无人飞行器集群仅测距初始相对位姿确定方法研究

Research on Determining Initial Relative Pose of Unmanned Aerial Vehicles Using a Range-Only Method

针对全球导航卫星系统(GNSS)拒止环境下无人飞行器集群成员间的相对定位问题,研究了一种基于机载惯性测量单元(IMU)、气压高度计与数据链测距组合的初始相对位姿求解算法。首先,在高度计稳定输出较为精确的高度信息的前提下,将飞行器的三维运动解耦成二维水平运动,给出了三维加速度和角速度、测距信息的水平坐标系投影等效模型。在此基础上,以待求量相对位置和航向角的非线性形式构造了新的待求状态量,并将相对位置和航向角的非线性求解问题转化成了新状态量的线性模型最小二乘求解问题。然后,通过引入递推最小二乘算法(RLS),建立了该相对位姿求解算法的实时输出递推形式,有效降低了机载在线计算的负载。接着,对所提算法进行了可观测性分析,并给出了使系统状态量不可观测的几种相对运动形式。最后,对所提算法进行了数值仿真实验,仿真结果表明,该算法能够有效、快速求解初始相对位姿,位置误差在初始相对距离的10%以内,航向角误差在初始相对角度的1%以内。

Aiming at the problem of relative localization among members of unmanned air vehicles(UAVs) under global navigation satellite system(GNSS) denied environment, an initial relative pose estimation algorithm based on the combination of on-board inertial measurement unit(IMU), barometric altimeter and data link ranging is studied. Firstly, under the circumstances that the barometric altimeter stably outputs relatively accurate altitude information, the three-dimensional motion of UAV is decoupled into two-dimensional horizontal motion, and the equivalent model of the horizontal coordinate system projection of the three-dimensional acceleration, angular velocity, and ranging information is given. Secondly, based on the horizontal coordinate system, a new state quantity to be obtained is constructed containing relative position and heading angle in a nonlinear form, and the nonlinear solution problem of relative position and heading angle is transformed into the linear least squares(LS) problem of the new state quantity. Thirdly, the real-time output of the relative pose solving algorithm is established in recursive form by introducing the recursive least squares algorithm(RLS), which effectively reduces the load of real-time on-board computing. Then, the observability analysis of the proposed algorithm is performed, and several forms of relative motion between UAVs that make the state quantity of the system unobservable are given. Finally, numerical simulations of the proposed algorithm are performed. The simulation results show that the proposed algorithm can effectively and quickly solve the initial relative pose estimation problem. The position and heading error are within 10% and 1% of the initial relative distance and heading, respectively.