This thesis sets up terrain parameters and locomotion strategies of a hexapod robot walking on variable and discontinuous terrain. Walking on this kind of terrain is the greatest advantage of legged robots compared with wheeled robots. First, establish a randomly distributed parameterized terrain. Second, set up morphological parameters and dimension parameters of the robot. Third, build kinematic model and generate continuous gaits of the robot, including crab gaits and turning gaits. The locomotion strategy can determine every AEP ,PEP and stride depending on terrain. Finally, verify the strategy through computer programming and find shorter path by calculating if foothold is available in advance. Because of applying randomly distributed parameterized terrain, in addition to describing the terrain more comprehensively, the terrain parameters can be adjusted easily according to different needs. This research will bring about more applications and developments of legged robots.

Dans cette thèse nous nous intéressons à deux types d'approches pour la planification de pas pour robots humanoïdes : d'une part les approches discrètes où le robot n'a qu'un nombre fini de pas possibles, et d'autre part les approches où le robot se base sur des zones de faisabilité continues. Nous étudions ces problèmes à la fois du point de vue théorique et pratique. En particulier nous décrivons deux méthodes originales, cohérentes et efficaces pour la planification de pas, l'une dans le cas discret (chapitre 5) et l'autre dans le cas continu (chapitre 6). Nous validons ces méthodes en simulation ainsi qu'avec plusieurs expériences sur le robot HRP-2. / In this thesis we investigate two types of approaches for footstep planning for humanoid robots: on one hand the discrete approaches where the robot has only a finite set of possible steps, and on the other hand the approaches where the robot uses continuous feasibility regions. We study these problems both on a theoretical and practical level. In particular, we describe two original, coherent and efficient methods for footstep planning, one in the discrete case (chapter 5), and one in the continuous case (chapter 6). We validate these methods in simulation and with several experiments on the robot HRP-2.