2A1-M05 Behavior Acquisition of a Virtual Robot by Gradual Learning
Ryosuke Ooe, Takashi Kawakami · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2015
This paper describes a behavior acquisition of virtual robots by a gradual learning. The gradual learning is a method in which initial states of simulation for evaluation is changing in evolution progress. Motion of virtual robot is calculated by the physical engine PhysX, and it is controlled by artificial neural network (ANN). Parameters of an ANN are optimized by particle swarm optimization (PSO) so that a virtual robot follows the given target. Experimental results show that the gradual learning is better than a general learning, and that a random initialization of parameters in the middle of evolution leads to better evaluation. The best motion given by the gradual learning is composed of some basic motions, which are controlled by the ANNs used for change of initial states.