基于移动机器人EKF-SLAM方法的一致性问题研究

发布时间：2018-02-26 00:14

本文关键词： 同时定位与建图机器人控制扩展卡尔曼滤波器能观测性分析估计不一致　出处：《沈阳建筑大学》2016年硕士论文　论文类型：学位论文

【摘要】：机器人技术由多种先进技术构成,同时也反映了一个国家的自动化科技水平,机器人技术在短短十几年时间里取得了飞速发展,并广泛应用在工业、军事、生活等各个领域。自主导航一直是移动机器人领域的研究热点,同时定位与地图创建(Simultaneous Localization and Mapping SLAM)使移动机器人具有更好的自主导航技术,因此对SLAM的研究成为了移动机器人领域的研究新的热点。首先,建立了移动机器人运动模型、传感器观测模型、环境特征模型及数据关联模型,并通过综合应用这些模型,建立解决SLAM问题的一般模型,为SLAM关键技术研究搭建了统一的平台。其次,对EKF-SLAM(Extended Kalman Filter-Simultaneous localization and mapping,EKF-SLAM)算法的一致性及收敛性问题进行了深入研究。经理论推导得出:当地标观测值趋近于无穷大时,机器人位姿的协方差存在极值,实现收敛;极值与地标位置估计精度均受机器人初始方向不确定性大小影响;如果在整个SLAM运行过程中,运动学模型和观测模型的雅可比矩阵都是在状态估计值处进行计算的话,该特征的方差会降低,也就是发生了方差不一致性现象;机器人方向的不确定性影响地标位置不确定性的极值或者下限。针对EKF-SLAM算法中的估计不一致问题,分析在两个条件下,EKF-SLAM算法产生不一致的基本原因,进行理论证明,并通过仿真结果验证推理过程及理论结果的准确性。最后,针对传统EKF-SLAM算法中存在的状态估计不一致问题,通过分析滤波器系统模型能观测性,提出一种增加能观测性约束条件的算法,使得EKF-SLAM系统与非线性SLAM系统观测方程能观矩阵的秩保持一致。主要通过构建补偿矩阵U,最优化求解约束条件,得到新的线性点,优化系统的雅克比矩阵,重构系统能观测矩阵。通过仿真实验,验证了所提出算法在状态估计的精确性和协方差一致性方面,明显优于传统的EKF-SLAM算法。研究工作和结论对车辆自主驾驶有重要参考价值。
[Abstract]:Robot technology is composed of a variety of advanced technologies, but also reflects the level of automation technology in a country, robotics technology has made rapid development in a short period of more than ten years, and widely used in industry, military, Autonomous navigation has always been a hot research topic in the field of mobile robot. At the same time, localization and map creation make mobile robot have better autonomous navigation technology. Therefore, the research of SLAM has become a new research hotspot in the field of mobile robot. Firstly, the motion model, sensor observation model, environment feature model and data association model of mobile robot are established, and these models are comprehensively applied. The general model of solving SLAM problem is established, which builds a unified platform for the research of SLAM key technology. Secondly, In this paper, the consistency and convergence of EKF-SLAM(Extended Kalman Filter-Simultaneous localization and mapping (EKF-SLAM) algorithm are studied in depth. The theoretical derivation shows that the covariance of robot pose exists extreme value and converges when the local standard observation value approaches infinity; The accuracy of the extreme value and the landmark position estimation is affected by the uncertainty of the initial direction of the robot, if the Jacobian matrix of the kinematics model and the observation model are calculated at the state estimation value during the whole SLAM operation. The variance of the feature will be reduced, that is, the variance inconsistency will occur, and the uncertainty of robot direction will affect the extreme value or lower limit of the uncertainty of the location of landmarks. This paper analyzes the basic reasons for the inconsistency of EKF-SLAM algorithm under two conditions, proves it theoretically, and verifies the reasoning process and the accuracy of the theoretical results by simulation results. Finally, aiming at the problem of inconsistent state estimation in the traditional EKF-SLAM algorithm, the paper analyzes the difference between EKF-SLAM algorithm and EKF-SLAM algorithm. By analyzing the observability of the model of filter system, an algorithm for increasing observability constraints is proposed. The rank of observable matrix of observation equation of EKF-SLAM system and nonlinear SLAM system is consistent. By constructing compensation matrix U and solving constraint conditions optimally, a new linear point is obtained to optimize the Jacobian matrix of the system. The simulation results show that the proposed algorithm is superior to the traditional EKF-SLAM algorithm in terms of state estimation accuracy and covariance consistency. The research work and conclusions have important reference value for autonomous vehicle driving.
【学位授予单位】：沈阳建筑大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP242

【相似文献】