Engineers develop robot that can self-learn with help of neural network

Scientists from Lobachevsky State University of Nizhny Novgorod (UNN) have developed a robot that can self-learn with the help of neural network
 
The robot is equipped with ultrasonic distance sensors and sensitive bumpers with touch sensors which allow it to feel obstacles on the right and left side, and the activity of sensory neurons directly depends on the activity of these sensors.
 
Upon receiving information from the sensors about the obstacle that lies ahead, the robot tries to find ways to get around it. After the first successful attempt, he fixes this algorithm as correct and uses it to prevent the next collisions, Hightech.fm reports.

---