The Oxford RobotCar Dataset is a large-scale dataset for autonomous driving, collected over a year (May 2014 to December 2015) using an autonomous Nissan LEAF. The dataset includes over 1000km of recorded driving, with 20 million images from 6 cameras, LIDAR, GPS, and INS data. It was collected in various weather conditions, including heavy rain, night, and snow, and includes data from a route in central Oxford that changed over time due to roadworks. The dataset is available for download and includes raw sensor data, sensor calibrations, and MATLAB development tools for data analysis. The dataset is designed to support research on long-term autonomous driving in dynamic urban environments. It captures a wide range of variations in scene appearance and structure due to illumination, weather, dynamic objects, seasonal effects, and construction. The data includes 23.15TB of information, with 1010.46km of recorded driving. The data is divided into individual routes and chunks, each corresponding to a 6-minute segment of the route. Each chunk is packaged as a tar archive, with timestamps and condition tags for easy access. The dataset includes data from multiple sensors, including 3 monocular cameras, 2 2D LIDARs, 1 3D LIDAR, and a GPS+Inertial navigation system. The data is formatted in standard data formats for portability, including images in lossless-compressed PNG format, 2D and 3D LIDAR scans in binary format, and GPS+Inertial data in ASCII-formatted CSV files. Visual odometry data is also provided, which can be used as a reference for local pose estimation. The dataset includes MATLAB development tools for image demosaicing, undistortion, 3D pointcloud generation, and projection of 3D pointclouds into camera images. These tools enable researchers to analyze the data and develop algorithms for autonomous driving. The dataset is intended to accelerate research towards long-term autonomy for autonomous vehicles and mobile robots. The authors also provide lessons learned from the data collection process, including the importance of logging raw data, using forward-compatible formats, and separating logged and processed data.The Oxford RobotCar Dataset is a large-scale dataset for autonomous driving, collected over a year (May 2014 to December 2015) using an autonomous Nissan LEAF. The dataset includes over 1000km of recorded driving, with 20 million images from 6 cameras, LIDAR, GPS, and INS data. It was collected in various weather conditions, including heavy rain, night, and snow, and includes data from a route in central Oxford that changed over time due to roadworks. The dataset is available for download and includes raw sensor data, sensor calibrations, and MATLAB development tools for data analysis. The dataset is designed to support research on long-term autonomous driving in dynamic urban environments. It captures a wide range of variations in scene appearance and structure due to illumination, weather, dynamic objects, seasonal effects, and construction. The data includes 23.15TB of information, with 1010.46km of recorded driving. The data is divided into individual routes and chunks, each corresponding to a 6-minute segment of the route. Each chunk is packaged as a tar archive, with timestamps and condition tags for easy access. The dataset includes data from multiple sensors, including 3 monocular cameras, 2 2D LIDARs, 1 3D LIDAR, and a GPS+Inertial navigation system. The data is formatted in standard data formats for portability, including images in lossless-compressed PNG format, 2D and 3D LIDAR scans in binary format, and GPS+Inertial data in ASCII-formatted CSV files. Visual odometry data is also provided, which can be used as a reference for local pose estimation. The dataset includes MATLAB development tools for image demosaicing, undistortion, 3D pointcloud generation, and projection of 3D pointclouds into camera images. These tools enable researchers to analyze the data and develop algorithms for autonomous driving. The dataset is intended to accelerate research towards long-term autonomy for autonomous vehicles and mobile robots. The authors also provide lessons learned from the data collection process, including the importance of logging raw data, using forward-compatible formats, and separating logged and processed data.