Multi-Sensory UAV Navigation Research Testbed

这项提案寻求资金，以获得基础设施，以支持纽芬兰纪念大学智能系统实验室的多型号机器人导航系统设计研究。无人驾驶飞行器(UAV)在远程操作和自主环境下为几个工业、测量和研究部门提供服务。目前，这些应用程序采用GPS航点导航控制来服务于客户要求的检查任务。然而，对于许多工业应用，在不断变化的环境条件下，需要在间歇和有限的GPS条件下进行精确的自主操作，例如，在具有零星GPS接收区的现场的结构检查任务、近海结构检查、船舶的干船坞检查、作战规模的无人机交付，以及支持恶劣环境导航的环境监视。这些应用需要一种多感官、多模式的方法和稳健的冗余措施，以实现平台的安全关键自治。该研究计划既有基础知识的开发，也有业务规模的验证，其中系统和传感器模型的档案以一致的方式使用，以可靠地观察机器人的导航状态，即空中机器人通过GPS有限区域、通过预先绘制的区域以及环境条件不断变化的区域，如阵风、磁扰和能见度下降的区域。该研究方法考察了概率估计器的设计，该设计采用了一系列可供其使用的模型，包括标称惯性导航系统(INS)机械化方程、平台的动态约束、学习的未建模动力学以及具有变化的随机噪声参数的模型。此外，还研究了视觉、惯性、超宽带、LiDAR和GPS等冗余传感器源的存在，以获得在扩展平台工作范围的情况下导航系统的稳健性能。这项NSERC RTI提案要求为“多传感无人机导航研究试验台”提供设备支持，该试验台包括一架配备定制视觉惯性激光雷达导航传感器包的六角无人机和一个RTK GPS系统。为了在运行过程中使用实际传感器数据反馈完成多模型机器人导航研究成果的现场实验验证，所提议的设备是必要的。该设备将使用最先进的软件和硬件工具，为HQP提供无人机导航技术操作规模验证方面的实践培训。这些能力可以高度转移到加拿大各地的技术和研究部门，并显著加强MUN ISL的行业合作研究伙伴关系。

This proposal seeks funding to acquire infrastructure to support multi-model robotic navigation system design research at the Intelligent Systems Lab of Memorial University of Newfoundland. Unmanned Aerial Vehicles (UAVs) are servicing several industrial, survey, and research sectors in both remotely operated and autonomous settings. Currently, these applications employ GPS waypoint navigation control to service inspection tasks required by clients. However, for many industrial applications, precise autonomous operation under intermittent and limited GPS conditions is desired in varying environmental conditions, e.g., structural inspection tasks at sites with areas of patchy GPS reception, offshore structures inspection, dry dock inspection of ships, operational scale drone delivery, and environmental surveillance to support harsh environment navigation. These applications demand a multi-sensory and multi-model approach with robust redundancy measures to achieve safety-critical autonomy of the platforms. The research program has both fundamental knowledge development and operational scale validation where an archive of system and sensor models is used in a consistent way to reliably observe the navigation states of robots, i.e., aerial robots transitioning through GPS limited areas, through pre-mapped regions, and areas with changing environmental conditions such as wind gust, magnetic disturbances, and degraded visibility. The research methodology examines probabilistic estimator design that employs a bank of models at its disposal that include nominal Inertial Navigation System (INS) mechanization equations, dynamic constraints of the platform, learned un-modeled dynamics, and models with varying stochastic noise parameters. Furthermore, the existence of redundant sensor sources ranging from visual, inertial, Ultra-Wide Band, LiDAR, and GPS are investigated to achieve robust navigational system performance for extending platforms' operational envelope. This NSERC RTI proposal requests equipment support for a “Multi-sensory UAV navigation research testbed” which comprises of a hexacopter drone equipped with a custom visual-inertial-Lidar navigation sensor package and a RTK GPS system. The proposed equipment is necessary to complete field experimental validation of multi-model robotic navigation research findings using actual sensor data feedback during operation. The equipment will provide HQP hands-on training on operational scale validation of UAV navigation technology using state of the art software and hardware tools. These capabilities are highly transferable to technology and research sectors across Canada and significantly enhance the industry collaborative research partnerships of MUN ISL.