MOSCOW, January 20. /TASS/. The Moscow Aviation Institute (MAI) is developing a unique autonomous navigation system that allows drones to operate successfully even, when a GPS signal is lost, it said in a statement.
The algorithms, cameras and sensors allow them to navigate in space, while the built-in neural network analyzes the video stream and detects people.
"We have created a system that allows a drone to ‘understand’ where it is based on its own ‘sensations’ in the form of an inertial system and computer vision. A special camera performs the function of vision, and inertial sensors are responsible for the perception of movements. Our algorithm analyzes incoming information in real time, estimating the position, speed and orientation of the aircraft. Combining these data will ensure high accuracy and stability of navigation," Pyotr Ukhov, associate professor of the Department of Computational Mathematics and Programming, Deputy head of the Department of the MAI IT Center said as quoted by the statement.
The technology is based on SLAM (Simultaneous localization and mapping) visual-inertial navigation, which allows the drone to build its own terrain map and track its position even with difficult maneuvers and loss of GPS signals. A team of students from the Institutes of Computer Science and Applied Mathematics, Aviation, Rocket Engines and Power Plants, and the Advanced Engineering School of MAI is working on the project. The key element of the research was the integration of machine learning algorithms that allow the drone not only to navigate in space, but also to detect people in real time.
Advantages
According to the developers, the solution proposed by the MAI team has important competitive advantages. Many existing analogues either require the installation of expensive and energy-intensive equipment, or operate as separate devices that are not fully integrated into the drone control system. In contrast, in the MAI project, navigation is fully integrated into a hand-made flight controller, which ensures more stable and reliable operation.
"The project was originally conceived as scalable. This will allow the technology to be adapted in the future for various types of unmanned aircraft systems and tasks, such as monitoring hard-to-reach areas, surveying buildings, industrial facilities and emergency zones," Ukhov said.
The development is undergoing field trials. Currently, the team is working to expand the functionality of the system and improve the accuracy of obtaining data on the course and coordinates of the aircraft. Besides, autonomy is increased: it is planned to use human detection to make further decisions, for example, automatically keeping an object in view, clarifying the trajectory, or re-circling the area of interest. Another promising area is to increase the reliability of work in difficult conditions - in poor lighting, smoke, and monotonous scenes. It is planned to bring the project to a complete solution in the second quarter of 2026.