Robotic rehabilitation therapies have been improving by providing the needed assistance to the patient, in a human-centered environment, and also helping the therapist to choose the necessary procedure. This thesis presents an adaptive "Assistance-as-needed" strategy which adheres to the specific needs of the patient and with the inputs from the therapist, whenever needed. The exertion of assistive and responsive behavior of the lower limb wearable robot is dedicated for the rehabilitation of incomplete spinal cord injury (SCI) patients. The main objective is to propose and evaluate an adaptive control model on a wearable robot, assisting the user and adhering to their needs, with no or less combination of external devices. The adaptation must be more interactive to understand the user needs and their volitional orders. Similarly, by using the existing muscular strength, in incomplete SCI patients, as a motivation to pursue the movement and assist them, only when needed. The adaptive behavior of the wearable robot is proposed by monitoring the interaction and movement of the user. This adaptation is achieved by modulating the stiffness of the exoskeleton in function of joint parameters, such as positions and interaction torques. These joint parameters are measured from the user independently and then used to update the new stiffness value. The adaptive algorithm performs with no need of external sensors, making it simple in terms of usage. In terms of rehabilitation, it is also desirable to be compatible with combination of assistive devices such as muscle stimulation, neural activity (BMI) and body balance (Wii), to deliver a user friendly and effective therapy. Combination of two control approaches has been employed, to improve the efficiency of the adaptive control model and was evaluated using a wearable lower limb exoskeleton device, H1. The control approaches, Hierarchical and Task based approach have been used to assist the patient as needed and simultaneously motivate the patient to pursue the therapy. Hierarchical approach facilitates combination of multiple devices to deliver an effective therapy by categorizing the control architecture in two layers, Low level and High level control. Task-based approaches engage in each task individually and allow the possibility to combine them at any point of time. It is also necessary to provide an interaction based approach to ensure the complete involvement of the user and for an effective therapy. By means of this dissertation, a task based adaptive control is proposed, in function of human-orthosis interaction, which is applied on a hierarchical control scheme. This control scheme is employed in a wearable robot, with the intention to be applied or accommodated to different pathologies, with its adaptive capabilities. The adaptive control model for gait assistance provides a comprehensive solution through a single implementation: Adaptation inside a gait cycle, continuous support through gait training and in real time. The performance of this control model has been evaluated with healthy subjects, as a preliminary study, and with paraplegic patients. Results of the healthy subjects showed a significant change in the pattern of the interaction torques, elucidating a change in the effort and adaptation to the user movement. In case of patients, the adaptation showed a significant improvement in the joint performance (flexion/extension range) and change in interaction torques. The change in interaction torques (positive to negative) reflects the active participation of the patient, which also explained the adaptive performance. The patients also reported that the movement of the exoskeleton is flexible and the walking patterns were similar to their own distinct patterns. The presented work is performed as part of the project HYPER, funded by Ministerio de Ciencia y Innovación, Spain. (CSD2009 - 00067 CONSOLIDER INGENIO.