Robotic assisted gait training (RAGT) for stroke rehabilitation is promising but it has not met the expectations of the clinical community yet. Several studies have demonstrated the effectiveness of this approach, but, as now, RAGT has not been demonstrated to be superior to standard gait training. The research community is thus focusing on developing novel training methodologies for RAGT in order to fully exploit its potential. Emphasis has been given to training methodologies that promote and maximize the active participation of the patients. The Error Augmentation (EA) paradigm has been successfully developed and proved effective for the upper limbs. In the EA paradigm, perturbations are used to amplify the movement error of the patient in a way that induces a compensatory motor plan that rejects the augmented error. This approach has been applied to the lower limbs by means of split-belt treadmill training, where the speeds of the belt are set in order to amplify gait asymmetries and promote an adaptation toward a less asymmetric gait pattern. However, EA has not been applied in RAGT. Its implementation in RAGT can extend the target of this therapy to several other gait parameters. In WALKEr we plan to develop the EA paradigm on both a robotic ankle device and a six degrees of freedom (DoFs) exoskeleton for gait rehabilitation. Both developments will then be tested on healthy subjects and chronic stroke survivors to assess the potential of this technique in improving motor adaptations, motor learning and rehabilitation.