机器人抓取检测技术的研究现状

作为机器人在工厂、家居等环境中最常用的基础动作,机器人自主抓取有着广泛的应用前景,近十年来研究人员对其给予了较高的关注,然而,在非结构环境下任意物体任意姿态的准确抓取仍然是一项具有挑战性和复杂性的研究.机器人抓取涉及3个主要方面:检测、规划和控制.作为第1步,检测物体并生成抓取位姿是成功抓取的前提,有助于后续抓取路径的规划和整个抓取动作的实现.鉴于此,以检测为主进行文献综述,从分析法和经验法两大方面介绍抓取检测技术,从是否具有抓取物体先验知识的角度出发,将经验法分成已知物体和未知物体的抓取,并详细描述未知物体抓取中每种分类所包含的典型抓取检测方法及其相关特点.最后展望机器人抓取检测技术的发展方向,为相关研究提供一定的参考.

Recent researches on robot autonomous grasp technology

As the most common and basic robotic movement in environment such as factory and household, the autonomous grasping has various application prospect. Although researchers have paid higher attention to the robot autonomous grasping in recent decades, grasping arbitrary objects accurately with arbitrary poses in non-structural environment is still a challenging and complex study. The robot grasping involves three main aspects: sensing, planning and controlling. The first step is detecting the targeted object and generating the grasping pose. It is the precondition of the successful grasp that helps achieve the grasp path planning and the entire grasp movement. The literature review in this study is mainly based on sensing. Firstly, the grasping technology is introduced from two aspects: analytical method and empirical method, and the empirical method is divided into the grasping of known objects and unknown objects from the perspective of whether there is prior knowledge of grasping objects. Moreover, the typical grasping methods and their related features contained in each category of unknown object grasping is detailly described. Finally, the development direction of robotic grasp detection technology is prospected with the aim of providing references for the relevant researchers.