ARCOL

Throughout our lifetime, we are acquiring and improving a wide range of motor skills, which enable us to perform certain movements faster, better and more accurately. The mechanisms at play in motor skill acquisition are complex and not yet fully understood but there is a consensus that skill acquisition requires practice and that motor variability has an operational role. However, practice design is heavily based on heuristics and it is not yet clear how to design good learning schedules. In the ARCOL project we envision computational solutions able to learn novice individual-specific practice design in order to support the acquisition of new motor skills. To reach this objective, the project focuses on two use cases involving complex movements. The first use case is motor learning in piano performance by novices. The second use case is motor sequence learning in prosthesis control. At the core of the project lays the use of machine learning as a means to understand and facilitate motor learning. The resulting dual process through which both human and machine learn reciprocally is what we call co-learning. The ARCOL project ambitions two main research objectives. First, we aim to build algorithms reinforcing co-learning. These algorithms will generate a practice schedule to be given to a human learner and assess its impact on the improvement of skill performance. This task is known as curriculum learning in machine learning and developmental robotics. However reinforcement-driven curriculum learning for motor skill acquisition poses new challenges that will be investigated in the project. Second, we aim to identify the benefits and limitations of human-machine co-learning in motor skill facilitation with non-experts. We focus on non-experts to better assess how new control policies are acquired. Important challenges stem from the complexity of the motor tasks considered and the resulting difficulty to assess learning progress. These challenges will be addressed in the project. Our long-term goal is thus to advance knowledge in skill acquisition mechanisms as well as producing operational systems able to facilitate motor skill learning in real-world situations.