DC6: Biomimetic model-based reinforcement learning of motor primitives
Project: Biomimetic model-based reinforcement learning of motor primitives (WP2)
Host institution: LUH (Germany)
Supervisor: Prof. T.Seel (LUH, Germany)
Co-supervisor(s): Prof. S. Ferrante (POLIMI), Dr. D. Laidig (Sensorstim)
Objectives:
To design and implement autonomous learning control schemes for assistive health systems by creating iterative learning strategies that enable reinforcement learning on small amounts of experimentally gathered data, to rapidly achieve accurate motor primitives.
Expected Results:
Enable assistive health systems to quickly learn to generate an accurate range of fundamental motor patterns, that generalise well to unseen patients and other assistive health systems. The methods will adjust to changes of the dynamics, utilizing continuous learning
Planned secondment(s):
- Sensorstim (3 months, M25-M27): Development of model-based reinforcement algorithms, with T. Schauer ( (KPI: joint conference paper)
- POLIMI (3 months, M33-M35): Validation of biomimetic motor primitives in a rehabilitation application field, with S. Ferrante (KPI: joint journal paper)
Enrolment in Doctoral degree: Doctoral School of Leibniz University Hannover (Germany)
Required profile: Background in control systems and machine learning, experience with Matlab or Python or both
Desirable skills/interests: Prior knowledge on real-time/embedded systems and biomechanics, interest in biomedical engineering and neuroprosthetics