基于交互式多模型卡尔曼滤波的AUV超短基线跟踪算法

AUV Ultra-short Baseline Tracking Algorithm Based on Interactive Multi-Model Kalman Filter

在复杂海洋环境下,利用超短基线对自主水下航行器(AUV)进行跟踪定位可能会受到各类误差的影响,通常采用以最小均方误差为准则的卡尔曼滤波对动态定位数据进行处理。构建起与目标实际运动相匹配的运动模型,是保证卡尔曼滤波精度和可靠性的重要基础,而AUV具有机动性较强的特点,往往难以先验性地确定单一的运动模型实现对所有运动状态的匹配。针对基于单模型卡尔曼滤波无法全程适应水下目标的所有运动状态的问题,采用交互式多模型卡尔曼滤波方法处理AUV的超短基线跟踪数据,运动模型之间通过概率矩阵转移来增强运动状态的适应性,实验结果表明该算法在多模型集合构建合理的情况下,其状态适应性优于单模型卡尔曼滤波算法。

Owing to complex marine environments,the tracking and positioning of autonomous undersea vehicles(AUVs)that use ultra-short baseline may be affected by various errors,and a Kalman filter based on the minimum mean square error is usually used to process the dynamic positioning data.It is important to ensure the accuracy and reliability of the Kalman filtering to construct a motion model that matches the actual motion of the target.However,the AUV is characterized by strong maneuverability,which often renders it difficult to a priori determine a single motion model to achieve the matching of all motion states.To address the inability of the single-model based Kalman filter to adapt to all the motion states of an underwater target,an interactive multi-model Kalman filter(IMMKF)algorithm was used to process the ultra-short baseline tracking data of an AUV.Furthermore,a probability matrix transfer between motion models was used to enhance the adaptability of motion states.The experimental results showed that the IMMKF algorithm was better than the Kalman filter algorithm for a single model when the multi model set was constructed reasonably.