基于Netvlad神经网络的室内机器人全局重定位方法

针对已知地图的室内机器人全局重定位、绑架恢复问题,提出一种基于改进的Netvlad卷积神经网络的室内机器人全局重定位方法,通过激光雷达获取的障碍物信息引导机器人到达空旷区域,粗定位阶段,使用栅格地图最短连通域距离作为正样本判据,并对Netvlad引入残差网络,通过图像检索得到机器人的粗略位置及角度信息。使用粗定位阶段得到的位置和角度信息作为自适应蒙特卡罗定位的初始值来估计机器人的精确位姿。实验结果表明,与传统定位方法相比,该方法可以使机器人从绑架问题中快速恢复准确位姿。     

基于EKF的全景视觉机器人SLAM算法

研究全景视觉机器人同时定位和地图创建（SLAM）问题。针对普通视觉视野狭窄, 对路标的连续跟踪和定位能力差的问题, 提出了一种基于改进的扩展卡尔曼滤波（EKF）算法的全景视觉机器人SLAM方法, 用全景视觉得到机器人周围的环境信息, 然后从这些信息中提取出环境特征, 定位出路标位置, 进而通过EKF算法同步更新机器人位姿和地图库。仿真实验和实体机器人实验结果验证了该算法的准确性和有效性, 且全景视觉比普通视觉定位精度更高。     

基于二维码视觉与激光雷达融合的高精度定位算法

为解决以蒙特卡罗定位算法为代表的激光室内定位算法存在的定位精度差和抗机器人绑架性能差的问题,以及传统二维码定位算法环境布置复杂且对机器人运行轨迹有严格要求的问题,提出了一种融合二维码视觉和激光雷达数据的移动机器人定位算法。机器人首先利用机器视觉技术搜索检测环境中的二维码,然后将检测出二维码的位姿分别转换至地图坐标系下,并融合生成先验位姿信息。而后以此作为初始位姿进行点云对准以得到优化后的位姿。同时引入里程计-视觉监督机制,从而有效解决了包括二维码信息缺失、二维码识别错误等由环境因素带来的问题,并保证了位姿的平滑性。基于移动机器人的实验结果表明,所提算法比经典的自适应蒙特卡罗定位（AMCL）算法的雷达采样点平均误差下降了92%,单次位姿计算时间减少了88%,可有效解决机器人绑架问题,并应用于以仓储机器人为代表的室内移动机器人。     

基于人工路标的易部署室内机器人全局定位系统

在一些布局易变或存在较多动态障碍物的室内,移动机器人的全局定位依然面临较大的应用挑战.针对这类场景,实现了一种新的基于人工路标的易部署室内机器人全局定位系统.该系统将人工路标粘贴在不易被遮挡的天花板上来作为参照物,仅依赖一个摄像头即能实现稳定的全局定位.整个系统根据具体的功能分为地图构建和全局定位两个过程.在地图构建过程中,系统使用激光SLAM算法所输出的位姿估计结果为基准,根据相机对路标点的观测信息来自动估计人工路标点在全局坐标系中的位姿,建立人工路标地图.而在全局定位过程中,该系统则是根据相机对地图中已知位姿的人工路标点的观测信息,结合里程计与IMU融合的预积分信息来对位姿进行实时估计.充分的实验测试表明,机器人在该系统所部署范围内运行的定位误差稳定在10 cm以内,且运行过程可以保证实时位姿输出,满足典型实际室内移动机器人全局定位的应用需求.     

基于视觉惯性里程计与语义信息的无人机SLAM方法研究

室内环境中存在丰富的语义信息,可以使机器人更好地理解环境,提高机器人位姿估计的准确性。虽然语义信息在机器人同时定位与地图构建(SLAM)领域得到了深入研究和广泛应用,但是在环境准确感知、语义特征提取和语义信息利用等方面还存在着很多困难。针对上述难点,提出了一种基于视觉惯性里程计算法与语义信息相结合的新方法,该方法通过视觉惯性里程计来估计机器人的状态,通过校正估计,构建从语义检测中提取的几何表面的稀疏语义地图;通过将检测到的语义对象的几何信息与先前映射的语义信息相关联来解决视觉惯性里程计和惯性测量单元的累积误差问题。在室内环境中对装备RGB-D深度视觉和激光雷达的无人机进行验证实验,结果表明,该方法比视觉惯性里程计算法取得了更好的结果。应用结合语义信息和视觉惯性里程计的SLAM算法表现出很好的鲁棒性和准确性,该方法能提高无人机导航精度,实现无人机智能自主导航。     

基于先验地图的视觉重定位方法综述

在自动驾驶、增强现实和智能移动机器人领域,视觉重定位是非常重要的基础问题.视觉重定位是指根据视觉传感器实时拍摄的数据,在已有先验地图中确定位置和姿态的问题.过去数十年间,该问题受到广泛关注,涌现出种类繁多的先验地图构建方法和视觉重定位方法.这些工作差异大,涉及范围广,技术概括和总结尚缺乏.因此,对视觉重定位领域进行综述具有重要的理论和应用价值.尝试为视觉重定位相关方法建立一个统一的蓝图,从图像数据在大规模地图数据库中查询的角度对相关工作进行分析和总结.综述不同类型地图数据库构建方法、不同特征匹配、重定位和位姿计算方法,总结目前视觉重定位的主流数据集,最后分析视觉重定位存在的挑战和潜在发展方向.     

基于视觉FastSLAM的移动机器人自主探索方法*

针对室内环境下机器人的移动和定位需要,提出基于视觉FastSLAM的移动机器人自主探索方法.该方法综合考虑信息增益和路径距离,基于边界选取探索位置并规划路径,最大化机器人的自主探索效率,确保探索任务的完整实现.在FastSLAM 2.0的基础上,利用视觉作为观测手段,有效融合全景扫描和地标跟踪方法,提高数据观测效率,并且引入地标视觉特征增强数据关联估计,完成定位和地图绘制.实验表明,文中方法能正确选取最优探索位置并合理规划路径,完成探索任务,并且定位精度和地图绘制精度较高,鲁棒性较好.     

低成本激光和视觉相结合的同步定位与建图研究

激光雷达和视觉传感是目前两种主要的服务机器人定位与导航技术,但现有的低成本激光雷 达定位精度较低且无法实现大范围闭环检测,而单独采用视觉手段构建的特征地图又不适用于导航应用。因此,该文以配备低成本激光雷达与视觉传感器的室内机器人为研究对象,提出了一种激光和视觉相结合的定位与导航建图方法：通过融合激光点云数据与图像特征点数据,采用基于稀疏姿态调整的优化方法,对机器人位姿进行优化。同时,采用基于视觉特征的词袋模型进行闭环检测,并进一步优化基于激光点云的栅格地图。真实场景下的实验结果表明,相比于单一的激光或视觉定位建图方 法,基于多传感器数据融合的方法定位精度更高,并有效地解决了闭环检测问题。     

多机器人协同SLAM技术研究进展

由于单机器人同步定位与建图（SLAM）技术在实际应用中的局限性,多机器人协同SLAM技术以较强的灵活性和鲁棒性受到研究人员的广泛关注,并且在农业生产、环境监测、海上搜救等领域具有巨大应用前景。多机器人协同SLAM是多机器人协同工作的核心及大范围复杂环境内及时获得场景感知信息的关键,能使多个机器人在协同工作时共同定位并构建任务空间地图,主要基于单机器人SLAM算法、多机器人系统架构、地图融合等技术实现。结合多机器人协同SLAM的发展历程,对比分析当前主流的多机器人协同SLAM算法。从传感器的角度,将多机器人协同SLAM分为激光协同SLAM、视觉协同SLAM以及激光视觉融合协同SLAM三类,并对多机器人协同SLAM的架构选择、多机通信、相对位姿、地图融合和后端优化问题进行讨论,同时指出异构机器人协同、基于深度学习的语义SLAM是多机器人协同SLAM的未来发展趋势。     

同时机器人定位与人体跟踪的多源感知协作粒子滤波方法

提出了分布式多传感器协作的条件粒子滤波算法以解决人与机器人位置的联合概率分布估计问题．全局视觉系统中,各视角独立运行图像平面上基于粒子滤波的目标跟踪,并利用地平面单应关系实现多视角目标主轴同步融合．视觉观测进一步与机器人激光数据以顺序滤波方式异步融合,提出包含人体位置假设的激光似然场模型以提高对机器人位姿误差的鲁棒性,并引入基于Kullback-Leibler距离的自适应采样以降低描述联合分布所需的粒子数目．实验验证了该方法能够在具有观测噪声且人—机位置均不确定的情况下利用多传感器协作实现基于地图的同时机器人定位与人体跟踪．     

A two-stage mobile robot localization method by overlapping segment-based maps

This paper presents a new method for accurately estimating the pose (position and orientation) of a mobile robot by registering a segment-based local map observed from the current robot pose and a global map. The method works in a two-stage procedure. First, the orientation is determined by aligning the local and global map through a voting process based on a generalized Hough transform. Second, it uses a coarse-to-fine approach for selecting candidate positions and a weighted voting scheme to determine the degree of overlap of the two maps at each of these poses. Unlike other methods previously proposed, this approach allows us to uncouple the problem of estimating the robot orientation and the robot position which may be useful for some applications. In addition it can manage environments described by many (possibly short) segments. This paper presents some experimental results based on our mobile robot RAM-2 that show the accuracy and the robustness of the proposed method even for poor quality maps and large dead-reckoning errors.     

Robust mapping and localization in indoor environments

Conventional localization methods have been developed for indoor static environments such as the home environment. In dynamic environments such as factories and warehouses, however, it is difficult to estimate the accurate robot pose. Therefore, we propose a novel approach for the estimation of the robot pose in a dynamic or large environment for which fixed features are used. In the proposed method, a ceiling-feature map is built using an upward-looking monocular camera. This map is created accurately from the robot pose using a laser scanner and an estimation based on the iterative closest point method. The ceiling-feature map consists of features such as lamps and the FREAK, and its creation can be more accurate if the sliding-window technique and bundle-adjustment schemes are used. During the post-mapping navigation, the robot pose is estimated using the Monte Carlo localization method based on the ceiling-feature map. In dynamic experiments, the proposed method shows a high repeatability and stability in real-world conditions and applications.     

Differential evolution solution to the SLAM problem

A new solution to the Simultaneous Localization and Modelling problem is presented in this paper. The algorithm is based on the stochastic search for solutions in the state space to the global localization problem by means of a differential evolution algorithm. This non linear evolutive filter, called Evolutive Localization Filter (ELF), searches stochastically along the state space for the best robot pose estimate. The set of pose solutions (the population) focuses on the most likely areas according to the perception and up to date motion information. The population evolves using the log-likelihood of each candidate pose according to the observation and the motion errors derived from the comparison between observed and predicted data obtained from the probabilistic perception and motion model.The proposed SLAM algorithm operates in two steps: in the first step the ELF filter is used at local level to re-localize the robot based on the robot odometry, the laser scan at a given position and a local map where only a low number of the last scans have been integrated. In the second step, the aligned laser measures and the corrected robot poses are used to detect whether the robot is revisiting a previously crossed area (i.e., a cycle in the robot trajectory exists). Once a cycle is detected, the Evolutive Localization Filter is used again to estimate the accumulated residual drift in the detected loop and then to re-estimate the robot poses in order to integrate the sensor measures in the global map of the environment.The algorithm has been tested in different environments to demonstrate the effectiveness, robustness and computational efficiency of the proposed approach.     

Simultaneous localization and mapping using the Geometric Projection Filter and correspondence graph matching

A common way of localization in robotics is using triangulation on a system composed of a sensor and some landmarks (which can be artificial or natural). First, when no identifying marks are set on the landmarks, their identification by a robust algorithm is a complex problem which may be solved using correspondence graphs. Secondly, when the localization system has no a priori information about its environment, it has to build its own map in parallel with estimating its position, a problem known as simultaneous localization and mapping (SLAM). Recent works have proposed to solve this problem based on building a map made of invariant features. This paper describes the algorithms and data structure needed to deal with landmark matching, robot localization and map building in a single efficient process, unifying the previous approaches. Experimental results are presented using an outdoor robot car equipped with a two-dimensional scanning laser sensor.     

基于扫描匹配的室外环境SLAM 方法

针对室外环境中的机器人“绑架”问题,提出了基于地图匹配的SLAM方法．该方法舍弃了机器人里程计信息, 只利用局部地图和全局地图的图形相关性进行机器人定位．方法的核心是多重估计数据关联,并将奇异值分解应用到机器人位姿计算中．利用Victoria Park数据集将本算法与基于扩展卡尔曼滤波器的方法进行比较,实验结果证明了本文提出的算法的有效性．     

激光—相机系统语义栅格建图和路径规划

目的 SLAM（simultaneous localization and mapping）是移动机器人在未知环境进行探索、感知和导航的关键技术。激光SLAM测量精确，便于机器人导航和路径规划，但缺乏语义信息。而视觉SLAM的图像能提供丰富的语义信息，特征区分度更高，但其构建的地图不能直接用于路径规划和导航。为了实现移动机器人构建语义地图并在地图上进行路径规划，本文提出一种语义栅格建图方法。方法 建立可同步获取激光和语义数据的激光-相机系统，将采集的激光分割数据与目标检测算法获得的物体包围盒进行匹配，得到各物体对应的语义激光分割数据。将连续多帧语义激光分割数据同步融入占据栅格地图。对具有不同语义类别的栅格进行聚类，得到标注物体类别和轮廓的语义栅格地图。此外，针对语义栅格地图发布导航任务，利用路径搜索算法进行路径规划，并对其进行改进。结果 在实验室走廊和办公室分别进行了语义栅格建图的实验，并与原始栅格地图进行了比较。在语义栅格地图的基础上进行了路径规划，并采用了语义赋权算法对易移动物体的路径进行对比。结论 多种环境下的实验表明本文方法能获得与真实环境一致性较高、标注环境中物体类别和轮廓的语义栅格地图，且实验硬件结构简单、成本低、性能良好，适用于智能化机器人的导航和路径规划。     

On the planning of collision-free movements of manipulators

This paper proposes an efficient position identification method for mobile robots in the environment of a building corridor using colour images and map information. A robot can usually estimate its position from its motion history (so-called dead reckoning); however, there are occasions when the robot's position needs to be estimated without the motion history, for example when self-tracking of the motion has failed, or just after the robot power is on. The proposed method is to identify the robot's position without the motion history. The method consists of the following three steps: (1) map information for the mobile robot is prepared; (2) a colour image is processed to detect a vanishing point and to generate an abstracted image. The robot moves to an appropriate position for the identification, if it is unable to identify the current position; and (3) the current robot position is identified from the map information, the vanishing point, and the abstracted image. The effectiveness of the proposed method is shown by the experimental results.