多传感器异步融合算法AUV对接导航系统

AUV docking navigation system based on multi-sensor asynchronous fusion algorithm

针对研发的自主水下机器人(autonomous underwater vehicle,AUV)在回收对接过程中,由于采样频率不同而导致多传感器组合导航系统精度较低的问题,提出一种基于联邦滤波结构的多尺度无迹卡尔曼(unscented Kalman,UK)异步融合滤波算法。该算法依据采样率划分多尺度信息,建立基于联邦滤波的多尺度系统误差模型,并针对不同尺度信息采用UKF算法进行非等间隔异步融合,从而得到全局的最优状态估计。仿真实验表明,同基于卡尔曼滤波(Kalman filtering,KF)的单一尺度融合算法和基于UKF的多尺度融合算法相比,该算法对AUV回收对接导航系统中异步采样信息具有更高的融合精度,提升了导航系统的可靠度,为AUV的回收对接提供了安全保障。最后通过开阔水域实验验证了AUV对接导航系统的有效性。

In order to solve the problem of low accuracy of multi-sensor integrated navigation system due to different sampling frequencies during the recovery and docking process of the developed autonomous underwater vehicle(AUV),a multi-scale unscented Kalman(UK)asynchronous fusion filter algorithm based on the federated filter structure is proposed.The algorithm divides multi-scale information according to the sampling rate,establishes a multi-scale system error model based on federated filtering,and uses the UKF algorithm to perform non-equal interval asynchronous fusion for different scale information,so as to obtain the global optimal state estimation.Simulation experiments show that compared with the single-scale fusion algorithm based on Kalman filtering(KF)and the multi-scale fusion algorithm based on UKF,this algorithm has higher fusion accuracy for asynchronous sampling information in the AUV recovery docking navigation system.Improve the reliability of the navigation system,and provide a safety guarantee for the recovery and docking of the AUV.Finally,the open water experiment verified the effectiveness of the AUV docking navigation system.