基于多普勒的改进粒子滤波地形辅助导航算法

An improved Doppler-based particle filter algorithm for terrain-aided navigation

关键词:  自主水下航行器  混合自适应重采样  粒子滤波  地形辅助导航  卡尔曼滤波

Key words:  Autonomous underwater vehicles  Hybrid adaptive resampling  Particle filter  Terrain-aided navigation  Kalman filter

摘要:

为满足低成本自主水下航行器（AUV）的高精度导航需求，选用多普勒计程仪（DVL）作为核心测速测深传感器。针对粒子滤波（PF）过程中重采样导致的粒子多样性匮乏问题，提出了一种混合自适应重采样粒子滤波（HARPF）的地形辅助导航算法。通过引入自适应高斯变异和交叉重采样策略，并设计粒子接受判断准则，以提高粒子分布的多样性和覆盖能力。为进一步提升导航性能，构建了基于HARPF与卡尔曼滤波融合（HARPF-KF）的导航算法框架，利用HARPF算法输出的位置信息与DVL的测速信息进行滤波融合，从而优化导航结果。实验结果表明，与传统PF算法相比，HARPF算法的导航精度提升了9.3%，HARPF-KF算法的导航精度提升了38.4%，有效改善了AUV的导航精度和鲁棒性。

Abstract:

To meet the high-precision navigation requirements of low-cost autonomous underwater vehicles (AUVs), the Doppler velocity log (DVL) has been selected as the primary sensor for measuring velocity and depth. To address the deficiency in particle diversity caused by resampling in the particle filter (PF) process, a hybrid adaptive resampling particle filter (HARPF) algorithm for terrain-aided navigation is proposed. The algorithm incorporates adaptive Gaussian mutation and crossover resampling strategies, as well as particle acceptance criteria, to enhance particle diversity and distribution coverage. To further improve navigation performance, a navigation algorithm framework based on the HARPF and Kalman filter fusion (HARPF-KF) has been constructed. This framework integrates the position information output from the HARPF algorithm with velocity information from the DVL for fusion filtering, thereby optimizing the navigation results. Experimental results demonstrate that the HARPF algorithm improves navigation accuracy by 9.3% compared to the traditional PF algorithm, while the HARPF-KF algorithm enhances it by 38.4%, effectively boosting the accuracy and robustness of AUV navigation.