Optimisation de la locomotion de robots bas coût à pattes. (Optimizing locomotion on low-cost legged robots).

A promise of legged robots is being able walking on irregular or uneven oor. It is already used nowadays in education and entertainment applications. We introduce the Metabot platform, an open-source quadruped robot developped for education. This takes place in current technological context which allows, thanks to an access to fast prototyping and o -shelf components, building legged robots that were formerly only present in laboratories. This platform was used for teaching in secondary schools, allowing students to discover robotics, and especially programming. We then describe an environment designed to study legged robots locomotion, extending the expert controller designed for Metabot. We realized some physics simulation experiments and analyzed it to get a better understanding of the legged locomotion underlying rules. At last, we get a closer look at biped locomotion, for which stability problems arise. When developping our Sigmaban platform, a small-sized humanoid robot created to participate in RoboCup soccer, we designed foot pressure sensors that allow us to estimate the robot center of pressure. We exploit these sensors to improve the lateral stability on the robot, creating a closed-loop walk. Optimisation de la locomotion de robots bas coût à pattes iii