AstroSLAM - A Robust and Reliable Visual Localization and Pose Estimation Architecture for Space Robots in Orbit

空间机器人技术对于当前和未来的空间探索和利用任务至关重要。先进的空间机器人技术将使卫星在轨服务和加油成为可能，并将支持更复杂的任务，如在轨大型柔性结构组装、碎片清除、检查、硬件升级等。例如，目前的通信卫星的典型寿命为10至15年，在这一点上，一颗本来功能完美的卫星会耗尽推进剂并退役。给卫星加油可以增加几年的寿命和收入。未来设想的月球、火星和其他地方的任务也将需要轨道上先进的机器人能力。许多未来的探索任务设想由宇航员和自由飞行的“协作机器人”组成的综合团队来支持或监控人类船员的活动。目前这些任务缺少的是使用自主机载感知和规划能力提供完整的4D（时空）态势感知的能力。最近地面机器人的技术突破为机器人在轨操作的类似进步铺平了道路，使机器人在不久的将来能够在太空中进行常规操作。其中包括感知和规划算法、基于机器学习的模式识别、自主性、新的计算机硬件架构、人机界面和灵巧操作等。这项研究的结果将是宇航员和空间机器人协同工作的能力，以实现对驻留空间物体的检查、监测和分类；机动和接近操作和对接；发生故障或翻滚的航天器的救助和回收；以及轨道资产的服务、建造、维修、升级和加油任务。该项目将开发新颖的视觉感知、定位、测绘和规划算法，为空间机器人提供新的态势感知能力，使其能够单独工作或与轨道上的宇航员一起工作。该研究计划包括利用深度神经网络架构开发新的自动特征提取和匹配算法，以适应具有挑战性的成像条件（准直光、高对比度、缺乏大气散射、缺乏独特特征等）和空间轨道运动约束。这将使鲁棒和可靠的相对姿态估计，三维形状重建和表征空间物体。基于因子图优化框架的新型优化规划和预测方法将与这些新的感知能力相匹配，以考虑燃料使用和轨道运动约束。该理论的实验验证将在佐治亚理工学院机器人空间操作自主航天器测试平台上进行，这是一个最先进的航天器模拟平台。这项研究将涉及研究生和本科生。这项研究的结果将通过期刊和会议出版物、组织应邀讲习班和研讨会发言以及有针对性地向大众传播媒介传播（新闻稿、采访）的方式向社区传播。该项目由跨部门机器人基础研究项目支持，由工程（ENG）和计算机与信息科学与工程（CISE）联合管理和资助。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Space robotics is essential for all current and future space exploration and utilization missions. Advanced space-robotic technologies will enable in-orbit servicing and refueling of satellites, and will support more elaborate missions such as in-orbit large flexible structure assembly, debris removal, inspection, hardware upgrades, etc. For instance, current communication satellites have a typical lifetime of 10 to 15 years, at which point an otherwise perfectly functioning satellite runs out of propellant and is decommissioned. Refueling the satellite can add several years of additional lifetime and revenue. Future envisioned missions to the Moon, Mars and beyond will also require advanced robotic capabilities in orbit. Many future exploration missions envision integrated teams of astronauts and free flying “co-robots” that support or monitor the human crew activities. What is currently missing from these missions is the ability to provide full 4D (space-time) situational awareness using autonomous on-board perception and planning capabilities. Recent technological breakthroughs for ground robots pave the way for similar advancements in robotic in-orbit operations to enable routine robotic operations in space in the not-so-distant future. These include perception and planning algorithms, machine learning based pattern recognition, autonomy, new computer hardware architectures, human-machine interfaces, and dexterous manipulation, among many others. The outcome of the research will be the ability of astronauts and space robots to work together to enable inspection, monitoring, and classification of resident space objects; maneuvering and proximity operations and docking; salvage and retrieval of malfunctioning or tumbling spacecraft; and servicing, construction, repair, upgrade, and refueling missions of assets in orbit.This project will develop novel visual perception, localization, mapping, and planning algorithms that will enable new capabilities in terms of situational awareness for space robots that can work alone or alongside astronauts in orbit. The research plan includes the development of novel automated feature extraction and matching algorithms using deep neural network architectures, adapted to the challenging imaging conditions (collimated light, high contrast, lack of atmospheric scattering, paucity of distinctive features, etc.) and orbital motion constraints imposed in space. This will enable robust and reliable relative pose estimation, 3D shape reconstruction and characterization of space objects. Novel optimal planning and prediction methods based on a factor-graph optimization framework will be matched to these new perception capabilities to account for fuel usage and orbital motion constraints. The experimental validation of the theory will take place at the Georgia Tech Autonomous Spacecraft Testing of Robotic Operations in Space platform, a state-of-the-art spacecraft simulation platform. The research will involve both graduate and undergraduate students. The results of this research will be disseminated to the community by journal and conference publications, organization of invited workshops and seminar presentations, and by targeted exposure (press releases, interviews) to popular media.This project is supported by the cross-directorate Foundational Research in Robotics program, jointly managed and funded by the Directorates for Engineering (ENG) and Computer and Information Science and Engineering (CISE).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.