基于融合感知的机器人实时定位与3D环境建模理论及轻型化装置研制

For realizing long-term autonomous of robots, this proposal investigates some key theorical problems of robot real-time 6D.O.F localization and online 3D mapping in GPS-denied environments by integrating a 3D lidar, a wide-angle monocular vision and a low-cost IMU as the perception system. On the basis of theoretical results, a lightweight prototype device will be carefully designed. And the device will be ultimately utilized for robot autonomous navigation experiments. The main feature of this proposal is twofold: online real-time calculation and multi-sensors fusion environmental perception. In the research, we will focus on some important aspects for practical applications. The first is the adaptability both for indoor and outdoor environments. The second is real-time calculation which can deal with faster 6D.O.F robot motions. And the last is the prototype device should be light enough for easy carrying by robots. The key scientific issues and techniques of the research include four parts: 1) real-time 6D.O.F odometry calculation, 2) online local environment 3D dense mapping, 3) mathematical description and robust solving for 6D.O.F simultaneous localization and mapping (SLAM) problem; and 4) development of lightweight prototype device. The scientific achievement of this research is to establishing a real-time robot 6D.O.F localization and 3D mapping theoretical framework, which will provide a technical basis for further researching on robots long-term autonomous in GPS-denied environments. Meanwhile, the results of the project will also provide technical supports for many other real-time online 3D mapping applications, such as online 3D reconstruction of the buildings, military reconnaissance, undermine online 3D mapping, underwater online 3D mapping, disaster site online 3D mapping, thus the technique has broad application prospects.

面向机器人长期自主的实际需求，申报项目以“3D激光、广角单目视觉、IMU融合感知”和“实时、在线计算”为研究特色来深入探讨具有室内外环境适应性、实时性、可适用于快速运动机器人载体应用的实时6D.O.F定位与在线3D环境建模理论与方法，并通过软硬件优化设计来研制轻型化样机装置，最终期望在自主机器人导航上实现应用。研究中重点解决紧耦合融合实时6D.O.F里程计算、局部致密环境建模、6D.O.F SLAM（同时定位与地图构建）问题的数学描述与鲁棒求解、轻型化样机研制四个方面的科学问题与关键技术，深入探讨不依赖于GPS的机器人实时定位与3D环境建模的新理论、新方法，为进一步的机器人长期自主性深入研究奠定基础，因而具有重要的学术价值。同时，项目成果也将为其它实时3D环境建模的应用开发，如建筑物3D重构、军事侦察，以及矿下、水下、灾难现场的在线3D环境建模提供理论技术支持，因而具有广阔的应用前景。

面向机器人长期自主的实际需求，项目以“3D激光、视觉、IMU融合感知”和“实时、在线计算”为研究特色来深入探讨具有室内外环境适应性、实时性、可适用于快速运动机器人载体应用的实时6D.O.F定位与在线3D环境建模理论与方法，通过软硬件优化设计研制了轻型化样机装置，并在自主机器人导航上实现了集成与应用。研究中重点解决紧耦合融合实时6D.O.F里程计算、局部致密环境建模、6D.O.F SLAM（同时定位与地图构建）问题的数学描述与鲁棒求解、轻型化样机研制四个方面的科学问题与关键技术，深入探讨不依赖于GPS的机器人实时定位与3D环境建模的新理论、新方法，为进一步的机器人长期自主性深入研究奠定基础，因而具有重要的学术价值。同时，项目成果也将为其它实时3D环境建模的应用开发，如无人机导航、建筑物3D重构、无人驾驶车建模与定位、以及矿下、水下、灾难现场的在线3D环境建模提供理论技术支持，因而具有广阔的应用前景。