Development of a self-learning mobile robot control system based on neural networks with long-term plasticity

The use of spiking neural networks as control systems is an urgent and promising task. Unlike artificial neural networks, spike networks are less demanding on energy resources and have greater computing potential. Also, they are much closer in their structure and functioning to biological neural networks. Therefore, it is very interesting to train a spiking neural network based on the experience obtained with the help of external sensors. Understanding the techniques and methods of training spiking networks will allow them to be used to solve other problems. A spiking neural network with long-term plasticity was trained to control the robot in one-dimensional space. The network independently learned to stop the robot when it receives a signal from a distance sensor that has noticed an obstacle, without touch sensors touching the latter.