Many motor disorders, including stroke, spinal cord injuries or multiple sclerosis, can lead to chronic disabilities limiting the autonomy and quality of life of hundreds of thousands of people in Europe. Motivation, social interaction and exercise intensity are key factors for improving recovery in post-stroke patients. Therefore, aim of this project is to advance in physical human-robot interaction (HRI) techniques with high level of safety and dependability using machine learning techniques, as well as take advantage of different neurological and biomechanical signals measured during functional movements for quantitative evaluation of patient’s recovery during and after physical training assisted by robotic devices.