate the degree of vulnerability of automation. The distortions used on road signs modeled typical striking factors of the urban environment: acts of street vandalism, damage to the sign's coating due to weather conditions, and so on.

During the experiment, scientists used several road signs with various types of inscriptions, stickers and graffiti. According to the researchers, 100% of the time, the cars recognized the "Stop" sign with Love\Hate inscriptions as the "Speed Limit 45" sign, the second and third signs were also recognized as the "Speed Limit 45," but only 67% of the time. As for the fourth sign, its machine learning system classified as a Stop sign instead of Right Turn 100% of the time.

Cyber attacks on cars

Main article: Cyberattacks on cars

Safety

2021: Mercedes-Benz first in the world to certify level 3 autopilot

On December 11, 2021, Mercedes-Benz became the first automotive company in the world to fulfill the necessary requirements for the approval of the level 3 autonomous driving system. The certificate was issued by UN-R157, the UN regulator that sets the standard for level three autonomous driving technology in vehicles. Read more here.

2020

California allows taxi without drivers

In mid-November 2020, it became known that the California Public Utilities Commission (CPUC) approved two new programs under which companies in the state of California will be able to deploy their projects to carry passengers using self-driving cars, charging a fare. Read more here.

Entry into the market of the Honda Legend sedan with autopilot of the 3rd level

In mid-November 2020, it became known that Honda became the world's first automaker to open sales of a car with self-driving equipment corresponding to the third level of automation according to the international classification. Read more here.

The organization, the main organization in Europe in crash tests, assessed autopilots

In early October 2020, the European Committee for Independent Crash Tests of Cars (Euro NCAP) published the results of tests of cars that are equipped with driver assistance systems. We are talking about assessing the operation of the Highway Assist system.

10 cars took part in testing: Tesla Model 3, Audi Q8, Volvo V60, BMW 3-Series, Peugeot 2008, Mercedes-Benz GLE, Renault Clio, Volkswagen Passat, Nissan Juke, Ford Kuga. Euro NCAP evaluated the operation of automotive systems according to two criteria: the balance between computer intervention and driver involvement in the process, as well as ensuring safety in unforeseen situations. According to the test results, the cars were given one of four ratings: Entry (initial), Moderate (middle), Good (good) and Very Good (very good).

The organization, the main organization in Europe in crash tests, assessed autopilots, the Tesla car system turned out to be worse than that of Mercedes-Benz, BMW and Audi

The following participants received the "Very Good" rating: Mercedes-Benz GLE, Audi Q8, BMW 3 Series. The Ford Kuga model received a "Good" rating. The "middle" category includes cars: Tesla Model 3, Nissan Juke, Volkswagen Passat, as well as Volvo V60. In the "initial" category were the Peugeot 2008 and Renault Clio models.

Thatcham research company Matthew Avery noted that the Tesla Model 3 demonstrated good results during tests for automatic emergency braking and collision prevention, in particular, experts praised the technology for updating autopilot using Wi-Fi. However, this model turned out to be worse than Mercedes-Benz, BMW and Audi. Euro NCAP noted that the main problem of this model is that it causes the driver to feel completely detached. Avery explained:

The algorithm of the system is more authoritarian than assisting. This means that she does not complement the driver, but replaces him.

Avery added:

Unfortunately, some drivers believe that they can purchase a self-driving car today. This is a dangerous misconception in which too much control is transferred to vehicles, and they are not yet ready to cope with all situations.[13]

ITMO University proposed using blockchain to monitor the safety of unmanned vehicles

Employees of the Faculty of Security of Information Technologies of ITMO University have proposed a method for using the blockchain system to monitor the traffic situation. The university announced this on February 21, 2020. With the help of the proposed algorithm, it will be possible to create changelessly protected logs of the activity of unmanned vehicles, allowing you to reliably record road accidents. Also, the technology can potentially help compile accident databases to identify the most dangerous sections of the road.

Describing the near future, engineers, economists, politicians around the world say that cities will become "smart," and cars, for the most part, will become unmanned. However, in order for this picture to become a reality, the development of technologies for collecting, analyzing, storing and protecting a huge amount of data is necessary. It should be remembered that many are interested in data distortion: drivers who want to hide the fact of participation in an accident; manufacturers of self-driving cars in need of the best possible safety indicators of their developments; authorities trying to improve the statistics of accidents and traffic jams. Employees of the Faculty of Security of Information Technologies of ITMO University Sergey Bezzateev and Vadim Davydov proposed a model for accumulating and protecting data on road conditions, which uses the blockchain mechanism. Their development should help protect data from manipulation.

The first possible application of mathematical algorithms proposed by employees of ITMO University is the creation of a system for collecting data on "behavior" on the roads of unmanned vehicles. All events that occur with the car: parking, accident, violation of traffic rules, malfunction - can be recorded on an on-board computer and sent to a common database.

China has a problem with improving the safety of self-driving cars. It is necessary to keep a log recording everything that happens, otherwise, according to the rules, security is not considered secured. As of February 2020, one of the proven and reliable ways to make this record reliable is blockchain, whose reliability has already been proven on cryptocurrency. At the same time, not only this car, but also all other drones, in the observation zone of which it will be, will keep records of the events that occurred with a specific car, tells Bezzateev

The technology can help not only for working with drones. The introduction of such a system will help collect big data on urban roads and suburban highways. The information will be accumulated in a single database, which will make it possible to compile a large, constantly updated accident database. With such information, road services and the traffic police will be able to see patterns and identify places with incorrect traffic lights, marking errors, roadway defects leading to constant accidents. At the same time, data collection can be carried out not only by on-board computers and video recorders, but also by ordinary passers-by who can transmit information about an accident through their mobile devices.

Potentially, scientists explain, such a system can also help collect reliable information about the history of a used car before buying it. So, a person who wants to buy a car could request data from such an information base and check whether the car participated in an accident or even what reliable mileage it has.

2019

Self-driving car can be tricked with special pictures

On November 10, 2019, it became eternal that specialists from the Max Planck Institute for Intelligent Systems and the University of Tübingen conducted a safety study of unmanned vehicles. Engineers checked how cars cope with the recognition of human figures.

There is a total system failure, an unmanned vehicle can leave the lane or unexpectedly brake. This is somewhat reminiscent of the effect when strobe flashes of a certain frequency cause epileptic seizures in some people. That is, through vision, the failure of the body's functions is artificially caused, in this sense, the human and drone brains have something in common.

According to scientists, in just four hours they managed to create a sample of color combinations that cause a self-driving car to be akin to panic, and this becomes a security threat. The pattern can be easily applied to T-shirts or made stickers on road signs or shopping bags. Hackers can also take advantage of this, the researchers warn.

The problem lies in the imperfection of artificial intelligence, which manifests itself when recognizing images. The algorithm uses a built-in camera to monitor the environment, such as the road in front of a car, and to detect obstacles. If recognition fails, the robot machine stops at best for safety reasons.

The authors of the study stressed that such a bug occurs with a probability of only a few percent, but this is enough for the drone to behave unpredictably. The experiment showed that if a car camera sees the same spot several times, its reaction will be special each time.

Of course, scientists and programmers will eventually solve this problem, but for now it remains. Researchers believe that now the task of automakers is to train their systems to be resistant to such attacks^[14].

The number of self-driving cars will grow to 10 million

July 2019 - In the future, the number of self-driving cars will rise to 10 million, according to scientists at the Georgia Institute of Technology. Scientists fear that cybercriminals will be able to paralyze urban traffic by hacking only a small part^[15] self-driving cars].

The main consequences of such cyber attacks on unmanned vehicles will be road accidents, as well as huge traffic jams, which will hit ambulances with wounded, sick and dying people.

Researchers simulated a situation in which hacking multiple self-driving cars could affect urban traffic in Manhattan (New York City area).

Stopping just 20% of cars during rush hour completely paralyzes traffic in the city, researchers say. The city will be divided into several sectors, which will allow you to move between neighborhoods, but it will no longer be possible to get to the other end. Hacking and forcibly stopping 10% of cars during rush hour will block the movement of ambulances. The results of the study also showed that such consequences can occur at any other time of the day.

Researchers recommend that self-driving car engineers link cars with multiple digital networks to prevent an attacker from accessing each car by compromising one or two networks.

In October 2017, speaking at the World Knowledge Forum in Seoul, South Korea, Mobileye Chief Executive Officer and Intel Senior Vice President, Professor Amnon Shashua proposed a way for the automotive industry to confirm the safety of self-driving cars. His decision, published in a scientific article and presented in a summary of this work for ordinary people, offers a mathematical formula using which you can confirm that a particular unmanned vehicle works in compliance with the norms of responsibility and cannot serve as the cause of an accident, the blame for which could be placed on this car.

The Responsibility Sensitive Safety model presented by scientists provides specific, measurable parameters that characterize human ideas about responsibility and caution, and defines the so-called "Safe State" (Safe State), supporting which an unmanned vehicle cannot cause an accident, regardless of what maneuvers or actions other vehicles perform.

In his speech, Amnon Shashua called on industry representatives and those who develop strategies to "work together to create standards that would allow for the unequivocal identification of the culprit" in the inevitable accidents involving cars driven by drivers and self-driving cars. He explained that all modern rules and regulations are based on the idea that the driver drives the car, therefore, to regulate unmanned vehicles, new parameters must be introduced into the rules.

to be able to identify the culprit behind the accident. Even the best drivers in the world are caught in road accidents, and self-driving cars will also not be able to avoid this fate due to the actions of other road users. But the likelihood that a responsible and cautious driver will have an accident through his own fault is very small, especially if a panoramic view is available to the driver, and he himself has a lightning reaction - like an unmanned vehicle, "explains Amnon Shashua. The proposed RSS model allows you to formalize the operation of unmanned vehicles, as a result of which "drones" will always work only within the framework of the model that is considered "safe," based on clear definitions of guilt, and which is approved by industry representatives and regulatory authorities.

Cyber ​ ​ threats are one of the calls for any connected devices, including cars. The head of the telecommunications company SoftBankMasayoshi Son at the end of February 2017 cited data that the number of cyber attacks on Internet-connected facilities quadrupled in 2016 compared to 2015.

In the case of cars, this is a special reason for concern, since people can suffer as a result of the actions of intruders. In theory, a hacker can hack the network, stop data transmission, turn off the brakes, or simply stop the machine.

In mid-2015, for example, computer security specialists at the Uber Center for Advanced Technology discovered a vulnerability in the Jeep car software, thanks to which they were able to remotely access some of the car's systems: air conditioner, wipers, audio system and brakes.

Cyber incidents are a problem for any automaker in the world. This is a matter of public safety, - noted earlier Mary Barra, CEO of General Motors.

Argus, a company specializing in cyber defense tools for cars, believes that some single product cannot be suitable for these purposes: various solutions designed for different parts of a connected car must integrate with each other so that full protection is provided.

Automakers and manufacturers of automotive solutions are investing in the development of this area of ​ ​ cybersecurity. A number of automakers, including Tesla, Fiat Chrysler and General Motors, have created special incentive programs for individuals who report security holes in their machine systems.

Responding to market requests, more and more companies are developing specialized solutions for cars. This direction appeared, for example, at Kaspersky Lab. In 2016, the company announced that it was developing a secure, secure operating system that, in particular, could be used for cars.

In 2016, carmaker Harman acquired Israeli cybersecurity technology TowerSec for its tamper-proof software. In the same year, another Israeli startup, Karamba Security, received investments to develop automotive cyber defense technologies.

Whose life is more important: a driver or a pedestrian?

In addition to technological challenges to the transition to mass use of drones, "moral" issues related to decision-making by autopilot will also have to be resolved. For example, should it be designed to protect the life of the driver at any cost, even if in an emergency it is necessary to ram a crowd of pedestrians?

Rules for self-driving cars in Germany

The German Federal Ministry of Transport and Digital Infrastructure has announced its intention to implement road rules for self-driving cars, their manufacturers and owners. As The Register writes in the summer of 2017, the document will oblige developers of self-driving cars to program their autopilot in such a way that in any unexpected situation on the road it puts human life above the life of animals or the safety of private or public property^[16].

Currently, no country in the world has uniform traffic rules that regulate the technical requirements for unmanned vehicles and regulate their movement on public roads. Some countries allow self-driving cars to be moved along common roads, but in this case a special permit is required. At the same time, the driver must always be driving an unmanned vehicle, ready to intercept control from the autopilot in case of any emergency.

The rules for self-driving cars in Germany were drafted by the advisory board of Germany's Federal Ministry of Transport and Digital Infrastructure, which includes 14 scientists and lawyers. In total, 20 requirements for unmanned vehicles, their manufacturers and drivers were included in the list. Thus, the requirement for the value of human life implies that autopilot in any emergency must drive a car in this way in order to save people's lives.

In the event of a dual emergency, the autopilot should not make a choice whose life should be saved - a driver or pedestrian, an elderly person or a child. Autopilot will have to do everything possible to save the lives of all participants in the accident. All self-driving cars registered in Germany must have a "black box," records from which can be used after an accident to find out who is responsible for it - on the driver or on autopilot.

At the same time, in all cases of an accident involving an unmanned vehicle, the "presumption of guilt" will operate, that is, the driver will always be considered guilty of the accident until the data of the "black sensor" or other results of the investigation of the incident prove the opposite. The rules also included the exclusive right of drivers to choose information that its manufacturers can receive from an unmanned vehicle. We are talking about location, speed, driver's data and a lot of other information that can be used, for example, to target ads.

Moral dilemma

Jean-Francois Bonnefon, a psychologist at the School of Economics in Toulouse, and colleagues tell^[17]people generally support the idea that in a critical situation, a car should crash into a wall or somehow sacrifice a driver to save more pedestrians. At the same time, the same people want to drive in cars that protect the driver at any cost, even if it entails the death of pedestrians.

Such a conflict puts manufacturers of computerized cars in a difficult position, Bonefon notes. Between the car, which is programmed for good for most and which is programmed for passenger self-protection, buyers will overwhelmingly opt for the second.

There are many emergency scenarios where the car will have to make a choice of who to donate

The authors of a study on the social dilemma of autonomous cars, published in the journal Science in 2016, believe that there are other complex moral questions in this area. Autonomous cars will have to make decisions in emergency situations, the consequences of which cannot be predicted in advance. Is it acceptable, for example, to program a car to avoid colliding with a motorcyclist by crashing into a wall? After all, the passenger of the car in this case has a better chance of survival than the motorcyclist who will collide with the car.

Autonomous cars can revolutionize the transportation industry, but they pose a social and moral dilemma that can slow down the spread of this technology, said Iyad Rahwan, a scientist at the University of California, one of the authors of this study.

Psychologist Kurt Gray of the University of North Carolina at Chapel Hill is confident that working compromises can be reached. If self-driving cars are programmed to protect a passenger in emergency situations, the number of road incidents will decrease anyway. Except in rare cases where such cars may pose a danger to passengers, they will in any case not exceed the speed, will not drink alcohol or type text messages on the go, which is why society will ultimately benefit.

The moral of artificial intelligence is one of the most talked about issues related to the onset of the robot era. In 2016, the Massachusetts Institute of Technology (MIT) in the United States developed a special test that helps to better understand what moral dilemmas developers of artificial intelligence face, and at the same time to deal with their moral guidelines^[18].

The test is very simple. In it, you need to put yourself in the place of the artificial intelligence of a self-driving car and choose who can be sacrificed in an accident - pedestrians at an intersection or passengers in a car. Sometimes you have to choose between who of the pedestrians needs to be crushed and who needs to be saved.

[4] There are 13 questions in the test. The number of passengers and pedestrians in some tasks is different, in some - the same. In addition, their age, gender, social position differ. There are pets in some matters - they are equated to other passengers and pedestrians.

For example, take the following task: a woman and two children (boy and girl) are sitting in the car, and a woman and two old women are walking along the pedestrian crossing. It is necessary to choose which of them to save artificial intelligence and which to sacrifice.

At the end of the test, the user is told who he donates most often and how other people who passed the test answered.

On the MIT website, you can also come up with your own moral task based on test questions and see what questions other users have posed.

Legislation

In addition to technological challenges, many issues at the level of legislative regulation will have to be resolved in order to move to the mass use of autonomous cars. Regulatory documents are needed that define the main technological and legal concepts in this area, regulation of the possibilities of using such technologies in general, responsibility in the event of incidents with unmanned vehicles, etc.

In one form or another, regulatory documents in this area have already been presented or are being developed in some countries. The United States has especially moved forward here. Nevada back in 2011 became the first state in the country to begin regulating the use of autonomous vehicles on the roads and issues related to their insurance, safety and testing.

The conditions for the movement of unmanned vehicles of varying degrees of freedom are now legislated in different states of the United States. In 2015, the governor of Arizona, United States, signed an order requiring self-driving cars in the state to register under the same conditions as conventional cars. There were no additional requirements for autonomous machines. In addition, state laws do not prohibit testing self-driving cars on the roads.

In late 2016, Michigan's governor signed a package of laws that directly addresses the scope of self-driving cars and effectively legalize their private and commercial use. They allow the sale of mass-produced self-driving cars that have passed certification, while cars are allowed to go on public roads without a driver behind the wheel and move in convoys. In addition, an unmanned taxi is now allowed in the state.

In Britain, in 2016, they began preparing amendments to the legislation, which should, firstly, allow insurance of the liability of unmanned vehicles, and secondly, update the Road Code (Code of Traffic Rules) Great Britain , taking into account the development of autonomous vehicles.

2021: $100 million fine threatens US companies for keeping quiet about drone accident

The National Highway Traffic Safety Administration of the US Department of Transportation (National Highway Traffic Safety Administration, NHTSA) issued a regulation on June 29, 2021, according to which all automakers are required to report all accidents involving autopilots from the second to fifth level of autonomy on the SAE scale. This became known on July 6, 2021.

According to the document, car manufacturers. The order lists all cases when the company is obliged to inform the management of the incident:

• If the systems were automated at least 30 seconds before the collision, or an accident led to the hospitalization of at least one person,
• fatal outcome of road accident participants,
• the need to tow the vehicle,
• airbag actuation
• injury to pedestrians or cyclists.

The report must be sent to NHTSA within one day of becoming aware of the accident, followed by an update within 10 days. Update reports monthly by adding any new information.

Failure to comply with this order entails a fine of $22,992 for each day of delay with the submission of the report. The maximum amount of the fine is $100 million. By the way, there are similar norms, for example, in the laws of the province of Ontario, Canada. There, companies are required to report on accidents involving 3-5 levels within 10 days from the date of the accident^[19].

Artificial intelligence in transport

Main article: Artificial intelligence in transport

Manufacturer News and Models

In Russia

• Yandex.Taxi Unmanned vehicle
• KamAZ unmanned vehicle
• KamAZ-1221 Shuttle Unmanned bus
• C-Pilot and Cognitive Agro Pilot Automatic Driving System
• Auriga Video Markup and Data Streaming Tool for Machine Learning
• MatrЁshka
• Traft Truck Project Self-driving truck
• Argo AI

In the world

• Tesla Motors, Tesla Semi Truck - Tesla Model
• Uber Advanced Technologies Group
• Waymo
• Otto
• Honda Autopilot
• Volkswagen Sedric
• Volvo Group
• Toyota Safety Sense, Toyota e-Palette
• Bosch Onboard computer with artificial intelligence
• Nvidia AI Traffic Jam Pilot, Nvidia Drive AI platform for self-driving cars
• Harman International = Samsung Electronics
• Ford Autonomous Vehicles
• Ford Transit Connect Language of Visual Communication for Self-Driving Cars
• Ford Fusion Hybrid Self-driving car with hybrid drive
• Ford Wrong Way Alert
• Mitsubishi Electric Autonomous Driving Technology
• CARLA Simulator Self-driving car training
• Renault Autopilot
• Alphabus
• Honda + SenseTime
• LG LTE V2X (communications for self-driving cars)
• Drvline (platform for unmanned vehicles)
• Apollo (self-driving car platform), Apolong (self-driving bus)
• Cisco Software Defined Vehicle, SDV + Hyundai Motor Company + Hyundai Self-Driving Vehicles
• ZF ProAI
• General Motors Robotaxi
• Renesas R-Car H Protection of Connected Cars
• Continental Autonomous Driving Systems
• Navya Arma Unmanned Bus
• Neolix Self-driving car

Self-driving trucks

Main article: Self-driving trucks

Self-driving cars in Russia

Main article: Self-driving cars in Russia

Since 2015, companies in Russia have been actively developing the technologies necessary to create unmanned vehicles.

Global market

Main article: Self-driving cars (global market)

The creation of a full-fledged self-driving car is one of the most exciting challenges for technological thought at the beginning of the 21st century for companies around the world.

Unmanned trains

Main article: Unmanned trains

Smart City - Smart Transport

• Transport Technology Portal

• Vehicle-to-Everytning (V2X) 5G in car evolution
• Electric vehicles

• AIC Safe City - as part of the implementation of the federal program "Safe City"
• Smart cities
• Intelligent Transport Infrastructure (ITS) Russia
• Unified Federal Center for Automatic Recording of Traffic Violations (CAFAP)
• Intelligent Transport System (Moscow)
• State IT giants: the 10 most expensive information systems in modern Russia
• Avtodata Platform
• AutoNet Roadmap
• Regional Navigation and Information System (RNIS)
• Center for Traffic Management of the Government of Moscow (GKU TsODD)
• Navigation and Information Center for the Administration of Freight Road Transport in Moscow
• Transport in Moscow

• Transportation Management System Transport Management Systems and _Projects
• Fleet management systems Auto management systems Systems and _projects
• Vehicle Safety and Control SystemsSystems and _Designs
• Satellite Communications and NavigationSystems and _Projects

Notes