Modern rehabilitation robots should operate close to a human and get in direct physical contact, which requires appropriate controllers. This thesis describes different interaction control concepts for an exoskeleton robot with compliant actuators on the basis of pneumatic direct drives with Rotary Elastic Chambers (REC). It is investigated which kind of interaction or effect can be achieved, if the complexity of plant models and control structures is kept as low as possible. The first part treats modeling of the REC actuators by using experimental & phenomenological approaches. The actuator model is divided into pneumatic and mechanic subsystems and static torque characteristics are combined with different friction compensation models. A solution to independently adjust the output torque and joint rigidity is implemented. The second part focuses on interaction controllers that manage without force sensors. For this purpose, a planar 3 degrees of freedom exoskeleton robot equipped with REC actuators is used, which can partly support the lower extremities of a person. As application example an activity of daily living is chosen, in which elderly people or slightly physically restricted stroke patients should perform repetitive sit-to-stand (StS) exercises. The goal is to combine REC actuators and supportive controllers to facilitate the StS process, such that the user is not passively guided along a fixed trajectory, but is mostly self-determined with own authority. The presented concepts begin from a master-slave approach with individual trajectory generation, to the usage of active virtual joint admittances through the realization of variable passive joint rigidity. An adaptive iterative learning concept eventually allows temporal and spatial freedom of movement.weiterlesen