392135 Models and Theories of human and robot motor learning (S) (WiSe 2025/2026)

Humans exhibit extraordinary adaptivity, seamlessly learning to perform complex motor tasks across a wide range of environments and developmental stages. This level of flexibility and robustness remains a significant challenge in artificial systems, particularly in autonomous agents and robotic platforms.

In this seminar, we will explore foundational and contemporary neuroscience theories that explain how humans acquire and refine motor skills—from the spontaneous movements of infants to the sophisticated coordination seen in adults. We will delve into key concepts such as motor primitives, internal models, sensory feedback integration, and neuroplasticity, grounding our understanding in experimental findings and theoretical frameworks.

Alongside the biological perspective, we will examine state-of-the-art deep reinforcement learning (DRL) strategies used in the training of humanoid robots. Topics will include model-free vs. model-based learning, curriculum learning, imitation learning, and sim-to-real transfer. We will analyze how these methods attempt to replicate—or diverge from—human-like learning processes, and critically assess their strengths and limitations.

The seminar aims to foster an interdisciplinary dialogue between neuroscience and artificial intelligence, highlighting how insights from human motor learning can inspire more adaptive, resilient robotic systems. Participants will be encouraged to think critically about the parallels and gaps between biological and artificial learners and explore how integrating perspectives from both fields can drive progress in motor control research and robotics.