SensoDat: 32,580 Simulation-Based Self-Driving Car Test Cases with 81 Sensors

SensoDat is a dataset of 32,580 executed simulation-based test cases for self-driving cars, generated with state-of-the-art test generators. It provides trajectory logs and time-series data from 81 different simulated sensors, such as rpm, wheel speed, brake thermals, and transmission. The dataset was created by Christian Birchler at the University of Bern to reduce dependency on expensive hardware and software for autonomous systems research.

Use Cases

• Training and validating AI models for autonomous driving based on the 81 simulated sensor streams.
• Developing regression testing techniques for simulation-based self-driving car systems using the executed test cases.
• Studying flakiness and reliability in simulation environments based on the generated trajectory and sensor logs.
• Analyzing vehicle dynamics and sensor correlations from the time-series data of components like rpm and brake thermals.

Strengths

• Contains 32,580 executed simulation-based test cases, providing a substantial corpus for analysis.
• Includes data from 81 different simulated sensors, offering a wide variety of vehicle state information.
• Generated with state-of-the-art test generators, suggesting a focus on methodological rigor.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count and file size are unknown, which may limit suitability assessment.

Provenance

Source

Christian Birchler, University of Bern

Collection Method

Generated via simulation-based test execution using state-of-the-art test generators for self-driving cars.