Usage

The RobotCycle dataset supports research in autonomous mobility, safety analysis, multimodal perception, and urban planning.

Initial Analysis

1. 3D Reconstruction and Odometry

• Challenging dataset due to sharp maneuvers and non-rigid motion.
• Serves as a benchmark for dynamic and real-world odometry estimation.
• Evaluated the robustness of KISS-ICP for LiDAR odometry on dynamic cycling data.

2. Risk and Interactions Analysis

• Classified segments into four infrastructure categories (Types A–D): no cycle lane, on-road lane, segregated track, and shared-use path.
• Correlated traffic intensity and lane geometry with vehicle proximity, TTC, and lane infingements.
• Identified high-risk areas such as The Plain Roundabout and the road works in Woodstock Road.

3. Eye Gaze and Attention Mapping

• Generated heatmaps of visual focus using Project Aria gaze tracking.
• Cyclists focus on junctions, approaching vehicles, and pedestrian crossings in high-traffic areas.
• Enables research on human attention, stress, and risk anticipation.

4. Safety Perception Analysis

• Combined self-reported safety ratings with sensor data.
• Found strong correlation with road layout (L1) and other road users’ behavior (L4).
• Supported by incident data from participant surveys and CrashMap.

Potential Applications

The RobotCycle dataset supports a wide range of research applications, including but not limited to:

• Benchmarking SLAM and odometry pipelines under highly dynamic conditions.
• Developing and evaluating multimodal perception, combining LiDAR, vision, IMU, and GPS.
• Studying how cyclists and vehicles interact and share space in real-world traffic environments.
• Quantifying how different types of cycling infrastructure influence safety, comfort, and risk.
• Analysing human attention and situational awareness using gaze, head-motion, and environmental context data.

For more information about the data collection methodology, sensor calibration, open-source toolkit, and analysis results, please refer to our paper:

The Oxford RobotCycle Project: A Multimodal Urban Cycling Dataset for Assessing the Safety of Vulnerable Road Users (PDF)