Simulated Robotic Autonomous Agents with Motion Evolution

Abstract

This research implemented autonomous control of robotic agents. The movement controls are simulated within a virtual environment. The control system algorithms were subjected to evolution. Genetic algorithms were implemented to enable the robotic agents to adapt in response to objects within the virtual environment. Additionally, each robot’s physical characteristics were subjected to evolution through a survival of the fittest system based on crossover with random mutations. Survival of the fittest was simulated by a shortage of food causing competition. When the food quantity was increased the evolution rate decreased. With increased food, there was reduced competition and average fitness stopped increasing over time. Removing the food bottleneck stopped the survival of the fittest mechanism.