They've gone crazy: the first Russian microcomputer with AI for drones has gone on trial

Russia has developed a microcomputer that combines the functions of a computing module for artificial intelligence and a flight controller for robotics. The solution is designed for drones, ground-based robots and other autonomous systems, allowing data processing and decision-making directly on board. According to the authors of the project, there are no analogues of such a platform on the domestic market yet. Now the technology has begun to be tested in conditions as close as possible to real ones. It should be borne in mind that it may take at least a year before practical implementation, experts noted. Whether a domestic microcomputer can replace Western developments is in the Izvestia article.

Universal "brain" with AI for robots and drones

The Russian technology company Aerospace Antenna Systems has developed a microcomputer for performing artificial intelligence tasks and controlling robotics. This was reported in the press service of the Platform of the National Technology Initiative. The solution can be integrated into unmanned aircraft systems, unmanned boats, ground-based robotic complexes, computer vision systems, smart cameras and industrial IoT (Internet of Things). There are no analogues in the domestic market, the developers emphasized.

— Our team has developed a flight microcomputer Brain Box. It is a powerful single-board computer specially designed for AI and robotics control tasks. The computer's "brain" is combined on one board, as well as a high—performance neuroprocessor, a specialized chip or computing unit designed for hardware acceleration of AI algorithms," explained project manager German Yangalin.

Photo: Provided by NTI

According to him, usually a drone or robot has an AI computer and a flight controller — these are two different boards that are connected by wires. Because of this, data is transmitted with a delay, and the design itself is cumbersome and unreliable. The microcomputer created is all in one: a powerful processor for artificial intelligence, engine control and navigation on a single board. This ensures an instant reaction — the robot sees an obstacle and makes a decision without delay.

— Low power consumption can also be highlighted among the advantages of the device: the entire system consumes only 12 watts, about the same as a regular light bulb. Another advantage is versatility: various sensors, including cameras, lidars and thermal imagers, can be connected to the platform. In addition, the built-in Russian encryption system protects against attempts to intercept control. Simply put, it is a computing module for various equipment that does not require additional boards to expand its functionality," said German Yangalin.

The technology supports Linux operating systems and can be ported to Astra Linux, Kaspersky Os, Alt, Aurora and others.

Quantum of Mercy: how Russian scientists have changed the balance of power in the technological race

In just five years, Russian specialists have become leaders.

The demand for Russian developments

According to the developers, American companies such as ModalAI, NVIDIA, and others have similar solutions. They work on foreign chips (Qualcomm, NVIDIA) and proprietary software. However, their own technologies are necessary for reasons of independence and security.

— We don't know if there are hidden bookmarks in the imported chips. Our system encrypts data according to Russian standards, and we control every step. We can adapt the board to the harsh climate, Russian satellite and domestic operating systems. This is a matter of both security and technological sovereignty," stressed German Yangalin.

Photo: RIA Novosti/Maxim Bogodvid

Today, industries increasingly need not just separate computing modules or flight controllers, but compact devices that are simultaneously capable of controlling equipment and performing artificial intelligence tasks directly on board, Anton Averyanov, CEO of the ST IT group of companies, TechNet NTI market expert, told Izvestia.

— This is the main value of such platforms. They allow you to process data from cameras, lidars, thermal imagers and other sensors without having to constantly transfer information to the cloud or a remote data center," the expert said.

The segment of single-board computers focused on machine vision and AI tasks is extremely competitive today. Global hardware architectures have already been formed around key global manufacturers of specialized chips, from NVIDIA to Qualcomm, Dmitry Kuzyakin, chief designer of the Center for Integrated Unmanned Solutions (CCDB), told Izvestia.

— There is a clear division in the industry into universal and highly specialized solutions. On the one hand, there are massive platforms like the NVIDIA Jetson line, designed for a wide range of tasks. On the other hand, there are specialized computing modules of the Qualcomm RB3 Gen 2 level, which are optimized for specific harsh conditions and are used, for example, in Hornet—class unmanned vehicles," he explained.

Revolution on the Table: Nvidia's New chip will Disrupt the Computer Market

The world's largest corporation has introduced a "killer" of competitors

According to him, all these solutions are created with a single goal — to integrate inside drones and robotic complexes to ensure their complete autonomy and independence from the operator due to AI algorithms.

Photo: IZVESTIA/Polina Violet

— However, it is critically important to understand that not only hardware plays a key role here. Every major vendor builds a powerful ecosystem around their processors, consisting of customized platforms and a global community of developers. It is the programmers who create libraries and ready—made software that determine the final scenario for using the platform," Dmitry Kuzyakin clarified.

The development demonstrates the transition to the practical creation of specialized solutions for edge computing - data processing directly on devices. Such technologies are of critical importance for the development of unmanned vehicles and robotics, said Yaroslav Seliverstov, a leading expert in the field of AI at University 2035. At the same time, there is still a significant way to go before the full-fledged practical implementation of the technology.

"It will take at least a year of intensive testing and substantial budgets, especially given the need to adapt the driver base to domestic operating systems," he said.

An experimental sample of a microcomputer is currently being tested in conditions as close as possible to real ones. These checks allow you to confirm the device's operability and assess its readiness for further implementation. The next stage will be field testing of the development.