Cognitive Agro Pilot Automatic Driving System

The main articles are:

• Digitalization in the agro-industrial complex of Russia
• What is the Internet of Things (IoT)
• Artificial intelligence in agriculture (agro-industrial complex)

Cognitive Agro Pilot is an artificial intelligence-based automatic driving system designed for installation on agricultural machinery (tractors, combines).

Composition of the complex

The automated complex, in addition to the artificial intelligence system, includes a video camera and a computer with additional interface units with the on-board systems of the combine.

"Brain" of the system

Cognitive Agro Pilot is based on deep learning technology for neural networks. During 2016, Cognitive Technologies developers in the real field worked to collect data sets (video images on which the main elements of the field scene are presented and on the basis of which the neural network is trained - approx.), and then training the neural network in objects and situations that the "smart" combine can and will meet during its operation and that artificial intelligence must recognize.

As of August 2017, the neural network is able to distinguish five main classes of field scene objects, which will ensure the safe operation of the smart combine:

• The first class is the unskilled part of the field;
• The second class is the processed, beveled part of it;
• The third class of objects corresponds to a roll;
• The fourth class corresponds to that part of the reaper (the mechanical device of the combine, which performs mowing of the agricultural culture), which falls into the field of view of the video camera;
• The fifth grade is all other objects (obstacles, agricultural machinery, forest, trees, weeds, etc.).

Additional features

In addition, the AI of the system is able to "understand" various types of field boundaries (edges, rolls), so that it is not necessary to install additional sensors on the equipment. At the same time, it is possible to promptly add new types of borders (row - a form of planting of individual crops, for example, corn - approx., As well as plowed and unspooled parts of the field). To do this, an update to the software version is enough, which will allow developers in the near future to expand the scope of the system to other agricultural crops, such as corn and sunflower. As of August 2017, the artificial intelligence of the system is trained to work with grain and feed crops.

FTP and investments

Cognitive Technologies is working on the development of the Cognitive Agro Pilot project in accordance with an agreement Ministry of Education and Science Russia with 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." The department allocated 34 million for the implementation of the technological part of the project. rubles

2023

Cultivation of more than 2.3 million hectares of land per year

On December 12, 2023, Cognitive Pilot announced a summary of the work of tractors equipped with the Cognitive Agro Pilot artificial intelligence (AI) autopilot system in Russia in 2023.

Tractors with AI processed more than 2.3 million hectares of land in the first year of use in Russia

This is the first case in Russia of mass industrial use of artificial intelligence on a key fleet of agricultural machinery - tractors. With their participation, the largest number of agricultural operations is carried out: sowing, tillage, fertilization, spraying, etc. We got a decent result. It should be borne in mind that the bulk of the equipment entered the economy and was able to start work only in the second half of the year. In fact, all our main smart technique worked a little more than a third of the season. told Olga Uskova, CEO of Cognitive Pilot

According to the company, in total in 2023, 312 tractors with Cognitive Agro Pilot worked in Russian farms in the period from March to November. Of these, 54 units are Kirovets tractors, mass-assembled by Cognitive Agro Pilot autopilots at the St. Petersburg Tractor Plant, and another 258 tractors from various manufacturers already in use in farms that have decided to robotize them.

According to the InterAgroTech association, the benefit of using the Cognitive Agro Pilot autopilot for one tractor averaged 2.6 million rubles. per year, based on the field size of 1000 hectares and the width of the gun - 5 m. The data obtained took into account resource savings: fuel, seed, chemicals, etc. Other autopilot capabilities that are difficult to assess were not taken into account in monetary terms, the association noted, for example, a decrease in fatigue of the machine operator and a time savings that allows you to complete work on time, reduce the risks of the weather factor, etc.

The association's specialists conducted a study of the results of the Cognitive Agro Pilot system from March to September 2023 in 17 farms in 6 regions of Russia (Krasnodar Territory, Volgograd, Ulyanovsk, Pskov Regions and others).

One of the most important features of the Cognitive Agro Pilot, according to users, was the accuracy of operations (up to 1 cm), as well as the ability to work in RAP conditions, a weak satellite signal and changeable weather.

As of December 2023, such situations arise quite often. They are most indicative in areas bordering the area of ​ ​ the SVO. Our systems have been applied in areas Belgorod and. Kurskaya regions In all these cases, the capabilities computer vision have shown the advantage of our solution over products using only satellite navigation. The equipment equipped with it at times simply turned into a pile of scrap metal. explained in the company

The Cognitive Agro Pilot artificial intelligence-based autopilot system for a tractor combines the capabilities of computer vision and satellite navigation, which ensures stability, accuracy of movement along a given trajectory (up to 1 cm.) And safety. Allows smart tractors to automatically perform turns.

The autopilot system allows you to perform almost all the main operations in autonomous mode: soil cultivation, cultivation, sowing, spraying, fertilizing, grass harvesting, caring for row crops, etc.

The Cognitive Agro Pilot system allows you to optimize the performance of equipment up to 25%, save time and fuel up to 7% due to the accuracy of operations, optimize yields up to 10%, and save fertilizers and seeds up to 20% and 40%, respectively.

The Cognitive Agro Pilot system allows you to automatically transmit to the messengers of tractor owners reports on the progress of equipment with confirmed photos and video data, etc. With remote connectivity, Cognitive Pilot technical support has the ability to work out customer questions online.

The beginning of mass deliveries of Russian autopilots for tractors

Cognitive Pilot (a subsidiary of Sberbank and Cognitive Technologies) at the end of May 2023 officially announced the start of mass deliveries of artificial intelligence-based tractor autopilot systems Cognitive Agro Pilot to Russian agricultural facilities.

Cognitive Pilot noted that in almost two months - from the beginning of April to the end of May 2023 - more than 100 kits were installed in more than 15 regions of Russia from the Pskov region to Blagoveshchensk. Among those using smart equipment: Kursk, Bryansk, Samara, Kurgan, Altai Territory, Krasnodar Territory, Primorsky Territory and other regions.

Cognitive Pilot officially announced the start of mass deliveries of tractor autopilot systems

Cognitive Agro Pilot autopilots based on artificial intelligence are supplied to agricultural facilities, both directly, at the request of farmers (they can be installed on almost all modern tractor models), and under the Cognitive Pilot contract with the St. Petersburg Tractor Plant, where serial equipment of Kirovets tractors is produced.

This is the most important event for our agro-market. For the first time in Russia, the massive industrial use of artificial intelligence began on a key fleet of agricultural machinery - tractors. With their participation, the largest number of agricultural operations is carried out: sowing, tillage, fertilization, spraying, etc., - commented Natalya Filippova, director of the leading industry association InterAgroTech.

According to Yaroslav Sinitsyn, Sales Director of Cognitive Pilot, Cognitive Pilot allows you to increase the performance of equipment up to 25%, save time and fuel up to 7% due to the accuracy of operations, increase yields up to 10% due to the preservation of the fertile layer, as well as save fertilizers and seeds up to 20% and 40%, respectively.

2022

Agreement to install Cognitive Agro Pilot on CNH Industrial combines prior to sale to consumers

The international developer of systems AI for robotic agricultural machinery, the company Cognitive Pilot signed an agreement with the official distributor agricultural equipment CNH Industrial, the distributor of brands Case IH and New Holland in the republic - Kazakhstan the company. Thus, BHK Agro AG Cognitive Pilot is expanding its cooperation with Kazakh suppliers of agricultural machinery. Cognitive Pilot announced this on November 23, 2022. More. here

Inclusion in the register of Russian radio-electronic products for import substitution

The Cognitive Agro Pilot system received an opinion Ministry of Industry and Trade confirming production on the territory RUSSIAN FEDERATION and is included in. register of Russian radio-electronic products This was announced in early November 2022 by the developer of this product Cognitive Pilot. The company became the first certified supplier of autopilots based AI on smart agricultural machinery in the market.

{{quote 'These documents are actually a legal form of complete replacement of foreign manufacturers of autopilot systems on the market with a Russian analogue of a new generation and a higher quality, a real and effective example of the import substitution program in Russia, - said the general director of Cognitive  Pilot Olga Uskova. }}

Russian company began to completely replace foreign agronavigators

By the beginning of 2022, the Cognitive Agro Pilo system was installed in more than 500 domestic agricultural farms in 30 regions of the country. By this time, the number of installations in Russia of foreign parallel driving systems based on satellite navigation, such as Trimble, TopCon, AgLeader and others, according to InterAgoTech, amounted to about 150 thousand.

In March 2022, foreign developers of agronavigation systems announced their withdrawal from the Russian market and the termination of services for their dealerships and client centers.

According to Natalia Filippova, General Director of the InterAgroTech Association, unlike "blind" foreign systems using satellite navigation, the Russian Cognitive Agro Pilot system has serious advantages: it allows technology, as a person, to see and understand the situations taking place on the field and make the necessary decisions. This helps smart technology to avoid incidents with collisions with obstacles, trips outside the field, etc., characteristic of navigation solutions, she added.^[1]

Adding a Navigation Module

Cognitive Pilot (a subsidiary of Sberbank and Cognitive Technologies), on October 25, 2022, announced the implementation of a navigation module in the Cognitive Agro Pilot agricultural machinery autonomous management system, which includes a satellite receiver and an inertial navigation system. Experts in the field of autonomous control call the navigation system the vestibular apparatus of robotic technology. Its task is to assess the change in the position of the machine in space, acceleration and direction of its movement.

The inertial navigation system, as well as the vestibular apparatus in the inner ear of a person, allows you to determine the position of the combine in space. It consists of three accelerometers measuring acceleration along three axes, and three gyroscopes measuring angular speed. The system, among other things, allows you to find the angles of orientation of the equipment - roll and pitch. This helps to compensate for the swinging of the combine on bumps, as well as the tilt of the field. Without taking into account roll and pitch, it is simply impossible to achieve our navigation accuracy of 2-3 cm, "said Anton Emelyanov, technical director of Cognitive Pilot.

The navigation module also provides high precision positioning (GNSS) of agricultural machinery and serves as a complement to the computer vision system. This mechanism is important for performing operations in areas where there are no visual landmarks, in places of delimitation of fields of different farms, when the work takes place on bare ground, in cases where computer vision control is impossible (poor visibility), etc.

It is important that the machine operator can switch the control modes of the combine by computer vision and satellite navigation.

The process of launching a robot in a straight line during autonomous sunflower harvesting along navigation in the Orenburg region

In addition, in agriculture, a lot of work is carried out using autopilot navigation - gouging, sowing, watering. Some cultures are simply more convenient to clean up by navigation. To start the GNSS mechanism, you need to drive 10-15 meters in a straight line. RTK amendments for autopilot are not mandatory, - explained in the company.

The satellite receiver included in the navigation module receives data from all possible global navigation satellite systems: GLONASS, GPS, Galileo and BeiDou. The more satellites, the more accurate the navigation solution. To improve the accuracy of the motion path, if necessary, you can additionally use information from differential correction systems - RTK correction.

The navigation module is installed together with the stereo viewing unit on the roof of the combine cabin

In the future, the possibility of satellite navigation will also allow automatic mapping of yields, setting up cleaning schemes, planning automatic turns of equipment and directions of movement, which can be corrected by the computer vision system.

Autopilot Tractor Sales Announcement

In early June 2022, the Russian developer of artificial intelligence systems for unmanned vehicles Cognitive Pilot announced the imminent launch of autopilot for tractors. Previously, the company developed Cognitive Agro Pilot technology for combines.

Starting July 1, 2022, sales of the Cognitive Agro Pilot tractor version are expected to begin in Russia and abroad. This step, according to Sergey Becker, Sales Director of Cognitive Pilot, will double or triple expand the company's target market in terms of the fleet.

Russian company launches autopilot for tractors on the global market

He noted that Cognitive Pilot expects serial production of about 1,000 tractors with Agro Pilot per year. In 2022, the production of an experimental batch of hundreds of tractors is expected. By the beginning of June 2022, the Company is negotiating with such manufacturers of agricultural equipment as Gomselmash and the Minsk Tractor Plant.

Almost all manufacturers of self-propelled agricultural machinery turn to us with proposals for cooperation. Previously, they equipped their cars with imported systems that are now not supplied to Russia, "Becker said in a conversation with Telesputnik, adding that only a small number of Chinese developers were represented in Russia from foreign manufacturers by June 2022.

The representative of Cognitive Pilot estimates the volume of the Russian market for unmanned control systems of agricultural machinery at 100-120 billion rubles a year. The target market, taking into account Europe and America, is hundreds of thousands of units of agricultural equipment.

Becker stressed that the company does not experience difficulties with the promotion and sale of its products abroad, since they are not subject to sanctions restrictions. Cognitive Pilot continues to sell abroad and open representative offices, conclude agreements with dealers.^[2]

2021

Russian Cognitive Pilot signed 47 contracts for the export of autopilots for agricultural machinery for the year

Cognitive Pilot for 2021 signed 47 contracts for the export of autopilots for agricultural machinery Cognitive Agro Pilot. The company announced this on December 8, 2021.

As the CEO of Cognitive Pilot Olga Uskova clarified, it was possible to start active deliveries of autonomous harvesting based on artificial intelligence only closer to mid-2021, but then the company began to quickly attract customers in Europe, North and Latin America.

Russian Cognitive Pilot signed 47 contracts for the export of autopilots for agricultural machinery for the year

In a sense, thanks to the Cognitive Pilot, 2021 has become a turning point in the world's unmanned vehicle market. If earlier in this segment, as a rule, only prototypes of drones were presented as players, now we are in full swing around the world there have been massive implementations of smart autonomous equipment - our agropilots, - said Uskova, adding that 47 contracts around the world are "a breakthrough result for the post-pandemic start."

In countries European Union , the company has signed 10 contracts. On 600 hectares in four farms Belgium Italy , equipment with Cognitive Agro Pilot collected 36 thousand centners of wheat and corn.

The average cost of a deal for the supply of Cognitive Agro Pilot worldwide in 2021 amounted to $8.9 thousand. Dealer rewards for exports range from 20% to 30%.

In addition, in the United States, where the company's technologies are used in eight out of 16 agricultural states, Cognitive Pilot Corporation (an American subsidiary of Cognitive Pilot) has partnered with Nvidia and entered the Nvidia Inception program, focused on supporting developers in the field of artificial intelligence, processing large amounts of data and high-performance computing. Cognitive Pilot Corporation will receive Nvidia resources to develop new solutions for agricultural machinery.^[3]

Availability for small businesses served in Sberbank

On June 25, 2021, Cognitive Technologies announced that the Cognitive Agro Pilot system for autonomous management of agricultural machinery based on artificial intelligence is now available to small entrepreneurs serving in Sberbank. Previously, only customers of large and medium-sized businesses could purchase this high-tech solution.

In May 2021, we launched a pilot project to sell Cognitive Agro Pilot to collective farms and small agricultural enterprises with from one to ten combines. And we have already implemented the first transaction in the amount of about 1 million rubles with a farm from Saratov. About 130 more customers have expressed their interest in this product, for June 2021 we are negotiating with them. The share of small enterprises in the agro-industrial complex is a significant part - 50%, so it is extremely important to increase the digital transformation of this segment to increase business efficiency not only of individual companies, but also of the industry as a whole, "said Alexey Shashkin, Director of the Maly and Microbusiness Division of Sberbank.

We have already covered more than 30 regions and are seeing an increased interest of small businesses in cleaning robotization. The decision-making mechanism here works faster, as we interact directly with the owners on many issues. To promote products in this zone, we use both traditional and non-standard channels, for example, unprofessional video blogs conducted by farmers and combines. There are millions of views. They appreciate the opinion of their colleagues and such a "word of mouth" mechanism works perfectly, - said Sergey Becker, Sales Director of Cognitive Pilot.

Small business can purchase the Cognitive Agro Pilot system for any brand of combines on favorable terms from SberLeasing. For the convenience of the client, the transaction can take place in a remote format, which greatly simplifies and speeds up the process. Leasing conditions - advance payment from 0%, term - up to 3 years, schedule - annuity, differentiable, seasonal, individual.

2020

Harvest by harvesters with Cognitive Agro Pilot 720 thousand tons of crop

Harvesters with artificial intelligence collected 720 thousand tons of crop. Sberbank announced this on November 3, 2020.

The Cognitive Agro Pilot system for autonomous management of agricultural machinery using artificial intelligence, developed by Sberbank and Cognitive Pilot, part of its ecosystem, was used industrially in 35 regions of Russia during the 2020 harvest season.

From June to October 2020, in autonomous mode, more than 350 combines equipped with Cognitive Agro Pilot processed over 160 thousand hectares of area and harvested more than 720 thousand tons of crop.

With the help of domestic robotic harvesting technologies on the basis of artificial intelligence, 590 thousand tons were harvested per 130 thousand hectares of grain crops (wheat, soybeans, barley, oats, sorghum, buckwheat, etc.), as well as about 130 thousand tons per 30 thousand hectares of row and roll crops (corn, sunflower, etc.) in,, Kaliningrad,, Kaluzhskaya,, Kurskaya,, Belgorod,,,,, and Tambovskaya Penza. Rostovskaya Tomsk Kurganskaya regions Krasnodar Krasnoyarsk Stavropol Territory

Due to the use of unmanned Cognitive Agro Pilot technologies this cleaning season, the cumulative scale of savings - on fuel, related materials, reducing the time of the harvesting campaign (the number of engine hours) and the active operation of combines, depreciation of equipment, extending the active operation of equipment to capital investments, reducing the human factor, optimizing business processes and other parameters - amounted to more than 500 million rubles.

Anatoly Popov, Deputy Chairman of the Management Board of Sberbank:

A few years ago, artificial intelligence in agriculture was a technology that was talked about a lot, but there were no practical examples of its implementation in Russia. For November 2020, according to the results of the first harvesting campaign with smart combines, it is safe to say that the future is becoming real before our eyes and, which is especially pleasant, with the direct participation of Sberbank. Our country is one of the leading countries in terms of agricultural production and is a key exporter of wheat, therefore, given the scale of production, the Cognitive Agro Pilot system can potentially save billions of rubles to the domestic agro-industrial complex.

Olga Uskova, CEO of Cognitive Pilot:

In total, under the control of the Cognitive Agro Pilot system, the equipment worked in 2020 for more than 230 thousand hours, covering over 950 thousand km. These are significant indicators that indicate the reliability of the system and its demand among farmers. Partnership with Sberbank and leading agricultural holdings allows us not only to scale the use of technology, but also guarantees our customers high-quality service on the ground.

Bierne Drexler, First Deputy General Director of the management company EcoNivaTechnika-Holding:

{{quote 'Equipping the combines with an autonomous control system made it possible to significantly increase the efficiency of harvesting work. Given the wide dealer network of the EkoNiva technical holding and the fact that the system can be installed on any combine, regardless of the manufacturer's brand and model of the combine, we can confidently talk about the implementation and participation in the largest industrial project for the robotization of agricultural machinery. One of the first capabilities of Cognitive Agro Pilot was appreciated by the combines. The system really helps in work, reduces the load, allows you to focus on other cleaning parameters. All this increases the efficiency of business processes and saves resources. " }}

Plan to include Agro Pilot in agroecosystem for accelerated digitalization of agriculture

On June 25, 2020, it became known that Digital Agro, Agrosignal and Cognitive Pilot will create a single agroecosystem for the digitalization of Russian agriculture as part of a strategic partnership .

As reported, the agroecosystem will include:

• Interface, algorithms for analyzing and processing big data (Big Data), as well as mobile applications for users from Digital Agro.
• The Cognitive Agro Pilot software and hardware complex is a system for autonomous control of agricultural machinery (combine harvester, tractor, sprayer) based on artificial intelligence (AI) technologies. Read more here.

2019: Cognitive Technologies to introduce self-driving system in some agricultural facilities of Tomsk Oblast

On May 13, 2019, Cognitive Technologies, a developer of artificial intelligence systems for unmanned vehicles, announced the signing of agreements on the introduction of unmanned driving systems for agricultural machinery with a number of leading agricultural enterprises in the Tomsk Region. Among them are Agro LLC, Agroholding Tomsky OJSC, Siberian Milk LLC, SEC Mezheninovsky and others. Read more here.

2018

Completion of field tests of agrodroid

On September 19, 2018, the company Cognitive Technologies announced the successful completion of field tests of agrodroid, a universal intelligent device for ensuring the autonomous movement transport of agricultural equipment. The device performing the function artificial brain can be operatively connected to combines, tractors, sprayers, etc. The agrodroid is supplied with video cameras and other sensors, depending on the specific task.

During the tests, harvesters equipped with agrodroid went offline for more than 200 km and participated in harvesting more than 60 hectares of grain crops. The system was tested in all possible operational modes.

author '= Yuri Minkin, Head of Unmanned Vehicle Development at Cognitive Technologies ' We conducted tests in difficult conditions. This is a strong dust that was created by five combines working in close proximity to us. We had to test the capabilities of our computer vision in border situations. This is also a different field texture that did not require additional neural network settings. And a complex terrain topology. We had to harvest on steep climbs and slopes when the combine's roll angles reached 25 degrees. In addition, about 20% of the work time was carried out in the dark.

Technical characteristics of agrodroide include:

• autonomous movement along the edge plowed/not plowed;
• autonomous movement beveled/unshaded;
• autonomous movement in rows (form of planting of individual crops: corn, sunflower, soybeans, etc.);
• autonomous movement on rolls (beveled and folded into a row of agricultural culture);
• autonomous u-turns at the end of the field;
• autonomous unloading of crops into a bunker or dump truck;
• the possibility of working in night conditions^[4].

Industrial Agrodroid C2-A2 (Cognitive2-Agro2 Droid1)

Cognitive Technologies introduced in the spring of 2018 an industrial agrodroid, that is, an "artificial brain" for the agricultural market. Agrodroid C2-A2 (Cognitive2-Agro2 Droid1) is an industrial model of a universal system for managing unmanned agricultural transport. That is, a kind of artificial intelligence for agricultural machines.

Agrodroid makes agricultural machines autonomous. The solution turns any farming into smart farming, and supplying it with a cradle prefix makes it possible to transfer the brain from one unit of agricultural machinery to another without ordering a new system every time.

The company plans to conduct industrial tests C2-A2 in the summer of 2018 in Russia and Brazil. It is assumed that the final cost of the device will be about $3 thousand, which is about 1.5% of the cost of the combine and about 3% of the cost of the tractor.

"We estimate the volume of the agrodroid market at $94 billion and expect to occupy at least 15% on it in the next 3-5 years. The world fleet of tractors alone, ready for cognitive droidization, today is about 27 million units, "notes Olga Uskova.

2017

Night tests of an unmanned combine

Cognitive Technologies on October 25, 2017 presented the next version of the automatic driving system of the Cognitive Agro Pilot unmanned combine. Unlike the previous one, this version is capable of working at night, which will allow combines to harvest around the clock, the company said.

According to Olga Uskova, President of Cognitive Technologies, according to preliminary estimates, this opportunity will minimize the impact of the weather and human factor and improve the quality of grain harvesting by 25% -30%.

In general, the updated Agro Pilot system has learned to perform more complex maneuvers than before. At any time of the day, the combine equipped with the system is capable of turning itself or moving along a complex and curved trajectory.

The combine works at a maximum speed of 11 km/h, while it automatically monitors the edge of the harvested crop. The system itself will determine if the collection boundary has begun to deviate from the straight line - the combine will simply adjust the movement and go further. So far, 11 km/h is the maximum possible speed at which a stable and efficient harvest is ensured.

At the same time, the combine under the control of Agro Pilot is able to determine not only the edge of the field: the system recognizes equipment, people, animals, understands where the field has already been processed and where it has not, the beveled culture for it differs from that which is still growing, emphasized in Cognitive Technologies.

In theory, the number of objects for segmentation can be increased, which will distinguish different cultures growing in the same field. The main sensor through which the system perceives the world is the camera. It is noted that the combine does not need to have high-precision GPS: it can be completely autonomous, relying only on the video signal. In addition to an artificial intelligence system and a video camera, Agro Pilot includes a computer with additional interface units with the combine.

The tests were carried out jointly with the manufacturer of combines - Rostselmash, which provided an experimental sample of the TORUM 760 combine, on which the system was installed. The robot was given a rice field with an area of ​ ​ 40 hectares. In the dark and without GPS, the machine successfully passed test harvest tests.

Night tests of an unmanned combine, October 24, 2017

Our computer vision system allows you to achieve similar results with one video camera as leading Western brands with three to four sensors. As a result, the cost of our equipment is three to four times less, - says Olga Uskova.

While the combine is working when the driver is in the cab. Cognitive Technologies promises to present a fully autonomous combine by 2023.^[5]

Field tests of an unmanned combine

On August 15, Cognitive Technologies announced the first field tests of the combine in unmanned mode. The work was carried out in the Rostov region together with the company's partner - Rostselmash. An experimental example of the RSM 181 TORUM combine was equipped with the Cognitive Agro Pilot automatic driving system.^[6]

According to the developer Cognitive Agro Pilot, in the Russian project, the set of sensors is only one video camera, unlike foreign analogues, which use laser scanners to move along the edge of the field and stereo cameras to work on the roll (beveled and folded into a row with/x culture).

We spent more than five years on developing the system and got excellent results, "said Olga Uskova, president of the Cognitive Technologies group of companies. - We were able to create a highly developed computer vision system that allows us to achieve similar results with one video camera as leading Western brands with three to four sensors. As a result, our equipment cost is three to four times less. This gives us a significant competitive advantage, especially now that we are actively entering international markets.

AI and combiner roles

During the harvesting works, the driver of the experimental model of the combine must be present in the cabin and act as a controller. At the same time, the level of his qualification should remain at the same level as that of a regular combine harvester. However, the intelligent system allows you to significantly reduce the routine load on the driver, as a result of which he has time to focus on choosing settings for technological process parameters that directly affect the quality of harvesting (reaper angle, speed, etc.), which is one of the main problems in cleaning, emphasized in Cognitive Technologies.

The technological process of controlling the unmanned combine is arranged so that its movement is carried out along the edge of the field or along the laid roll. The AI system signals that it sees an edge or roll and is ready to take control. With this message, the driver just needs to press the autopilot button to transfer control and confirm this action. At the same time, the system itself will understand where the border of the beveled culture passes and where it is necessary to steer to ensure movement along the edge, "explained Alexei Panchenko, project manager at Cognitive Technologies.

Under the control of Cognitive Agro Pilot, the combine is able to automatically steer, make turns until it reaches the end of the run - until the end of the circle (to the place where the field boundary breaks) or to the perpendicular edge. In this case, the combine harvester is given a signal to take control. If control is not taken, the combine stops.

Unforeseen situation

According to Alexei Panchenko, an unforeseen situation arose during the tests: instead of planned work on harvesting wheat with an ear height of more than 80 cm, the system had to be tested on small grain crops in a barley field with an ear height of 30-40 cm.

In such a situation, the roll was almost impossible for the human eye to distinguish from the beveled culture. This is an unprecedented case. It was a challenge for us. For a day and a half, we trained the neural network on new data sets. As a result, the neural network was able to determine the laid roll better than a person. The result exceeded all expectations, - concluded Alexei Panchenko.

System Development Plans

Following the summer tests, Cognitive Technologies plans to complete the necessary improvements and move on to the pilot series.

With the possibility of unmanned driving, Russian combines will be able to compete on equal terms with the world's leading agricultural brands, which already have a similar their own function. We expect to create a fully unmanned combine by 2023-2024, "added Olga Uskova.

AI system for unmanned combine harvester

Cognitive Technologies announced in April 2017 computer vision technology for automatic driving of a combine harvester.

"It is assumed that our system will first be installed on an experimental sample of agricultural machinery," said Olga Uskova, President of Cognitive Technologies. - We spent more than three years on its development and got excellent results. Having the function of automatic taxiing during harvesting, Russian combines will be able to compete on equal terms with world agricultural brands that already have similar functionality. "

"The development is based on the technology of deep learning of neural networks, or deep learning," said Alexei Panchenko, project manager at Cognitive Technologies. Over the past year, Cognitive Technologies developers have been collecting data sets in real field conditions from a camera installed on a combine (video images on which the main elements of the field scene are presented and on the basis of which the neural network is trained), and then training the neural network in objects and situations that the "smart" combine can and will meet during its operation and that artificial intelligence must recognize.

As a result, today the neural network is able to distinguish five main classes of field scene objects, which will ensure the safe operation of the "smart" combine. The first is the undisturbed part of the field. The second is the processed, beveled part of it. The third class of objects corresponds to a roll (beveled and folded into a row of s/x culture). The fourth corresponds to that part of the reaper (the mechanical device of the combine, which performs mowing of the agricultural culture), which falls into the field of view of the video camera. Finally, the fifth grade is all other objects (obstacles, agricultural machinery, forest, trees, weeds, etc.).

According to the developers, during the harvesting work, the driver of the experimental model of the combine will still have to be present in the cabin and act as a controller. At the same time, the level of his qualification should remain at the same level as that of a regular combine harvester. However, he will have free time in order to focus on choosing settings for technological process parameters that directly affect the quality of harvesting (reel angle, speed, etc.), which today is one of the main problems in cleaning, Cognitive Technologies emphasized.

"Automatic movement of the combine with an artificial intelligence system will be carried out along the left edge of the field or along the beveled roll. At the same time, artificial intelligence itself will understand where the border of the beveled culture passes and where it is necessary to steer to ensure movement along the edge, "Panchenko explained

In certain situations, for example, when completing the passage through the field, as well as in cases of a booster on the signal of the system, the driver must be ready to take control.

The automated complex of the smart combine, in addition to the artificial intelligence system, includes one domestic video camera and a computer with additional units for interfacing with the onboard systems of the combine.

The competitive advantage of the new solution Cognitive Technologies sees the possibility of understanding by artificial intelligence of various types of field boundaries (edges, rolls), in contrast to the solutions of competitors who understand only one of the types of boundaries, and to determine it, it is necessary to install additional sensors corresponding to them, which ultimately determines the higher cost of the solution.

In addition, the Cognitive Technologies system provides the ability to promptly add new types of borders (row - a form of planting of individual crops, for example, corn, as well as plowed and unswept parts of the field). To do this, an update to the software version is enough, which will allow developers in the near future to expand the scope of the system to other agricultural crops, such as corn and sunflower. Today, the artificial intelligence of the system is trained to work with grain and feed crops.

Work on the project is carried out by Cognitive Technologies in accordance with an agreement with the Ministry of Education and Science of Russia 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." For the implementation of the technological part of the project, the department allocated p34 million.

See also

• IT in the agro-industrial complex of Russia
• AgroBot Unmanned tractor
• Case IH Magnum
• FoodNet Roadmap (Smart Agriculture)
• National Technology Initiative (NTI)

• thematic section of news and articles "Agriculture and fisheries"
• IT systems implemented in R&D companies
• catalog of implementation projects in s/x companies
• ttop IT companies working with s/x companies
• developers of systems that are used in s/x companies
• s/x companies in TAdviser catalog

Notes