The research presented in this thesis report focuses on the optimization of crossing capabilities for "wheel-legged" hybrid mobile robots. We consider that a robot is able to overcome an obstacle if he is able, for each geometric configuration of the contact points encountered during the crossing , to balance the external forces applied to it while maintaining contact adhesion. Our approach consist in using two features of poly-articulated robots in order to optimize the distribution of contact forces. Firstly, the actuation redundancy allows the application of internal forces. Secondly, the kinematic redundancy allows to change the CoM position, changing the set of possible solutions to the force distribution problem. The criterion used for the optimization measures the contacts stability robustness relatively to uncertainties in the force control. It was chosen based on a comparative analysis of different criteria used in the field of grasping. The resulting control algorithm allows for simultaneous control of both robot movements and contact forces. Experimental results on robot HyLoS 2 show that the algorithm allow for the crossing of an obstacle represented by a slope of 60\char6 with a height of 19 cm, while the wheels radius is only 7 cm.
