2P1-B04 視覚によるモデル獲得に基づく動力学シミュレータを用いた予測行動プランニングシステム(動作計画と制御の新展開)

Abstract

To realize the prediction motion planning, it requires making models by recognition using vision, planning using dynamics calculation, and reducing the difference between real and simulation. These needs relation between recognition action programs and dynamics simulators. This paper picks up the block stacking problem and realize the prediction motion planning on EusDyna, which is one of the environments that integrate robot brains and simulations. Recognition, modeling, planning and decrease the error of prediction are illustrated in a humanoid robot.