A software system for modeling evolution in a population of organisms with vision, interacting with each other in 3D simulator.

Abstract

Development of computer models imitating the work of the nervous systems of living organisms, taking into account their morphology and electrophysiology, is one of the important and promising branches of computational neurobiology. It is often sought to model not only the nervous system, but also the body, muscles, sensory systems, and a virtual three-dimensional physical environment in which the behavior of an organism can be observed and which provides its sensory systems with adequate data streams that change in response to the movement of the organism. For a system of hundreds or thousands of neurons, one can still hope to determine the necessary parameters and get the functioning of the nervous system more or less similar to that of a living organism - as, for example, in a recent work on the modeling of the Xenopus tadpole. However, of greatest interest, both practical and fundamental, are organisms that have vision, a more complex nervous system, and, accordingly, significantly more advanced cognitive abilities. Determining the structure and parameters of the nervous systems of such organisms is an extremely difficult task. Moreover, at the cellular level they change over time, these including changes under the influence of the streams of sensory signals they perceive and the life experience gained, including the consequences of their own actions under certain circumstances. Knowing the structure of the nervous system and the number of nerve cells forming it, at least approximately, one can try to optimize the initial parameters of the model through artificial evolution, during which virtual organisms will interact and survive, each under the control of its own version of the nervous system. In addition, in principle, the rules by which the brain changes during the life of the organism can also evolve. This work is devoted to the development of a neuroevolutionary simulator capable of performing simultaneous functioning of virtual organisms that have a visual system and are able to interact with each other. The amount of computational resources required for the operation of models of the physical body of an organism, the nervous system and the virtual environment was estimated, and the performance of the simulator on a modern desktop computing system was determined depending on the number of simultaneously simulated organisms.