分布式微小型火星漫游器导航方法研究

During the past decades, NASA has achieved tremendous success in Mars exploration missions, among which worthy of mention are “curiosity” and “opportunity” rovers that are still operating today. However, NASA has not alter the situation of highly limited exploration range of rover ,usually only several meters per day, which will not satisfy the requirements of future deep space planetary exploration, leading to the promotion of the concept of coordinated multi-mini rovers exploration in formation, under which exploration can be operated in many flexible and effective patterns, such as multiple rovers simultaneously explore as individual robot or hierarchically receives order from a central robot. Based upon above mentioned background, this project researches on providing autonomous navigation algorithms for coordinated multi-rovers exploration. The main research contents include: Internal relative navigation methods for inner formation; global navigation methods for multi-rovers; ground simulation tests, etc. Historic observation information is utilized to solve for the problem of owe observation in internal relative navigation. Also, method of Sparse Grid estimation is introduced to solve for the problem of large-scale non-linear estimation in coordinated navigation, and method of distributed single landmark SLAM algorithm is introduced to solve for the problem of high computation complexity and weak fault-tolerance due to dimensional dynamic variation of system model in the process of map construction, which leads to the development of a completed autonomous navigation method for multi-overs. This project can serve as an effective reference plan and provides theoretical supports for future planetary surface exploration missions for our country.

以NASA的 “机遇号”和“好奇号”为代表的大型单体漫游器虽然在火星探测任务中获得了成功，但却未能改变每日数十米探测范围的状况，已经无法满足未来深空天体探测的需求，这促使微小型漫游器编队协同探测概念的产生。多个漫游器构成的编队可以以多点同时或单点协同等多种灵活的方式高效开展行星探索。本项目以此为背景，研究微小型漫游器协同探测所需的自主导航方法。主要研究内容包括：漫游器编队内部相对位姿确定，漫游器协同全局导航以及地面仿真实验等。通过运用历史观测信息和定性导航方法，解决微小型漫游器相对定位中的欠观测问题；提出基于稀疏网格的估计方法，解决漫游器协同导航中的大规模非线性估计问题；建立单路标点分布式协同SLAM方法，解决地图构建过程中由系统模型维数动态变化造成的计算复杂度高、容错性差的问题，形成一套完整的微小型漫游器自主导航方法。本研究可为我国未来的行星表面探索任务提供方案借鉴和理论支撑。

由微小型漫游器组成编队对火星等深空天体展开探索，具有探测范围大、可靠性高和研制生产成本低等优势。但是实现漫游器编队探测深空天体，需要解决编队内部相对位置确定和编队整体定位问题。.本项目通过对深空天体表面特征检测与匹配方法、地形相关导航方法以及基于稀疏网格的估计方法的深入研究，建立了火星表面地形特征、广义特征鲁棒检测与匹配方法，构建了基于路标点的漫游器间相对导航方法，构建了基于稀疏网格的高斯滤波方法和积分点继承策略，以及单路标点分布式协同SLAM方法，建立了基于集合描述的三维地形特征匹配方法以及导航特征数据库构建与快速查询方法等，形成了漫游器编队内部相对导航和编队全局定位方法。创新性的研究成果包括：通过对图像特征提取与匹配方法深入研究，建立的仿射SURF特征检测方法有效的提高了强仿射变化条件下的图像特征检测精度，使特征误匹配率和计算量下降；针对火星表面地形建立的三维特征检测与匹配方法，创新性的引入集合描述特征，并利用集合相似性匹配特征，有效的解决了显著性较低带来的特征难以匹配的问题；提出的导航数据库构建方法通过特征相似性和空间分布筛选入库特征，在减小了数据库尺寸同时也减少了误匹配现象，并使导航定位精度得到提高；提出的基于稀疏网格的滤波方法通过积分点继承策略保留了系统高阶矩信息，提高了预测状态精度，从而提高了状态估计精度。这些针对火星漫游器展开的研究，其研究成果不但解决了火星漫游器相对/全局导航中的共性问题，同时也对机器人、航天器导航领域中的相关问题有重要的借鉴意义。