Mohamed bin Zayed University of Artificial Intelligence
This course explores how robot hands manipulate, sense, and learn from in-hand interactions via model-based control, tactile perception, and data-driven learning. Students program goal-directed, highly dynamic dexterous skills by integrating motion/force control, reinforcement learning, and sim-to-real deployment.
Students implement and test model-based and learning-based in-hand manipulation on a physics simulator and a dexterous hand platform using tactile and visual feedback. Activities include building goal-conditioned controllers and RL policies, sim-to-real transfer, benchmarking on manipulation tasks, and analyzing performance, safety, and robustness.