SLAM: Suction, Localisation And Manipulation

SLAM：吸力、定位和操纵

• 批准号: 10023566
• 金额: $ 22.32万
• 依托单位: RECYCLEYE LTD
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10023566
• 关键词: SLAM; Suction; Localisation; Manipulation

Recycleye has already developed a low-cost, AI-powered system replicating the power of human vision. It uses advanced machine learning algorithms to provide automatic, image-based detection of individual items in co-mingled waste streams, at a material and object level. It leverages a cutting-edge synthetic data generation pipeline, and Recycleye's own WasteNet - the world's largest visual database of labelled waste items.In a previously funded R&D project, Recycleye augmented the vision system with a robotic arm to deliver a robotic sorting solution. Now, Recycleye will develop and test a novel robotic grasping technology, that will drastically improve picking accuracy (and therefore the efficiency) of Recycleye's current commercial offerings, enabling the application of Recycleye's technology to waste streams beyond Municipal Solid Waste (MSW), whilst providing significant technical development for the wider industrial automation industry,The method currently used in automation industry when picking objects, is to aim for the centre of the detected object when grasping. This is optimal if the item is flat, however in numerous industrial applications items are crushed and deformed leading to a sub-optimal or failed item pick. By applying the latest computer vision algorithms and techniques to the grasping problem, Recycleye will develop the first optimal grasp prediction technology that can be deployed real-time to maintain accuracy in complex fast-paced industrial applications. This technology has a huge potential for boosting Recycleye's system capabilities within the waste industry, while providing a step forward in grasping technology relevant to a range of industries.

Sparkleye已经开发了一种低成本的人工智能系统，复制了人类视觉的力量。它使用先进的机器学习算法，在材料和物体层面上对混合废物流中的单个物品进行自动的、基于图像的检测。它利用了尖端的合成数据生成管道和Biscleye自己的Biscleye Net-世界上最大的标签废物可视化数据库。在之前资助的研发项目中，Biscleye用机械臂增强了视觉系统，以提供机器人分拣解决方案。现在，Sparkleye将开发和测试一种新的机器人抓取技术，这将大大提高采摘的准确性（因此效率）的提升，使其技术能够应用于城市固体废物（MSW）以外的废物流，同时为更广泛的工业自动化行业提供重大的技术发展。自动化行业目前使用的方法，当拾取物体时，是在抓取时瞄准被检测物体的中心。如果物品是平的，这是最佳的，然而在许多工业应用中，物品被压碎和变形，导致次优或失败的物品拾取。通过将最新的计算机视觉算法和技术应用到抓取问题中，Sparkleye将开发第一个最佳抓取预测技术，该技术可以实时部署，以在复杂的快节奏工业应用中保持准确性。这项技术具有巨大的潜力，可以提高废弃物处理行业的系统能力，同时在掌握与一系列行业相关的技术方面向前迈出了一步。