In Russia, Neural Networks Are Set to Train Robots
Scientists at Lobachevsky University have developed a robot that can learn autonomously through interaction with its environment.
Researchers at Lobachevsky State University of Nizhny Novgorod (UNN) have created a neuro-robot controlled by an artificial neural network modeled on principles similar to those of the human brain. During the study, scientists found that this virtual neural network can effectively control a physical device and adapt to changes in the external environment without relying on predefined scenarios. Products of this kind are currently almost absent from the market.
Obstacles Are No Obstacle
The robot is equipped with ultrasonic distance sensors and a sensitive bumper with touch sensors, allowing it to detect obstacles from multiple directions. As it moves, sensory signals activate neurons in the network, and the robot gradually learns to adjust its behavior. As a result, it begins to avoid obstacles in advance without collisions, demonstrating self-learning without external intervention.
For the Depths and for Space
One of the key features of the development is the system’s ability to relearn. When the placement of sensors changes, the robot adapts to the new configuration and rebuilds its behavioral model. This makes the technology promising for use in rapidly changing conditions, including hard-to-reach environments such as underwater sites or space missions.
From Prototype to Commercial Product
The next stage of the project will be the creation of a commercial product known as a “neuro-constructor.” It could find demand in education, where it can be used to study how neural networks work, as well as in the gaming industry.
The project is being carried out as part of Russia’s National Technology Initiative under the Neuronet track, which focuses on the development of neurotechnologies and intelligent human–machine systems.
