CAREER: Robust and Collaborative Perception and Navigation for Construction Robots

职业：建筑机器人的稳健协作感知和导航

• 批准号: 2238968
• 负责人: Chen Feng
• 金额: $ 60万
• 依托单位: New York University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-01 至 2028-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2238968&HistoricalAwards=false
• 关键词: CAREER; Robust; Collaborative; Perception; Navigation

The construction industry is responsible for maintaining aging civil infrastructure and build new facilities that can accommodate the social needs of the 21st century. This is in addition to the ongoing critical need to address long-standing problems in occupational safety, labor productivity, costs, and labor shortage. A promising technical solution is to introduce mobile robots on construction jobsites. It is possible to leverage recent discoveries in robotics and artificial intelligence (AI) to tackle those aforementioned challenges. However, unlike manufacturing automation or self-driving cars, construction robots face unique challenges due to the need to navigate dynamic environments. Such robots are also required to work closely with humans in a variaty of tasks and often handle heavy payloads. This award supports fundamental robotics research to allow better perception and navigation for construction jobsite monitoring robots. It will produce an intelligent mobile robot team equipped with cameras to autonomously monitor construction progress and operations to improve jobsite efficiency and safety. The results of this research will be widely applicable to scenarios beyond construction, ranging from connected and autonomous vehicles to service robotics in smart and accessible cities. The project will facilitate collaboration between robotics, artificial intelligence, and civil and mechanical engineering. Furthermore, it aims to broaden participation of underrepresented groups in engineering via educational games, multi-disciplinary robotics curriculum, and workforce training workshops.Mobile robotics in construction jobsites are often limited by perception challenges due to occlusion and limited field of view. In dynamic jobsites, limited perception leads to navigation and inefficient assistance. To improve the robustness, reliability, and scalability of the vision system in mobile robots, novel self-supervised and graph-based representation learning will be used to extract, organize, and reason about places and objects from high-dimensional sensory inputs. This research will advance the state of the art along three directions: (1) robust navigation from topological representations for monitoring in dynamic and ever-changing jobsites, (2) collaborative perception for providing safer operation monitoring and collision warnings on busy jobsites, and (3) integrated perception and navigation at both the algorithm, system, and dataset levels. The research will be validated in real construction jobsites through industry partners, and the resulting software, hardware design, and dataset will be open source to stimulate future research.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

建筑业负责维护老化的民用基础设施并建造能够满足21世纪社会需求的新设施。除此之外，还迫切需要解决职业安全、劳动生产率、成本和劳动力短缺等长期存在的问题。一个有前景的技术解决方案是在建筑工地引入移动机器人。可以利用机器人和人工智能 (AI) 领域的最新发现来应对上述挑战。然而，与制造自动化或自动驾驶汽车不同，建筑机器人由于需要在动态环境中导航而面临独特的挑战。此类机器人还需要在各种任务中与人类密切合作，并且经常处理重型有效载荷。该奖项支持基础机器人研究，以便为建筑工地监控机器人提供更好的感知和导航。它将生产一支配备摄像头的智能移动机器人团队，自动监控施工进度和操作，以提高工地效率和安全性。这项研究的结果将广泛适用于建筑以外的场景，从互联和自动驾驶车辆到智能和无障碍城市中的服务机器人。该项目将促进机器人、人工智能以及土木和机械工程之间的合作。此外，它的目标是通过教育游戏、多学科机器人课程和劳动力培训研讨会来扩大代表性不足的群体对工程的参与。建筑工地的移动机器人通常因遮挡和视野有限而受到感知挑战的限制。在动态的工作现场，有限的感知会导致导航和低效的协助。为了提高移动机器人视觉系统的鲁棒性、可靠性和可扩展性，新颖的自监督和基于图形的表示学习将用于从高维感官输入中提取、组织和推理地点和物体。这项研究将沿着三个方向推进最先进的技术：（1）从拓扑表示中进行稳健的导航，以在动态和不断变化的工作现场进行监控，（2）协作感知，以在繁忙的工作现场提供更安全的操作监控和碰撞警告，以及（ 3）在算法、系统和数据集层面集成感知和导航。该研究将通过行业合作伙伴在真实的建筑工地进行验证，最终的软件、硬件设计和数据集将开源以刺激未来的研究。该奖项反映了 NSF 的法定使命，并通过使用基金会的评估进行评估，被认为值得支持。智力价值和更广泛的影响审查标准。

项目成果

DeepMapping2: Self-Supervised Large-Scale LiDAR Map Optimization

• DOI: 10.1109/cvpr52729.2023.00898
• 发表时间: 2022-12
• 期刊: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
• 作者: Chao Chen;Xinhao Liu;Yiming Li;Li Ding;Chen Feng-