Résumé
With this paper we focus on one feature of our contact planner, namely the posture generation, that was under-used so far. We describe it carefully and show we can take advantage of the underlying optimization-based scheme to perform richer planning by the introduction of generic tasks that are not related to locomotion. We then apply with success the resulting improved planner to a challenging scenario and experiment its output plan on a HRP-2 humanoid robot.