Pre Grasp

Pre-grasping in robotics focuses on the crucial steps a robot takes before actually grasping an object, encompassing actions like repositioning or reorienting the object to achieve a stable grasp. Current research emphasizes developing robust and adaptable pre-grasping strategies using reinforcement learning, diffusion models, and vision-language models to handle diverse objects and environments, often incorporating extrinsic dexterity (using surfaces or tools). These advancements aim to improve robotic manipulation capabilities, particularly for complex tasks involving ungraspable or oddly-shaped objects, leading to more efficient and versatile robotic systems in various applications.