Rehabilitation robots support rehabilitation of patients with neurological movement disorders. Newly developed patient-cooperative control approaches aim at enabling the robots to perform more effective trainings, which are tailored to individual patients. This paper presents two approaches to automatically adapt robotic support: The iterative learning feedforward control is able to support movements with defined timing. The iterative learning conservative force field can also support movements with free timing. Both approaches are demonstrated in an example application with the gait rehabilitation robot Lokomat.