基于自适应零速修正机制的低速无人车定位方法

A Low-Speed Unmanned Vehicle Positioning Method Based on Adaptive Zero Velocity Update Mechanism

针对缺失全球导航卫星系统(GNSS)信号条件下的低速无人车定位问题,在多状态约束卡尔曼滤波(MSCKF)框架下,提出了一种基于自适应零速修正机制的低速无人车定位方法(AZUPT-MSCKF)。传统MSCKF算法利用惯性测量单元(IMU)传播车辆运动信息,并利用相机测量实现对运动信息的校正。然而,当无人车处于静止状态时,相机测量更新停止。受到IMU累积误差的影响,无人车的定位性能将迅速下降。对此,本文提出的AZUPT-MSCKF方法通过新增的自适应零速修正机制校正IMU的信息传播,使得无人车定位方法能较好地适应静态场景。实验结果表明,相比于传统MSCKF算法及VINS-Mono算法(关闭回环检测),AZUPT-MSCKF方法具有更高的定位精度和更强的鲁棒性。

In order to solve the problem of low-speed unmanned vehicle positioning without GNSS signals, a low-speed unmanned vehicle positioning method(AZUPT-MSCKF)based on adaptive zero velocity update mechanism was proposed under the framework of Multi-State Constraint Kalman Filter(MSCKF). Traditionally, the IMU was adopted by the MSCKF to spread the vehicle motion information, and the cameras were used for the correction of the motion information by image measurements. However, the measuring process by the cameras stopped when the vehicle was in the static scene. And then, the positioning performance of the unmanned vehicle will decline rapidly due to the cumulative error of IMU. In this paper, a novel algorithm named AZUPT-MSCKF was developed, and an adaptive zero velocity update mechanism was adopted to correct the information transmission of IMU. Hence, the unmanned vehicle positioning method can be adapted to the static scene. Experimental results showed that the AZUPT-MSCKF has higher positioning accuracy and stronger robustness compared with the traditional MSCKF and the VINS-Mono(the loop-closure is prohibited).