Non Prehensile Planar Manipulation

Non-prehensile planar manipulation focuses on moving objects on a flat surface without grasping them, typically using pushing or poking actions. Current research heavily utilizes reinforcement learning, often incorporating model predictive control or learning from demonstrations to overcome the challenges of hybrid dynamics and contact uncertainty inherent in these tasks. These methods are being applied to improve the efficiency and robustness of robot manipulation, particularly in scenarios involving diverse object shapes and cluttered environments. The resulting advancements have implications for automating tasks in manufacturing, warehousing, and other domains requiring dexterous but non-grasping interactions.