A Study on Acquiring Models Criterion Focusing on Learning Efficiency

M. An, K. Kagawa, and T. Taura (Japan)


Imitation Learning, Robotics, Generic Algorithm, GeneticProgramming


This paper presents a method using Genetic Algorithm (GA) and Genetic Programming (GP) to presume mimetic object's behavioral criterion which can be regarded as intention of an action. The latest research results obtained from our simulation have proved a theory that if a robot presumes a behavior criterion of a mimetic object correctly, it will learn the object's behavior efficiently. Therefore, it can be judged whether a robot got a mimetic object's behavioral criterion by evaluating the robot's learning efficiency. For this purpose of the investigation, two robots are created in our simulation to performance a reaching behavior and a learning behavior. The simulation results based on these methods also will be presented.

Important Links:

Go Back