Doppler Velocity Log (DVL) Measurements

该数据集记录了地中海海域自主式水下航行器（AUV）上搭载的多波束测速声纳（DVL）在作业期间的数据，重点在于对缺失的多普勒速度记录进行回归分析。该数据集包含了多样化的动态数据，以便进行全面的训练，并且是通过使用Teledyne RDI的Work Horse Navigator DVL进行收集的。规模上，数据集包含了11个独立的部分，每部分持续400秒，用于训练，同时还提供了额外的测试片段。任务是对速度进行预测，并对缺失的多普勒速度记录光束进行回归分析。

This dataset records data collected during the operation of a multibeam velocimetry sonar (DVL) mounted on an Autonomous Underwater Vehicle (AUV) in the Mediterranean Sea, with the core focus on regression analysis for missing Doppler velocity records. This dataset contains diverse dynamic data to support comprehensive model training, and was gathered using Teledyne RDI's Work Horse Navigator DVL. In terms of scale, the dataset consists of 11 independent training segments, each with a duration of 400 seconds, alongside additional test segments provided separately. The core task of this dataset is to conduct velocity prediction and regression analysis on the missing Doppler velocity record beams.

搜集汇总

背景与挑战

背景概述

该数据集包含由Snapir AUV在地中海哈法海岸附近采集的约7小时实验数据，用于A-KIT（自适应卡尔曼信息Transformer）模型的研究。数据涵盖不同机动、深度和速度，采样率为INS 100 Hz和DVL 1 Hz，旨在改进传感器融合中的过程噪声估计，提升导航精度。数据集分为训练集和测试集，训练集包含11个400秒片段，测试集包含2个额外片段，用于评估模型性能。