Collective Perception in Swarm Robotics

群体机器人中的集体感知

• 批准号: RTI-2023-00281
• 负责人: Beltrame, Giovanni
• 金额: $ 10.37万
• 依托单位: École Polytechnique de Montréal
• 依托单位国家: 加拿大
• 项目类别: Research Tools and Instruments
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=754185
• 关键词: Collective; Perception; Swarm; Robotics

Automation has been, and still is, the driving force behind the transformation of our economies. However, to this day robots are still mostly relegated to tightly controlled industrial assembly lines, where all their motions are carefully planned and tuned for very specific and limited tasks. Despite the best efforts of scientists and engineers, robots are not yet ready to leave their supervised settings to wander the wild and the unknown. We have been carrying a research program to make swarm robotics technology into practical, every day applications. In the past 6 years, we have produced contributions to connectivity maintenance, long-term autonomy, pattern formation, collaborative localization with minimal infrastructure, multi-robot mapping, object tracking, human-swarm interfaces, space exploration and collaborative transport. These contributions were made possible using robots obtained via previous RTI competitions, that are now in dire need of upgrade and replacement to continue the program. Our current focus is collaborative perception: the need for reliable, resilient and safe perception is a core challenge of autonomous robotics. One of the core perception algorithm in robotics is Simultaneous Localization And Mapping, or SLAM. Situational awareness, as provided by SLAM, is essential to most robotic systems. While it can be done on each robot individually, it is often beneficial to perform it collaboratively when working with groups of robots. In Collaborative SLAM, or C-SLAM, each individual robot performs single-robot SLAM, then communicates and performs data association with the other robots in order to build a common global map between them. This way, they can share a common reference frame for localization and a mutual understanding of the environment for decision-making. Our general objective is to reduce the rate of failures of multi-robot localization and mapping algorithms deployed in unknown and uncontrolled environments. We request a set of 20 AgileX Limo: a small but capable robot featuring a powerful onboard computer, a 3D camera, an inertial unit, and a lidar. Having at least 20 robots would allow us to experiment hierarchical swarms, collaborative transport, energy sufficiency, and collaborative SLAM. We also request 2 Bunker Mini robots to support our research on multi-robot exploration and mapping of GPS-denied environments. Our research output will be tested in collaboration with the European Space Agency (ESA), the Canadian Space Agency (CSA), and NASA JPL. Over the next five years, we expect 6 Postdoctoral fellows, 10 PhDs, 8 Master's and 15 undergraduates from the applicant's laboratory to be trained using this equipment.

自动化一直是，并且仍然是我们经济转型背后的驱动力。然而，时至今日，机器人仍然大多被归入严格控制的工业装配线，它们所有的动作都是经过精心规划和调整的，用于非常具体和有限的任务。尽管科学家和工程师尽了最大的努力，但机器人还没有准备好离开他们的监管环境去野外和未知的地方游荡。我们一直在进行一项研究项目，将群体机器人技术应用到日常生活中。在过去的6年里，我们在连接维护、长期自治、模式形成、基于最小基础设施的协同定位、多机器人映射、目标跟踪、人群界面、空间探索和协同运输方面做出了贡献。这些贡献是通过以前的RTI竞赛获得的机器人实现的，现在迫切需要升级和替换以继续该计划。我们目前的重点是协作感知：对可靠、有弹性和安全感知的需求是自主机器人的核心挑战。同时定位与映射（Simultaneous Localization And Mapping，简称SLAM）是机器人领域的核心感知算法之一。SLAM提供的态势感知对大多数机器人系统至关重要。虽然可以在每个机器人上单独完成，但在与机器人组一起工作时，协作执行通常是有益的。在协同SLAM （C-SLAM）中，每个单独的机器人执行单机器人SLAM，然后与其他机器人进行通信和数据关联，以便在它们之间建立一个共同的全局地图。通过这种方式，他们可以共享一个共同的本地化参考框架，并对决策环境有一个共同的理解。我们的总体目标是降低部署在未知和不受控制环境中的多机器人定位和映射算法的失败率。我们需要一套20 AgileX Limo：一个小型但功能强大的机器人，具有强大的车载计算机，3D摄像机，惯性单元和激光雷达。拥有至少20个机器人将使我们能够实验分层群、协作运输、能源充足和协作SLAM。我们还需要2个Bunker迷你机器人来支持我们在gps拒绝环境下的多机器人探索和测绘研究。我们的研究成果将与欧洲航天局（ESA）、加拿大航天局（CSA）和美国宇航局喷气推进实验室合作进行测试。在未来五年内，我们预计将有6名博士后、10名博士、8名硕士和15名本科生从申请人的实验室使用该设备进行培训。