Simultaneous evolution of leg morphology and walking skills to build the best humanoid walker

Nicolas Jouandeau (1) , Vincent Hugel (2)

1 LIASD - Laboratoire d'Informatique Avancée de Saint-Denis
2 LISV - Laboratoire d'Ingénierie des Systèmes de Versailles

This paper presents an optimization process designed for simultaneous tuning of humanoid's physical characteristics and walking skills' parameters. Starting from the NAO's model used in the 3D Simulation Soccer League, this process focuses on 5 specific parameters to improve individual walking performances. Two policies allow to generate best-first agent or best-average agent. Further improvements may be possible by addressing more parameters or by linking improvements to existing high-level analysis. With these first physical modifications, we achieved to walk 1.5 times faster than the original NAO. In addition results produce more realistic, safer and more precise walk.

Domaines: Intelligence artificielle [cs.AI]

Nicolas Jouandeau, Vincent Hugel. Simultaneous evolution of leg morphology and walking skills to build the best humanoid walker. IEEE-RAS International Conference on Humanoïd Robots 2013, 8th Workshop on Humanoïd Soccer Robots, Oct 2013, Atlanta, United States. ⟨hal-02317135⟩