Neuromorphic model of associative memory in the formation of the working environment description of a robotic agent

Abstract

The article considers an approach to solving the problem of describing the workspace of a robotic agent based on the neuromorphic principle. A segmental spike model of a neuron with the possibility of structural adaptation was used as a basic element in the modeling. The main features of the neuron model used in terms of the possibilities of its structural reconfiguration are indicated. As the basis of the model for such a description, a scheme for the formation of a conditioned reflex in living organisms is chosen. The structural scheme of the neuromorphic model of conditioned reflex formation is given, as well as the general scheme of the model of associative memory formation. A step-by-step description of the algorithm for forming associative links in such a scheme is given. The principle of the formation of inhibitory connections when competing features appear in several objects is presented. The article presents the results of computer modeling on a model example. The conclusion is made about the applicability of the chosen neuron model and the scheme of organizing neurons into a network to solve the problem of describing the workspace of a robotic agent.