In this paper, a bearing-based three-dimensional self-localization and distributed circumnavigation with connectivity preservation and collision avoidance are investigated for a group of quadrotor-type unmanned aerial...In this paper, a bearing-based three-dimensional self-localization and distributed circumnavigation with connectivity preservation and collision avoidance are investigated for a group of quadrotor-type unmanned aerial vehicles (UAVs). A leader–follower structure is adopted, wherein the leader moves with reference dynamics (a target). Different from the existing approaches that necessitate full knowledge of the time-varying reference trajectory, in this paper, it is assumed that only some vehicles (at least one) have access to the bearing relative to the target, and all other vehicles are equipped with sensors capable of measuring the bearings relative to neighboring vehicles. In this paper, a consensus estimator is proposed to estimate the global position for each vehicle using relative bearing measurements and an estimate of neighboring vehicles received from a direct communication network. Then, a continuous robust integral of the sign of the error (RISE) control approach is effectively integrated with the distributed vector field approach to ensure UAV formation orbiting around the moving target while avoiding obstacles and maintaining network links within available communication ranges. In contrast to the classical RISE control rule, a \(\tanh (\cdot )\) function is used instead of the \(\text {sgn}(\cdot )\) function to further decrease the high-gain feedback and to obtain a smoother control signal. Furthermore, by using the localized radial basis function (RBF) neural networks (NNs) in a cooperative way, deterministic learning theory is employed to accurately identify/learn model uncertainties resulting from the attitude dynamics. The convergence of the entire closed-loop system is illustrated using the Lyapunov theory and is shown to be uniformly ultimately bounded. Finally, numerical simulations show the effectiveness of the proposed approach.展开更多
SLAM(simultaneous localization and mapping)是无人载体实现自主导航定位的关键技术。针对传统视觉SLAM系统在动态场景下导航定位精度低的问题,在视觉SLAM系统的基础上引入惯性传感器(inertial measure-ment unit)。在ORB-SLAM3系统...SLAM(simultaneous localization and mapping)是无人载体实现自主导航定位的关键技术。针对传统视觉SLAM系统在动态场景下导航定位精度低的问题,在视觉SLAM系统的基础上引入惯性传感器(inertial measure-ment unit)。在ORB-SLAM3系统的基础上设计了一种面向动态环境的视觉惯性SLAM系统。提出一种基于向量场一致性(vector field consensus,VFC)的稀疏光流法来追踪图像的特征点并计算基础矩阵,分别利用光流对极几何约束和惯性传感器信息计算特征点的动态概率,提出一种联合的动态特征检测方法计算特征点的总动态概率,并将动态概率大于阈值的特征点进行剔除,在SLAM系统的前端实现了视觉信息与惯性运动信息的紧耦合。在数据集上的实验结果表明,该视觉惯性SLAM改进算法有良好的性能表现。展开更多
文摘In this paper, a bearing-based three-dimensional self-localization and distributed circumnavigation with connectivity preservation and collision avoidance are investigated for a group of quadrotor-type unmanned aerial vehicles (UAVs). A leader–follower structure is adopted, wherein the leader moves with reference dynamics (a target). Different from the existing approaches that necessitate full knowledge of the time-varying reference trajectory, in this paper, it is assumed that only some vehicles (at least one) have access to the bearing relative to the target, and all other vehicles are equipped with sensors capable of measuring the bearings relative to neighboring vehicles. In this paper, a consensus estimator is proposed to estimate the global position for each vehicle using relative bearing measurements and an estimate of neighboring vehicles received from a direct communication network. Then, a continuous robust integral of the sign of the error (RISE) control approach is effectively integrated with the distributed vector field approach to ensure UAV formation orbiting around the moving target while avoiding obstacles and maintaining network links within available communication ranges. In contrast to the classical RISE control rule, a \(\tanh (\cdot )\) function is used instead of the \(\text {sgn}(\cdot )\) function to further decrease the high-gain feedback and to obtain a smoother control signal. Furthermore, by using the localized radial basis function (RBF) neural networks (NNs) in a cooperative way, deterministic learning theory is employed to accurately identify/learn model uncertainties resulting from the attitude dynamics. The convergence of the entire closed-loop system is illustrated using the Lyapunov theory and is shown to be uniformly ultimately bounded. Finally, numerical simulations show the effectiveness of the proposed approach.
文摘SLAM(simultaneous localization and mapping)是无人载体实现自主导航定位的关键技术。针对传统视觉SLAM系统在动态场景下导航定位精度低的问题,在视觉SLAM系统的基础上引入惯性传感器(inertial measure-ment unit)。在ORB-SLAM3系统的基础上设计了一种面向动态环境的视觉惯性SLAM系统。提出一种基于向量场一致性(vector field consensus,VFC)的稀疏光流法来追踪图像的特征点并计算基础矩阵,分别利用光流对极几何约束和惯性传感器信息计算特征点的动态概率,提出一种联合的动态特征检测方法计算特征点的总动态概率,并将动态概率大于阈值的特征点进行剔除,在SLAM系统的前端实现了视觉信息与惯性运动信息的紧耦合。在数据集上的实验结果表明,该视觉惯性SLAM改进算法有良好的性能表现。
