Automation, Intelligence and Adaptation of Robot Training for Motor Learning and Neurorehabilitation
Project Start: 2018
Although rehabilitation robots have a huge potential to increase the efficiency of motor learning and neurorehabilitation, nowadays even the state-of-art robotic systems lack the intelligence to adapt their support to the specific trainees’ needs. This lack of adaptation results in sub-optimal training interventions that reduce the trainers’ effort and, therefore, training outcomes.
The goal of this project is to exploit machine learning algorithms and develop novel robot-training-controllers to add intelligence to robotic therapy and efficiently enhance neurorehabilitation outcomes. The first part of the project focuses on automatizing the assessment of subjects’ performance using the sensor data integrated in the robotic devices; more specifically, identifying which special features of the trained task/movement need special attention.
The second part focuses on the generation of small training sessions which aim to improve these detected features and, therefore, the overall task performance. The last step focuses on allowing robotic devices to self-modify their own control laws over time, based on the measured effects that the applied control laws had on the subjects’ rehabilitation outcomes. We believe that such an autonomous and optimal training routine would boost the effectiveness of robotic neurorehabilitation.
Keywords: Motor Learning, neurorehabilitation, robotics, haptics, motion control, machine learning