Simulating virtual character's learning behavior as an evolutionary process using genetic algorithms

Abstract

In this paper, we describe a genetic algorithm approach to simulate complex virtual character's learning behaviours as an evolutionary process. The method presented here enables virtual character to have abilities to learn for specific assigned tasks. The skill for the task can be developed and evolved through the experiences of performing the task. The animation system presented here has two tightly coupled simulation units, which are an artificial brain unit for learning and controlling and a physics-based motion simulation unit driven by simulated muscle forces.