Experimental study of acquisition of orientational concept based on embodiment of virtual agent

In this study, we discuss a virtual agent´s acquisition of orientational concepts based on physical experience. A virtual agent moves in a computer´s virtual space under restrictions similar to that in the real world. We use a genetic algorithm to simulate and evolve the agent. We then discuss an orientational metaphor that has a basis in physical experience, and we attempt to retrieve clues for solving the generation of concepts by simulating concrete examples, for example, by detecting objects that build a pattern of leg locomotion to approach the object. The pattern is an insignificant set of weight matrices; however, this pattern is grounded in the environment and the body. We then discuss what is represented in the agent action by using the simulation results. In addition, we mention the construction of concepts, such as the GOOD-BAD concept.