KamAZ launched mass production of unmanned trucks

See also: Autopilot (self-driving car)

2024: KamAZ launched mass production of unmanned trucks

February 22, 2024 it became known about the production of unmanned trucks KAMAZ"." By September 2024, the company is going to make 12 such machines.

The production of unmanned vehicles for the transportation of goods at the KamAZ plant was reported by TASS. According to him, by February 2024, nine cars are participating in unmanned cargo transportation on the Moscow-Petersburg highway: six from KamAZ and three from Sberbank. By the end of 2024, their total number should increase to 43 pieces.

It became known about the production of unmanned trucks "KamAZ"

KamAZ is developing self-driving trucks based on the KamAZ-54901 tractor. As noted in the company, the machines operate using hubs, with the help of which there is a link from traction to a tractor moving in unmanned mode.

On February 5, 2024, the KamAZ press service announced that the automaker's leasing subsidiary had financed the purchase of five KamAZ-54901 unmanned vehicles. The leasing agreement was concluded for 18 months with zero advance and a subsidy from the Ministry of Industry and Trade.

In February 2024, KamAZ also announced the conclusion of an agreement with Kuzbass State Technical University named after T.F. Gorbachev. The agreement involves the joint development of an unmanned pit dump truck with a carrying capacity of 220 tons. Partners agreed to perform the following tasks:

• theoretical study of the dynamics of the movement of a mining dump truck by simulation modeling
• development of a software and hardware complex for the design of an unmanned pit dump truck,
• creation of a draft design of a dump truck and a technical design of a self-driving system, a technical design and working design documentation for a digital dispatching system of a coal quarry.

It is assumed that it will be possible to control the operation of an unmanned dump truck remotely.^[1]

2023: Unmanned Kamazy began work in the Arctic

Gazprom Neft began transporting goods in the Arctic by unmanned vehicles, which the company announced in May 2023. Self-driving trucks went on flights at the East Messoyakhskoye field, located on the Gydan Peninsula of the Yamalo-Nenets Autonomous Okrug. Read more here.

2022

Magnitogorsk Metallurgical Plant began to use unmanned KamAZ

December 27, 2022 Magnitogorsk Iron and Steel Works () MMK announced the use of pilotless the truck "" KAMAZ to transport refractories from production preparation management warehouses to the construction site of coke battery No. 12. More here.

KamAZ built a testing ground for unmanned vehicles for 1 billion rubles

In early October 2022, it became known about the launch of KamAZ a new landfill for testing unmanned vehicles. About 1 billion rubles were invested in this project.

At this training ground, it will be possible to test the entire range of KamAZ equipment, including promising drones. There are conditions here for testing handling, stability, brakes, noise insulation and more. In addition, with the help of this database, they will test the reliability of components and assemblies, configure electronic safety systems, bring driver assistance systems (ADAS) and control systems for robotic cars.

The test complex includes ten structures. The main test track is 770 meters long and 16 meters wide. At the landfill there is a cobblestone road for testing vibration loading and a road with different irregularities - both four meters wide. Controllability will be checked on a dynamic platform with diameters of 120 meters with artificial irrigation.

In early October 2022, Vedomosti, with reference to Deputy Minister of Transport Dmitry Bakanov, reported that Sberbank and KamAZ were ready to start testing three unmanned trucks on public roads, but their applications for the creation of experimental legal regimes (EPR) were not approved by the government.

According to a source in one of the companies engaged in the development of drones, two cars produced by Sberbank and one by Kamaz are ready for testing. "Sberavtotech" expects the adoption of the relevant document by the government within a month, a company representative told Vedomosti. The fleet of unmanned passenger cars "Sberbank" now has 180 cars, he added.^[2]

Announcement of an unmanned mining truck

At the end of June 2022, KamAZ presented an unmanned dump truck, which received the KAMAZ-6559 index and the name Jupiter 30. A vehicle deprived of a cab for a driver was created to transport loose rock or ore using deserted technology - without the presence of people in the dangerous area of ​ ​ operation of heavy trucks and excavators. Read more here.

2021: The cost of the KamAZ unmanned truck is 14.5 million rubles

In December 2021, KAMAZ"" named the price of its unmanned truck, which will be tested on the highway MoscowSt. Petersburg"-" M11 (). We are talking about a cost of 14.5 million rubles (about twice as expensive as a standard analogue), it is indicated in the presentation of an unmanned logistics corridor (BLK) at the M11 that it prepared. Ministry of Transport

According to Vedomosti, the Kamaz-54901 experimental car with an autopilot system will be delivered with a diesel engine, a semi-trailer and a 3 + autonomy class. This means the presence of an adaptive cruise control system: it requires the presence of a driver and will not allow him to remove his hands from the steering wheel, a KamAZ representative explained to the publication. The company is ready to deliver the first four cars in 2022, and in 2024 the automaker will supply 30 more such tractors.

An ordinary tractor "KamAZ-54901," but with a three-year service contract, will cost 7.98 million rubles from January 1, 2021, a company representative said. 68 such machines could be collected for the amount spent on the creation of a test site for KamAZ drones - its construction in Naberezhnye Chelny will cost more than 1 billion rubles.

Another unmanned model - "KamAZ-54907" - should appear in 2025. It will also be offered with a semi-trailer, but already with a liquefied gas engine and with an autonomy class of 4 (the system will require prisut­stviya a person in the cab, but the steering wheel can no longer be held). Its price will be from 19 million rubles.

The need of carriers for drones for 2024 by the end of 2021 is already estimated at 189 units. It is assumed that Evocargo Ozon they are also needed for their own logistics. Another 25 cars will be required, "" Russian Post Office seven -, X5 Retail Group 12 - "." Business lines^[3]

2016

Cognitive Technologies and Kamaz will release an industrial version of the car with AI elements by the end of 2017

Cognitive Technologies and Kamaz are planning to release an industrial version of Kamaz with elements of artificial intelligence by the end of 2017.

"We plan to

release the completed version of the first level ADAS with artificial intelligence elements for 'Kamaz' by the beginning of next year. This will already be a real pre-industrial prototype, which can be brought to the stage of industrial production by the end of 2017, "said Olga Uskova, President of Cognitive Technologies. "Last week, we passed one of the last stages of work under an agreement with the Russian Ministry of Education and Science and reached the finish line - the release of a pre-industrial version of the driver assistance system or ADAS (Advanced Driver Assistance System) for Kamaz with elements of artificial intelligence," she added.

It is assumed that the developed ADAS will correspond to the first level of the generally accepted international classification of SAE International, which provides for the operation of warning systems for the danger of collisions with cars, pedestrians, crossing the marking line, as well as the identification of road signs. The second level of ADAS will already be active assistance to the driver, providing intervention in the control system (slowing down, keeping in the lane, etc.). According to the SAE International classification, there are 6 classes of autonomy: from level 0 - fully manual control to 5 - a completely unmanned car.

A feature of the Cognitive Technologies approach is the use as a technological basis of the so-called passive (signal absorbing) model of computer vision, when the information necessary to form a control effect is read from video cameras. Actually, according to the same principles, a person acts. This approach is used in Russia in addition to the active (radiating) part of the model on which many foreign projects are based, for example, Google Car, the company said.

According to representatives of Cognitive Technologies, the most important feature of the project is the possibility of real work in Russian conditions. Unlike foreign developers, who focus largely on ideal traffic conditions (high-quality markings, favorable weather conditions, the absence of damage to the road surface, etc.), the approach of the Russian company to creating a machine vision system allows you to recognize the road scene (including road borders, lane width, etc.) in domestic realities. For example, Cognitive Technologies scientists created a model that allows the drone to operate in off-road conditions, when significant vertical fluctuations and movements of the car are taken into account. It allows even in such conditions to stably recognize all objects of the road scene.

Project dynamics

In September 2016, participants in the project to create an unmanned KamAZ announced the continued development of the project.

In February 2015, Cognitive Technologies and Kamaz announced the start of joint work to create an unmanned truck. Cognitive Technologies took up the creation of the "brain" for the car. In the fall of 2015, the developers showed the first prototype of the KAMAZ unmanned truck^[4].

According to the statement of PJSC Kamaz, within the framework of the project, work is carried out in three main areas:

• Create a smart assistant for the driver, capable of performing adaptive cruise control and urgently slow down in critical situations
• Create an autopilot system under certain conditions (for example, in a traffic jam)
• Create an autopilot system

KamAZ - drone, (2015)

In the Unmanned Kamaz project, Cognitive Technologies used all the experience and achievements. The company's specialists taught the piloting system to create a zone of interest - a narrow area of ​ ​ foveal vision that surrounds the observed object.

The developed technology makes it possible to process only the necessary information, and with higher quality indicators. Imagine the advantage this gives us in saving computing resources and increasing the speed of information processing, when instead of all the images taken from the video camera, it is enough for us to process 5-7 percent of it. Yuri Minkin, Head of Unmanned Vehicle Development at Cognitive Technologies

According to company representatives, the quality of technical vision in the end turned out to be better than the physical vision of a person.

The developers reported attention to the problem of safe use of unmanned vehicles.

Work in the direction of ensuring security is in full swing. Cognitive Technologies has a separate group working on this problem and has extensive development experience. To ensure safety in terms of reducing accidents, the company engages external experts, creates a base of precedents, which allows us to learn from the mistakes of others, as in the case of a Tesla accident. Many situations are worked out by specialists at the virtual training ground. Yuri Minkin

The developers did not forget about protecting the car from external interference.

Financing

As of September 2016, the state represented by the Ministry of Education and Science allocated about 300 million rubles to the Unmanned Kamaz project. Within the framework of the federal target program "Research and development in priority areas of development of the scientific and technological complex of Russia for 2014-2020." About 90 million rubles. investments belong to PJSC Kamaz. Taking into account the total cost of the project, additional funding is being provided (the cost of the project can reach 20 billion rubles, this is the development of an unmanned truck, the creation of a landfill, practical tests and the transition to the stage of industrial use).

Tests

According to a statement from the companies for September 2016, the project is actively working to create a landfill for practical tests of an unmanned truck. The artificial intelligence system will be tested at the "virtual training ground," which will open on the basis of the robotics center of the Moscow Institute of Steel and Alloys. According to the developers, the "virtual polygon" will help save about 2 billion rubles.

An idea of ​ ​ training an artificial intelligence system at a virtual training ground, (2013)

Experts from the Cambridge and Massachusetts Institutes of Technology are taking part in the creation of the "virtual landfill." Their estimates of the chances of Russian companies for global success are very restrained.

To start testing an unmanned Kamaz in real conditions, it is necessary to resolve issues in the legislative field - to determine what responsibility the manufacturer of the vehicle will bear for driving an unmanned vehicle. Appropriate amendments must be made to the Criminal and Civil Codes, traffic rules, Administrative Code and some technical regulations. The Interior Ministry suggests that many criminals will try to use the autopilot system as a tool for committing crimes.

Representatives of the Unmanned Kamaz project also work with regulators in terms of legislation.

{{quote 'author
= Yuri Minkin' We are actively working to create a legislative framework. This year, we initiated hearings in the State Duma on the problem of legislative support for the movement of unmanned vehicles in Russia. We have proposed options for determining what is an unmanned vehicle, what is a subject of responsibility, what is an operator, and what is a driver. Given this, the movement of unmanned vehicles on public roads will be able to begin only in a few years. According to our estimates, such tests may not begin until 2019 - 2020.}}

The unmanned truck is being developed only for civilian purposes, however, according to experts, in the future the range of use of this technique will be expanded.

The start of mass production of unmanned vehicles is scheduled for 2025.

Yandex will introduce artificial intelligence into KamAZ cars

Specialists of the company Yandex"" and PJSC KAMAZ"" entered into a cooperation agreement in the summer of 2016, which involves equipping the cars of the Kama Automobile Plant with systems, artificial intelligence the press service of Yandex reports.

Joint work will be carried out in several areas. In particular, we are talking about adapting services for trucks and introducing voice control of the car's multimedia system. In addition, Yandex's artificial intelligence technologies such as computer vision, machine learning and speech technologies will become the basis for the possibility of semi-autonomous car control.

More: Project:KamAZ Shuttle Unmanned bus

In order not to crash an unmanned vehicle, Russians allow a pedestrian and a dog to be hit

The developers of an unmanned vehicle based on KamAZ asked people to morally assess the situations that artificial intelligence will face. Respondents were asked to choose a behavior model in situations when a person or^[5] dog suddenly appears on the road^[6].

Every fourth participant in the survey (25.7%) agreed with the thesis that it would be more correct to crush a person who entered the road in front of a moving car with autonomous control than to move into a ditch and allow the death of a passenger inside this car.

If a dog suddenly appeared on the road, twice as many respondents are ready to crush it - 55.2%. Another 40.1% are ready to save the animal's life by applying emergency braking, even if this leads to a collision with a vehicle following the distance. It is noteworthy that 13% more women think about the fate of the dog than men.

The majority of respondents (71%) in one way or another advocated directing an unmanned vehicle into a ditch in order to avoid injury to people who accidentally entered the roadway in front of the car, even though falling into a ditch can cause irreparable damage to an unmanned vehicle and lead to injuries critical to the life and health of the passenger.

A fairly large number of respondents (43%) spoke in favor of the idea of ​ ​ developing a system for determining the value of the lives of accident participants. "It was proposed to use the results of an instant search for information on the Internet and social networks to compile a detailed portrait of an accident participant, so that based on this data, artificial intelligence can make a decision on who to sacrifice and who to save. For example, if in front of an autorobot in the cabin of which there is a person with a tarnished reputation, a production leader runs out onto the road, then artificial intelligence will have to be guaranteed to send the car into a ditch, "the researchers explained.

In total, 1,384 respondents from Russia took part in the survey, including 73.8% of men and 26.2% of women. The age of the main group of respondents is up to 30 years (60.6%). The share of respondents aged 31 to 50 years - 33.5%, from 50 years - 5.9% (of which 3.7% are pensioners). The survey was conducted via the Internet.

Designation of roads for unmanned vehicles

On June 29, 2016, Cognitive Technologies announced the development of draft warning road signs with Artemy Lebedev's design studio, designed to inform road users about the appearance of unmanned vehicles: "Attention, unmanned vehicles on the site," "The beginning of the road section with the participation of unmanned vehicles," "The end of the road section with the participation of unmanned vehicles."

The package of documents with design layouts of the corresponding road signs was sent to the State Specialized Installation and Operation Enterprise of the Ministry of Internal Affairs of the Russian Federation (FSUE GOSSMEP). MINISTRY OF INTERNAL AFFAIRS Russia

Road signs are planned to be conceptually approved and, after agreement with the State Traffic Inspectorate, officially and in advance put into operation by 2020 - when, according to the developers of unmanned transport technologies from Cognitive Technologies, drones will first appear on public roads as full-fledged road users.

The primary use of signs is expected by 2018, when the first unmanned vehicles will appear on the roads (including on specially designated lanes) of large cities. The Moscow Moscow region Republic of Tatarstan has also been chosen as pilot regions for testing unmanned vehicles.

The signs were developed by order of Cognitive Technologies, taking into account the visual continuity and legal compliance with all modern requirements: first of all, we are talking about the observance of GOST R 52290-2004 standards in signs and other standards on road signs adopted in Russia.

Road signs for drones must have a simple symbolic series and capaciously convey the essence of the message, without causing motorists to false associations and confusion with other generally accepted signs. Lebedev's studio specialists managed to accurately cope with the task. Andrey Chernogorov, CEO of Cognitive Technologies

Signs are planned to be included in section No. 5 "Signs of special prescriptions" of traffic rules. According to the current version of the SDA, signs of special prescriptions introduce or cancel certain traffic modes. The sign associated with traffic on a section of the road, along with piloted cars, also unmanned, should oblige drivers to abandon sharp changes while driving along a section of the road in the area of ​ ​ the sign. In addition, drivers should especially monitor the observance of the distance in order not to artificially create stressful situations for the BPTS system, which can at first perceive the impulsive actions of drivers as an accident.

2015

300 million rubles for an unmanned vehicle

In February 2015, Cognitive Technologies and Kamaz announced the start of a joint project, as a result of which it is planned to create a new generation unmanned vehicle on the basis of the Kamaz car by 2020.

For the implementation of the project, the state represented by the Ministry of Education and Science allocated 300 million rubles. as part of the implementation of the federal target program "Research and development in priority areas for the development of the scientific and technological complex of Russia for 2014-2020." Cognitive Technologies won the corresponding competition in 2014. According to its terms, another 90 million rubles. Kamaz will invest in the project.

Cognitive Technologies told TAdviser that the project involves the creation of a joint venture with Kamaz in the future. Later, the issue of ownership of development rights under the project will also be discussed.

According to preliminary estimates, the cost of an unmanned vehicle will not be higher than the cost of an ordinary Kamaz car by more than 10% -15%, TAdviser cited data in the company.

Representatives of Cognitive Technologies also told the portal that the project uses not only previous developments, but also the entire school of image recognition and processing, conscious on the basis of their company and its scientific team over the past 40 years.

To create a drone car, it is planned to use Cognitive Technologies technologies in the field of computer vision. In particular, a passive (signal absorbing) model of computer vision, when the information necessary to form a control effect is read from video cameras. The company notes that this approach is used in Russia in addition to the active (radiating) part of the model on which many foreign projects are based, for example Google Car.

The most important feature of the project in Cognitive Technologies is the possibility of real work in Russian conditions.

"Unlike foreign developers, who focus in many ways on ideal traffic conditions - high-quality markings, favorable weather conditions and more - Cognitive Technologies' approach to creating a machine vision system allows you to recognize the road scene, including road boundaries, lane width even in the absence of any markings," the company notes.

The complex under development will have to ensure the automatic operation of vehicles in various weather conditions, except for cases of insufficient visibility determined by the requirements of traffic rules. The system is obliged to detect pedestrians regardless of the direction of their movement. The time allotted for hazard detection and response will not exceed 0.3 seconds.

It also provides for the possibility of detecting almost any obstacles on the road, including animals, including dogs, which qualitatively distinguishes the Russian solution from the foreign analogues that exist today.

In addition, the system will be able to detect and recognize road signs, traffic participants regardless of angle, partial blocking and direction of movement, as well as traffic light signals.

Oleg Slavin, project manager for the creation of Cognitive Technologies machine vision systems, notes that many of the world's leading automakers already have prototypes of unmanned vehicles, but they are unlikely to be able to count on their use in Russian realities.

"The challenge of our developments is to create a better system. If it works effectively on Russian roads, definitely, we will get a competitive advantage in the world market

, "he says.

Cognitive Technologies seeks talent to develop self-driving vehicle

In February 2015, Cognitive Technologies announced the Cognitive Vehicle Detection Benchmark Competition for the best development software () ON of detecting vehicles moving on roads, in images obtained from a moving video camera (recorder video).

Participants who have offered the most interesting solutions to the task will be offered work in the department of development and software projects of Cognitive Technologies with the ability to carry out scientific activities in the field of machine vision, as well as cash prizes: for I place - 100 thousand rubles, for II place 80 thousand rubles and for III place 50 thousand rubles.

Applications will be evaluated according to the following criteria: the quality of detection, the performance of the program, the resource intensity of the program, the level of scientific and practical novelty.

The competition committee was formed from employees of Cognitive Technologies and the Institute for System Analysis of the Russian Academy of Sciences, specialists in the field of machine vision.

According to Oleg Slavin, project manager for creating Cognitive Technologies machine vision systems, "The organizers of the competition do not expect commercial quality programs from its participants and do not plan to use the programs presented for the competition for commercial purposes. Its goal is to recruit promising developers into the creative group, which faces interesting practical problems, the result of which should be the creation of an unmanned vehicle by 2020. Participation in such projects is a real opportunity to feel as part of a team that creates breakthrough technologies. "

The competition will be held in two stages. At the first stage, participants will have to present a solution to the task set in its conditions. Those of them who, according to the Tender Committee, present the most successful ones, in terms of meeting the criteria for evaluating the decision, will be invited to a conference in which they will have to prepare presentations of their results.

The results of the competition are expected to be summed up at the end of April 2015.

The first tests of the KAMAZ unmanned truck

The installation of equipment at the first unmanned KAMAZ was completed in early May at the KAMAZ Scientific and Technical Center (NTC). The serial KAMAZ-5350 installed the components necessary to work out different modes of autonomous movement of the car: radars, lidar (an active optical sensor that emits laser beams towards the target while driving) transport , video cameras, communication systems and on-board computers. A mobile remote control station was assembled - a joystick with which you can drive a car. Earlier in Naberezhnye Chelny, the truck passed pre-road tests that confirmed the quality of assembly and the operation of all systems, the KAMAZ press service said.

According to the general director of KAMAZ Sergey Kogogin, in the future the tests will continue at one of the landfills of the Ministry of Emergency Situations of the Russian Federation. "Serial production of the machine may begin no earlier than in two years," the head of the auto giant emphasized.

According to Sergei Nazarenko, chief designer for innovative products of the KAMAZ Scientific and Technical Center, two cars are occupied in the first field tests: a drone based on a serial truck and a regular KAMAZ, which is required to simulate movement in a convoy and other modes.

Specialists from VIST Group and Cognitive Technologies have begun setting up and training the car in different driving modes. The test plans include about ten movement options, including a remote control mode, a movement mode along a given route, movement in a convoy and others. It is also planned to test the autonomous movement of the car according to information received only from video cameras and signals/, GPSGLONASS without the help of radars and lidars. Tests should show what the machine "sees," whether the distance is correctly determined. The object recognition system by video signal is constantly debugged to bring the probability of recognition to 100%. During the tests, of course, the driver in the car will be, for safety net.

Work on the Avtorobot project is carried out in three directions: SmartPilot, AirPilot and RoboPilot. The first is the creation of smart assistants for the car that help the driver: they can brake the car in case of danger, carry out adaptive cruise control. The second direction, AirPilot - the creation of machines with remote control. A third system, RoboPilot, allows the car to operate without a driver at all or in Autopilot mode, when the driver has the opportunity to distract himself from driving a truck for a while. It is expected that such drones may appear on public roads in 2025-27.

"An unmanned vehicle is already able to independently control the gas and brake pedals, steering wheel drive and automatic transmission. Developments are proceeding as usual, "RIA Novosti quoted Olga Uskova, president of Cognitive Technologies, as saying

.

Reportedly, the tests were carried out on a limited, closed area of ​ ​ 200×50 meters, while the speed did not exceed 10 km/h. In the near future, tests will continue at one of the landfills of the Ministry of Emergency Situations in the Moscow region, the company notes. It is planned to work out the detection modes of the border of the road, traffic participants, road signs, detour, or stop in front of an obstacle, traffic in a convoy and others.

"The development of the detection mode of the road and other elements of the road scene is one of the most important elements of the tests, since it is associated with the assessment of the work of artificial intelligence in a" pure form, "without the help of any additional devices and devices.

NUST "MISIS" joined the project for the creation of an unmanned truck

On June 25, 2015, it became known about the development of NUST MISIS software for KAMAZ drones.

NUST MISiS acts in the project as a developer of a software complex for active safety of trucks based on technologies of stereoptic computer vision and analysis of the road environment. In addition, the university conducted testing and adaptation of location equipment for project tasks.

At the tests of a prototype unmanned vehicle conducted at the Ministry of Emergency Situations training ground in Noginsk near Moscow, developed by engineers of NUST MISIS, the software responsible for detecting damage to the roadway and predicting the position of vehicles on the road showed excellent results and was highly appreciated by representatives of KAMAZ and Cognitive Technologies.

Director of the Center for Nanotechnology, Professor of Materials Science and Engineering Sciences at the University of Toronto, Harry Ruda, praised the proposed solution to problems related to "deviations from standard road situations, for example, in the case of difficult weather conditions (fog or heavy rain), when the basic computer vision system (use of road markings) can fail." In addition, the model of machine vision used by the project participants creates the conditions for "improving the efficiency of the decision-making process in real time."

As of June 25, 2015, none of the developers of unmanned vehicles, except for the consortium KAMAZ-Cognitive Technologies - NUST MISIS, managed to solve the problem of the drone operating in rain or snow.

Olga Uskova, President of Cognitive Technologies, noted: "The positive assessment of the scientific and technical approach of the consortium participants, received from world-renowned scientists, confirms the high prospects for the announced developments and gives reasonable expectations of the wide possibilities for applying its results not only in Russia, but also in the world market."

Cognitive Technologies is responsible for the creation of an artificial intelligence system within the framework of the project, which will allow an unmanned vehicle to analyze the road situation and carry out control in autonomous mode.

Alevtina Chernikova, rector of NUST MISIS, stressed: "The University attracts young and talented engineers, students and graduates of NUST MISIS from the Student Design Bureau and startup Karfidov Lab to work together with KAMAZ and Cognitive Technologies. Our cooperation is not limited only to current developments: at the suggestion of partners, the university will create a targeted master's program "International Center for Robotics" to attract the most gifted and successful students who will be able to take part in this project in the future. "

The project is expected to be implemented by 2020. According to preliminary plans, in 2016 such trucks will be commissioned at KAMAZ sites.

Technology simulating elements of human vision

Cognitive Technologies, the main developer of the artificial intelligence system in the Russian project to create an unmanned vehicle based on KAMAZ, has developed a technology that allows an autorobot to see and understand participants in the road scene according to the same principles as a person does. The company believes that this allows you to more effectively recognize the road scene and, as a result, can serve as a serious competitive advantage in rivalry with well-known foreign autorobot developers.

It is known that the ability to see distant objects to a person is provided by a narrow area of ​ ​ foveal vision (or central) located in the center of the visual axis of the eye (see Figure 1). It is she who allows you to see them in high resolution. At the same time, the rest of the scene elements located in the zones of peripheral and close to peripheral vision are observed in low resolution.

"In a sense, we managed to build a computer model of foveal vision," Vasily Postnikov, general designer of Cognitive Technologies, comments on the situation. "At any given time, we learned to build a so-called zone of interest - a narrow area of foveal vision surrounding the observed object (see Figure 2)."

This approach allows you to see at any time in high resolution only those elements of the road scene (roadway, shoulder, vehicles, signs, etc.) that are needed to analyze the current situation. "The developed technology makes it possible to process only the necessary information, and with higher quality indicators. Imagine the advantage this gives us in saving computing resources and increasing the speed of information processing, when instead of the entire image taken from the video camera, it is enough for us to process 5-7 percent of it, "says Postnikov. The technology successfully works on all elements of the road scene that are both close and far from the observer. "Essentially, we provide a quality of vision that is better than a person's physical vision," he concludes.

The proposed method is based on the principle of internal self-similarity of the road scene. The developers have learned to identify the most common, fundamental signs inherent in the roadway, be it a highway, country or dirt road. This allows you to recognize the roadbed with high accuracy and ensure the stable operation of algorithms developed on the basis of the new Cognitive Technologies technology on various road configurations and in various conditions: turns in different directions, climbs, descents, at night, winter, as well as in adverse weather and climatic conditions.

The created technology was called the "virtual tunnel," since it is this shape that resembles a moving sequence of rectangular zones of interest.

What is it like - the moral right to murder?

On December 14, 2015, Cognitive Technologies announced the formation of scenarios for the behavior of an unmanned vehicle.

One of the most important problems that developers have to solve is the behavior of the autorobot in critical situations, when it is impossible to avoid an accident and damage to health and even life is inevitable for participants in the traffic situation. To determine moral law - which of the road users will have to sacrifice in such cases, Cognitive Technologies, with the support of NUST MISIS, MIPT conducts a sociological study among citizens of the Russian Federation. The company intends to use its results as the main material for preparing legislative initiatives regulating the behavior of an unmanned vehicle, both during tests and, subsequently, on public roads.

Studies are supposed to be carried out for various target groups: youth (from 14 to 30 years old), middle (31-50 years old) and senior (over 50). Attention will also be drawn to the difference in the answers of the male and female audience of respondents. In addition, it is assumed to consider the difference in responses among individual social groups.

The organizers of the study prepared descriptions of 5 typical situations involving an autorobot, when it is impossible to avoid an accident, involving 16 behaviors. At the same time, the model considered in the study assumes the complete autonomy of the car, when the driver's intervention in the control process is completely excluded. Respondents will need to answer questions about what they think the behavior of an unmanned vehicle should be in each of the cases.

Among them are the following situations:

• Situation 1. Suppose a self-driving car travels at a speed of 80 km/h. Suddenly, a dog appears in front of him on the road. What should the autorobot do, realizing that emergency braking in this situation is useless? Continue moving and crush the dog, break the rules, with the threat of losing the driver's license, rebuilding into the oncoming lane, separated by a double solid marking line or making an exit into a ditch, realizing that this will lead to imminent global damage to the car and critical injuries to the life and health of the autorobot driver?

840px'upright 'Situation 1 (2015)

• Situation 2. Suppose a self-driving car A moves at a speed of 80 km/h. The scenario involves the sudden appearance of several people on the road in front of him. What should the autorobot do, realizing that emergency braking in this situation is useless? Continue to move and hit people, make a rebuild into the oncoming lane, separated by a double solid lily of markings along which the car B is moving, or make an exit into a ditch, realizing that this will lead to imminent global damage to the car and critical injuries to the life and health of the autorobot driver?

840px'upright 'Situation 2 (2015)

author = Olga Uskova, President of Cognitive Technologies For unmanned vehicles, today it is necessary to create an appropriate regulatory framework - a kind of moral code that regulates the behavior of the artificial intelligence of an unmanned vehicle, including in critical situations that, unfortunately, inevitably arise on the road.

The study is posted on the Internet at: www.cognitive.ru/survey/. Anyone can take part in it.

The company expects to take stock by the end of January 2016.

Notes