EAGER: GOALI: Distributed Embedded Vision System for Multi-Unmanned Ground Vehicle Coordination in Indoor Environments

EAGER：GOALI：用于室内环境中多无人地面车辆协调的分布式嵌入式视觉系统

• 批准号: 1547934
• 负责人: Christophe Bobda
• 金额: $ 27.36万
• 依托单位: University of Arkansas
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1547934&HistoricalAwards=false
• 关键词: EAGER; GOALI; Distributed; Embedded; Vision

This collaborative project between academics and industry provides a configurable and portable vision-based infrastructure for decentralized coordination of industrial trucks, also known as Automated Guided Vehicles (AGVs) in indoor environments. AGVs have the potential to revolutionize operations in areas such as manufacturing and distribution, health care, and military by efficiently accomplishing the mundane and often repetitive task of transporting materials in distribution and manufacturing. The research addresses limitations of recent successful introduction of commercial robot systems in distribution centers and manufacturing to transport items, but typically only to specific and restrictive environments. The research extends these technologies to environments that are quickly reconfigured by moving stations and objects around. Generic models, tools, and technologies are developed to actively capture the world with semantically labeled objects, actions and events, and to generate goals, priorities, and plans. The solutions proposed in this research address the requirements of decentralized coordination, real-time environmental changes. The approach uses a set of distributed ceiling-mounted smart cameras with overlapping field-of-view for global view and coordination at the facility level, and cameras mounted on AGVs for short-range truck navigation. Multi-truck coordination is then framed as the problem of routing packets in a dynamic and hierarchical network where cameras represent routers and trucks represent packets. To address the complexity of image processing tasks, the hardware/software implementation utilizes an FPGA-based target platform from a previous NSF-Funded project, where low-level inherent parallel image processing tasks are mapped onto hardware while high-level reasoning is kept in software. This project will provide an interface formalism to specify component integration and system composition along with methods to optimize run-time and resource usage. This research is expected to produce an infrastructure and methods for decentralized vision-based coordination of autonomous vehicles activities in a dynamically changing indoor environment. The developed infrastructure will enable manufacturing and distribution companies, including lean and agile entities, to optimize indoor transportation activities in existing arrangement, without modification of available infrastructure, and reduce labor and operating costs by redeploying employees to value-added roles. In addition, AGVs have the potential to enable autonomous mobile robot applications in numerous other unstructured environments, including hospitals, malls, retail stores, critical infrastructure, airports, schools, and sports venues. The project is conducted as a joint effort between the University of Arkansas in Fayetteville and R-Dex in Atlanta and will provide undergraduate and graduate students opportunities to perform their work in academic and industrial environments. The involvement of under-represented groups will be increased with the support of the University of Arkansas Engineering Career Awareness Program, and the University of Arkansas chapters of the National Society of Black Engineer (NSBE) and the Society of Hispanic Professional Engineers (SHPE).

这个学术界和工业界之间的合作项目为工业卡车(也称为室内环境中的自动导引车(AGV))的分散协调提供了一个可配置的、基于视觉的便携式基础设施。AGV通过高效地完成配送和制造中平凡且经常重复的材料运输任务，有可能给制造和分销、医疗保健和军事等领域的运营带来革命性的变化。这项研究解决了最近在配送中心和制造业成功引入商用机器人系统以运输物品的局限性，但通常仅限于特定和受限的环境。这项研究将这些技术扩展到通过移动站点和物体来快速重新配置的环境。开发通用模型、工具和技术是为了主动捕获具有语义标签的对象、动作和事件的世界，并生成目标、优先事项和计划。本研究提出的解决方案解决了分散协调、实时环境变化的要求。该方法使用一套安装在天花板上的分布式智能摄像机和安装在AGV上的摄像机进行短程卡车导航，这些摄像机具有重叠的视场，用于设施层面的全局查看和协调。然后，多卡车协调被框架化为在动态和分层网络中的数据包路由问题，其中摄像头代表路由器，卡车代表数据包。为了解决图像处理任务的复杂性，硬件/软件实现利用了来自先前NSF资助项目的基于FPGA的目标平台，其中低级固有的并行图像处理任务被映射到硬件上，而高级推理被保持在软件中。该项目将提供一个接口形式化来指定组件集成和系统组合，以及优化运行时和资源使用的方法。这项研究有望为自动驾驶车辆在动态变化的室内环境中基于视觉的分散协调活动提供基础设施和方法。开发的基础设施将使制造和分销公司，包括精益和敏捷的实体，在不修改现有基础设施的情况下，在现有安排下优化室内运输活动，并通过将员工重新分配到增值角色来降低劳动力和运营成本。此外，AGV有可能在许多其他非结构化环境中实现自主移动机器人应用，包括医院、商场、零售店、关键基础设施、机场、学校和体育场馆。该项目是费耶特维尔的阿肯色大学和亚特兰大的R-Dex联合开展的，将为本科生和研究生提供在学术和工业环境中开展工作的机会。在阿肯色大学工程职业意识计划、全国黑人工程师协会(NSBE)和西班牙裔专业工程师协会(SHPE)阿肯色大学分会的支持下，代表不足的群体的参与将会增加。