The actual developer of the concept of “Marusya” as an unmanned vehicle is the company Yandex, which launched the project in 2017 to test autonomous driving technologies. It was Yandex engineers who integrated complex neural networks, lidars and radars into the body of a conventional vehicle, turning it into a smart robot capable of recognizing road markings and signs without human intervention. The name “Marusya” has become an unofficial but stable brand for these test prototypes, which are regularly found on the streets of Moscow, Innopolis and other cities with a million people.
Yandex He did not just modify existing machines, but created a complete ecosystem where software interacts with hardware in real time. Unlike standard driver assistance systems such as cruise control or lane retention, algorithms are used to control the driver. self-driving make decisions on rebuilding, braking and bypassing obstacles independently. This requires enormous computing power placed directly in the trunk or under the hood of the car.Many people mistakenly believe that there is only one person behind the project, but the creation of such a complex mechanism is the result of the work of hundreds of specialists. The project was worked by mathematicians, programmers, mechanical engineers and computer vision specialists. Their common goal was to create safe-haven Traditional management, which can reduce the number of accidents caused by human factor, which is statistically the cause of most accidents.
History and evolution of prototypes
The history of the appearance of Marusi originates in the research unit of Yandex, where in the mid-2010s, active work on cartography and computer vision began. The first full-fledged prototype, which went on public roads, was a converted Lexus RX450h. It is on the basis of this model that engineers honed algorithms for behavior in complex urban traffic, collecting data for training neural networks. Later, the park was replenished with models Hyundai Sonata and Kia K5, which also received equipment for autonomous driving.
The development process involved several stages, each of which was marked by an increase in the level of autonomy. While initially the car could only keep a lane and distance, with each iteration of the firmware it learned to recognize flashing traffic lights, respond to traffic controller gestures and predict pedestrian behavior. The key is the integration of high-precision maps, which are updated in real time thanks to data from other machines of the company.
- 🚗 2017 – the launch of the first tests at closed landfills and the start of the trip to the streets of Moscow.
- 🗺️ 2018 – introduction of HD maps and improvement of sensors in winter conditions.
- 🤖 2019 – a demonstration of fully unmanned driving in Innopolis without a test engineer in the cabin.
- 🚦 2020 – scaling up tests and starting work with partners in the field of taxi and freight transportation.
It is important to note that the name “Marusia” is not an official marketing name for the product, but rather a folk or internal designation that has taken root in the media and among enthusiasts. Officially, the company uses the terminology of “Yandex unmanned vehicle” or “robotax”. However, it was the image of Marusi that became a symbol of a technological breakthrough in the Russian automotive ind-oyster, demonstrating that domestic developments can compete with global giants like Waymo or Tesla.
Technical equipment and sensors
Visually, the Marushu is easy to distinguish from a conventional car by the characteristic equipment installed on the roof and bumpers. The basis of the sensory complex is lidar (LIDAR) are laser rangefinders that create a three-dimensional point of a cloud of surrounding space. They allow the machine to “see” objects even in complete darkness, measuring their distance with millimeter precision. The number of such devices on the roof can vary, but there are usually several of them to provide a circular view.
Additionally, the perimeter of the car is controlled by radars operating on radio waves. They are particularly effective at measuring the speed of moving objects and work in poor visibility conditions, such as during heavy rain or snowfall, when cameras may go blind. The combination of radar and lidar data creates redundancy, which is critical for security: if one type of sensor fails or becomes contaminated, others continue to transmit correct information to the data. flight-computer.
⚠️ Warning: The sensors of the driverless car require constant cleaning. Mud, snow or ice on lidar lenses and cameras can cause the system to crash or incorrectly recognize the traffic situation.
High-resolution cameras perform the function of the eyes, reading the colors of traffic lights, road markings and signs. The processing of the video stream is carried out using deep neural networks trained on millions of kilometers of mileage. The system not only sees the picture, it understands the scene: distinguishes the shadow from the pit, predicts the trajectory of the ball rolled out onto the road, and identifies the special vehicle by the characteristic beeps and flashing lights.
How Artificial Intelligence Works at the Driving Room
The heart of the Marousi is a complex software package, often referred to as the car’s “brain.” This software module receives data from all sensors and builds a real-time model of the world around it. Based on this model, the system predicts the situation a few seconds ahead. For example, if the car in front begins to shift to the side, the algorithm will pre-plan a detour maneuver or braking, without waiting for a critical approach.
For navigation, it is not a regular GPS, which has an error of several meters, but a high-precision positioning system that works in conjunction with HD maps. The car knows its position on the road with an accuracy of centimeters. This allows her to know in advance about a narrowing lane, road repair or a difficult intersection before she even gets there physically. Neural network It constantly compares what the cameras see with what should be on the map, and adjusts the route in case of discrepancies.
How a neural network is trained
The training takes place on huge datasets collected from test cars. Engineers detect complex situations (such as non-standard pedestrian behavior) and the algorithm learns from them. In addition, a simulation is used, where virtual Marousi travel millions of kilometers in digital copies of cities, getting into rare and dangerous situations that are difficult to recreate in reality.
Interaction with other participants of the movement is based on predictability. The drone is programmed to comply with traffic rules strictly, but with safety in mind. If the rules require a stop and the car is approaching quickly from behind, the system may decide to use a smoother braking or maneuver to avoid being hit from behind, sacrificing literal follow-up to the instructions for the sake of preserving the integrity of the car.
Safety and emergency scenarios
Safety is a priority in the development of any autonomous system. Marusya is equipped with a multi-level fault tolerance system. All critical components, such as steering and braking, have duplication. If the main control unit fails, the backup immediately takes control and initiates a safe stop of the car on the side of the road with the alarm on.
Special attention is paid to scenarios, which in the development environment are called edge cases - rare and complex situations. It can be an unusual load on the roof of an oncoming car, a person in unusual clothes or destroyed markings. To handle such cases, the system has conservative algorithms: if there is any uncertainty in the correctness of recognition, the object is treated as potentially dangerous, and the car slows down or stops.
- 🛑 Automatic emergency braking in case of sudden appearance of an obstacle.
- 🔄 Reservation of communication and power channels for critical control systems.
- 📡 Permanent telemetry: The control center sees the state of the machine in real time.
- 🚧 Geozonation: the vehicle will not move beyond the permitted area for testing.
The main safety principle of Marusi is the priority of accident prevention over compliance with formal rules of movement in non-standard situations.
Legal status and regulation
The operation of unmanned vehicles in Russia is regulated by separate government decrees and federal laws. For legal movement of the "Marusi" on public roads, a special permit is required. The test driver (engineer) must be in the cabin and be ready to take control at any time, although modern versions allow the car to pass certain areas completely independently.
The legislation is gradually adapting to new technologies. The concept of “automated driving system” has been introduced and requirements for equipment to be installed on a vehicle have been defined. The owner or operator of the drone is responsible for its actions, which implies special insurance and investigative procedures in the event of incidents. Data from flight recorders (“black boxes”) are the main source of information in the analysis of accidents.
| Parameter | Requirement/Status | Note |
|---|---|---|
| Level of autonomy | 4 (high) | Human intervention is required in extreme cases |
| Permission to test | Needless to say | Issued by the Ministry of Industry and Trade and the Ministry of Internal Affairs |
| Driver's presence | Will do (bye) | Test engineer in the cabin |
| Geography | Special zones | Moscow, Tatarstan, Skolkovo, etc. |
Future plans and commercialization
The developers do not stop there and plan to move from test modes to commercial operation. The first step will be to launch robotic taxi services in major cities, where Marusya will be able to transport passengers along predetermined routes. This is expected to reduce the cost of travel and increase their affordability, especially during rush hours when there are not enough human drivers.
In parallel, the possibility of using unmanned technologies in cargo transportation is being considered. Autonomous trucks moving along the tracks in columns (platooning), can significantly improve the efficiency of logistics. In this scheme, the first truck leads the column, and the following automatically repeat its maneuvers with a minimum interval, saving fuel due to aerodynamics.
⚠️ Fully unmanned driving (without an engineer in the cabin) on public roads of a mass nature is still limited by the legislation of most countries, including the Russian Federation.
The technological potential of the project is enormous. In the future, the integration of drones with urban infrastructure (“smart traffic lights”) will completely eliminate traffic jams at intersections and optimize traffic flows. Marusya is not just an experiment, but a step towards a fundamental change in how humanity interacts with the car in the coming decades.
☑️ Ready for the Drone Era
Frequently Asked Questions (FAQ)
Can I buy a marusia in a regular car dealership?
At the moment, Marusya is not a serial product for retail sale. This is an experimental prototype owned by the developer. However, the technologies developed for the project can be gradually introduced into production vehicles of partners in the form of driver assistance systems.
What happens if the internet goes missing while you’re driving?
The car is able to function autonomously without a constant connection to the Internet. All necessary maps and algorithms are downloaded in advance. Communication is used primarily for telemetry, map updates and communication with the dispatcher, but critical control of the machine occurs locally.
Who is responsible if Marusya violates the traffic rules?
Under current legislation, the vehicle owner or the organization conducting the tests is responsible. The test driver is responsible for monitoring the situation and preventing violations. In the future, with the advent of fully autonomous services, the responsibility may be shifted to the system operator or insurance company.
Does the system work in heavy snow?
Working in difficult weather conditions is one of the most difficult tasks. Snow, rain and fog can make it difficult for cameras and lidars to operate. The system has algorithms for cleaning data and protecting sensors, but in extreme conditions, safety requires a slowdown or a stop, so fully autonomous driving in a blizzard is limited.
Why is the project called Marusya?
The project is officially called “Yandex Unmanned Technologies”. The name “Marusya” has become a household name for test cars, perhaps by analogy with popular female names given to the technique, or as a reference to the internal code name in the early stages of development.