Application of genetic algorithm on autonomous agents for virtual navigation in Unity3D

Abstract

This paper presents a software implementation of genetic algorithms on autonomous agents that can navigate a virtual environment created in Unity3D to find an indicated destination. The fitness of the autonomous agents is calculated mainly through a proposed fitness function, which evaluates the performance to heuristically find a destination on three different scenarios with different difficulty and simulation parameters. Based on the obtained results, it is determined that the complexity of the scenarios, the established parameters, and the randomness of the algorithm affect the performance of individuals in finding their destination, but without the need to establish a map of predefined routes.