devCusto/vehicle-motion-sensors-fusion

DriveFusion是一个真实世界的多模态车辆传感器数据集，结合了同步的GPS、速度、航向、陀螺仪(IMU)和里程表数据。数据集已完全匿名化，非常适合用于轨迹预测、传感器融合、车辆动力学、时间序列建模和驾驶行为分析等研究。数据集包含Speedometer.csv、Gyroscope.csv和Mileage.csv三个文件，分别记录了GPS+速度+航向、IMU角速度和里程表日志的原始数据。数据集支持多种模态，包括GPS（纬度、经度、高度、精度、卫星数）、速度计（速度km/h、移动标志）、航向（罗盘方向）、陀螺仪（角速度x、y、z）和里程表（累计距离）。数据集合并采用基于时间戳的策略，确保数据对齐。数据集适用于GPS轨迹预测、IMU-GPS传感器融合、车辆运动建模、驾驶行为分类、预测性维护、时间序列预测、深度学习序列模型（LSTM、Transformer、TCN）以及卡尔曼滤波/SLAM研究。数据集在MIT许可证下发布，允许广泛重用，但需注明出处。

DriveFusion is a real-world multimodal vehicle sensor dataset combining synchronized GPS, speed, heading, gyroscope (IMU), and odometer data. The dataset is fully anonymized and ideal for research in trajectory prediction, sensor fusion, vehicle dynamics, time-series modelling, and driving behaviour analysis. The dataset includes Speedometer.csv, Gyroscope.csv, and Mileage.csv files, which record raw GPS + speed + heading, IMU angular velocity, and odometer logs, respectively. The dataset supports multiple modalities, including GPS (latitude, longitude, altitude, accuracy, satellites), speedometer (speed km/h, movement flag), heading (compass direction), gyroscope (angular velocity x, y, z), and odometer (cumulative distance). The dataset merging uses a timestamp-based strategy to ensure data alignment. The dataset is suitable for GPS trajectory prediction, IMU-GPS sensor fusion, vehicle motion modelling, driving behaviour classification, predictive maintenance, time-series forecasting, deep learning sequence models (LSTM, Transformer, TCN), and Kalman filtering/SLAM research. The dataset is released under the MIT License, allowing broad reuse with attribution.