Analysis of the cognitive properties of neural systems based on biofeedback

Abstract

Problems of cognitive system reengineering, i.e., the development of devices with cognitive properties on the basis of their biological prototypes, cannot be solved without understanding the basic features of the architecture of biological systems, information properties, and molecular organization of the primitive units forming the architecture: nerve cells. The construction of learning models makes it possible to study the activity of individual cells, not only in terms of behavioral responses to natural stimuli but also in experiments on isolated preparations with excitation of peripheral bodies and isolated cell structures. The software-tool complex NeuroFeedBack was developed; it includes a system of living neurons and a neurocomputer interface feedback. The complex provides the reception and processing of input signals from neurons, their visualization and storage, as well as the generation of output reinforcing stimuli applied to the neurons. Analysis of the functional activity of neurons of the right parietal ganglion of the Lymnaea stagnalis mollusk was performed with the complex in three models of reinforcement. It was shown that optimization of neural activity occurred under conditions of biofeedback, allowing the neuron to minimize the quantity of the reinforcing stimuli. The results provided grounds for the design of a hybrid robotic system in which living neural systems using a neurocomputer interface could solve navigation tasks, controlling a real-time mechanoelectronical device operating in a real environment. In the experiment, the use of the proposed programs of hybrid reinforcements allowed the robot to find a glowing light bulb and reach it in a few minutes.