Synthetic Data for Autonomous Mining Vehicles

The SynDAB system generates high-fidelity, safety-critical synthetic data to train and validate autonomous mining equipment. It reproduces off-road mining scenarios in a virtual environment, enabling the simulation of rare conditions such as landslides, low visibility, and unexpected object detection.

• Supports AI robustness in rugged mining environments
• Integrates with ROS-based autonomy stacks
• Enables cost-efficient development of autonomous mining operations
• Accelerates safety validation against ISO standards

• Autonomous haul trucks and loaders
• AI model development for mineral extraction robotics
• Simulation-based training for mining operations

In both underground and open-pit mining operations, autonomous vehicles must function in visually degraded, GPS-denied, and topographically complex environments. The lack of real-world training data for such conditions limits the performance and safety validation of AI-driven systems. Traditional test methods are resource-intensive and dangerous, offering limited reproducibility and low coverage of operational edge cases. Meeting standards for functional safety and system reliability necessitates advanced simulation environments with synthetic data capabilities.

The invention originates from the Institute of Industrial Automation and Software Engineering (IAS) at the University of Stuttgart, under the leadership of Prof. Michael Weyrich (Institute Director) and Prof. Christof Ebert. Their research focuses on AI-based testing and validation of autonomous and safety-critical systems operating in complex, high-risk environments such as mining, construction, and manufacturing.

Building on previous work in simulation-driven validation and software quality assurance, this invention extends those methods to the mining sector, enabling AI training and certification under extreme environmental conditions that are difficult to replicate in reality.

Current validation of autonomous mining systems requires testing in remote, hazardous, and visually degraded environments. These testing efforts are logistically complex, time-intensive, and carry significant risk for operators and assets.
Conventional testing methods cannot generate or reproduce the edge cases critical for functional safety, leading to potential blind spots in system behavior. With safety standards becoming more stringent, there is an urgent need for controlled, repeatable, and risk-informed synthetic testing environments that simulate the full range of mining operations.

The University of Stuttgart’s solution allows for the synthetic generation of operational and failure scenarios tailored to autonomous mining vehicles operating underground or in open-pit mines. These environments often present extreme topographies, dust interference, and poor lighting conditions, making real-world testing infeasible for many scenarios.

The system’s AI engine selects risk-prioritized scenarios to generate a Minimum Viable Test Set (MVTS) that ensures safety validation with fewer but more targeted simulations. Synthetic sensor data (e.g., degraded vision, sensor dropout) supports training and validation across a wide range of operational contexts.

The technology can be used in ROS-based digital twins or custom mining simulators and supports standards such as IEC 61508 (functional safety), ISO 17757, and ISO 19011 for internal audits. It reduces testing burden, enables simulation of high-risk scenarios, and accelerates approval for autonomous systems in mining operations.

The technology can be commercialized through licensing, providing a direct route to industry adoption. Alternatively, the inventors’ team at the University of Stuttgart has implemented several of the algorithms in software, which can be adapted through cooperation or consulting projects for specific mining automation, risk management, or digital twin applications.

Ebert, C. et al. (2023). AI-Based Testing for Autonomous Vehicles. ResearchGate link
IAS (2022). Testing Software Systems. University of Stuttgart. Link
 

Fickinger, M. et al. (2023). Simulation-based Validation for Construction Robotics. Automatisierungstechnik, De Gruyter. Link