The goal of this project is to improve robotic neurorehabilitation by developing robotic artificial intelligence, which can automatically adapt the robotic training strategy to the trainee’s special needs and the trained motor task.
We had previously developed a virtual training environment that consisted a motor task where the participants inverted a pendulum and tried to keep it inverted. We conducted a two-day motor learning experiment on this training environment to see how the arm weight support affects the movement and the motor learning of the participants, especially when the environment is haptically rendered.
During this experiment, we also collected performance data on a second day when participants performed the task with randomized arm weight support levels (between 0, i.e., no support, and 100% support). These data will be employed in the project’s second step, where we will use machine learning methods to build and test a model that selects the optimum arm weight support level based on the participants' ongoing performance. The model's effectiveness will be evaluated with new participants and compared to the results obtained from the first experiment (i.e., the participants who had no arm support or 100% fixed support during training).
The application of machine learning is a promising approach to enhance motor learning and neurorehabilitation. Data science in robotic neurorehabilitation has the potential to accelerate the motor recovery of neurologically injured patients by means of personalized training.