SGER: Persistent, Autonomous Localization and Mapping of Unstructured Environments

点击翻译按钮获取中文摘要

The proposed research will investigate new representation, state estimation, and data association techniques to enable sustained, full 6-DOF motion simultaneous localization and mapping (SLAM) in diverse, large-scale environments. The driving vision is to realize an autonomous robot (and sensor suite) capable of sustained exploration of an arbitrarily large environment mixing indoor and outdoor elements, and human-built and natural scene structure. The work combines (1) a basic science component that will investigate fundamental issues of data representation, generality of environments, and algorithmic scaling behavior; (2) a development component with long-duration excursions with real sensors in a variety of environments using both robotic and human platforms; and (3) a validation and assessment component that will formulate and apply objective assessment metrics. The intellectual merits of the proposed research stem from the scientific challenges of developing persistent, autonomous, large-scale mobile robot navigation and mapping systems. Autonomous navigation is a compelling research goal because it embodies key issues that underlie many difficult problems, including choosing a representation, modeling sensor physics, managing uncertainty, detecting features, and selecting actions in real-time. The broader impacts of achieving a sustained, broad-area SLAM method will form the basis of a number of fundamentally useful capabilities, both from technical and societal standpoints. An enormous number of civil, commercial and military applications use maps, and gain greater utility from maps that are accurate and current with respect to the actual state of the world. Among these applications are vehicle and pedestrian navigation aids, accessibility studies for the disabled, urban planning, rapid emergency response, search and rescue operations, equitable taxation and land-use reform, zoning compliance, geographical information systems, and military planning, training and rehearsal. Autonomous operations, including navigation, mapping, and path planning, are vitally important to mobile robots operating on land, in air, space, underground, and underwater environments. Beyond SLAM, the proposed research can have broad impact on the large class of problems facing the challenge of performing real-time state estimation with highly uncertain data