Generating Human-like Reaching Movements with a Humanoid Robot : A Computational Approach $

Michel Taïx (1) , Minh Tuan Tran (1) , Philippe Souères (1) , Emmanuel Guigon (2, 3)

1 LAAS-GEPETTO - Équipe Mouvement des Systèmes Anthropomorphes
2 ISIR - Institut des Systèmes Intelligents et de Robotique
3 AGATHE - Assistance aux Gestes et Applications THErapeutiques

This paper presents a computational approach for transferring principles of human motor control to humanoid robots. A neurobiological model, stating that the energy of motoneurons is minimized and that dynamic and static efforts are processed separately, is considered. This paradigm is used to produce humanoid robots reaching movements obeying the rules of human kinematics. A nonlinear programming problem is solved to determine optimal trajectories. The optimal movements are then encoded by using a basis of motor primitives determined by principal component analysis. Finally, generalization to new movements is obtained by solving of a low-dimensional optimization problem in the operational space.

Mots clés: nonlinear programming; optimization; motor primitives; humanoid robot; human motor control

Domaines: Robotique [cs.RO]; Automatique / Robotique; Intelligence artificielle [cs.AI]; Automatique

Michel Taïx, Minh Tuan Tran, Philippe Souères, Emmanuel Guigon. Generating Human-like Reaching Movements with a Humanoid Robot : A Computational Approach $. Journal of computational science, 2013, 4 (4), pp.269-284. ⟨10.1016/j.jocs.2012.08.001⟩. ⟨hal-01537858⟩