SummaryAVL develops and validates technologies that increase the reliability of automated vehicles in adverse weather and low-visibility conditions. Within the European ROADVIEW project (Robust Automated Driving in Extreme Weather), AVL developed, integrated and validated methods under both real-world and controlled environmental conditions to reduce weather-induced sensor disturbances and improve downstream perception and localization.
Key contributions and capabilities- Participation in ROADVIEW with 14 partners from industry and academia to address perception, localization, decision-making and vehicle control under adverse weather.
- Development and integration of 3D LiDAR outlier detection and filtering methods to remove measurement points caused by rain, fog or snow.
- Measured improvements on perception after filtering: semantic segmentation improved by approximately 5%; object detection improved by up to approximately 2%.
- Extensive multi-season sensor data collection capturing rain, fog and snow across a temperature range from -10 °C to 50 °C.
- Controlled, reproducible validation at the AVL Mobility and Sensor Center in Roding, including a weather chamber able to generate repeatable rain, fog and lighting conditions.
- Investigations of V2X approaches where infrastructure sensors share environment, object and weather data with vehicles to extend perception when on-board visibility is limited.
- Research into LiDAR-based localization matching live sensor data against high-resolution, semantically enriched environment representations for GNSS-degraded or unavailable conditions.
Testing & validation environment- Real-world data collection across seasons and environments to capture varied adverse-weather phenomena affecting sensors (rain, fog, snow).
- Controlled weather-chamber testing with reproducible scenarios to systematically evaluate automated driving functions under extreme conditions.
Representative technical details and findings- 3D LiDAR outlier detection: methods to identify and remove weather-generated spurious points to improve perception quality.
- Perception improvement metrics observed in project tests: semantic segmentation ≈ +5%; object detection up to ≈ +2% after filtering.
- Tested LiDAR performance across temperatures from -10 °C to +50 °C.
- Weather chamber capabilities: simulated rainfall rates up to 120 mm/h; dense fog simulation with visibility down to ~7 m; adjustable lighting including complete darkness for sensor and perception testing.
Positioning and system integrationAVL combined algorithm development, system integration, test infrastructure and validation capability to demonstrate how perception filtering, precise localization, connected infrastructure and realistic validation scenarios can jointly improve reliability of automated driving functions in everyday conditions.
Quote“Automated driving must prove itself not only under ideal weather conditions. If the technology is to become part of everyday life, it must also be able to cope with rain, fog, snow, and limited visibility. ROADVIEW has demonstrated how robust perception, precise localization, connected infrastructure, and realistic validation can work together. For AVL, it is essential not only to develop technologies, but also to integrate and validate them under both real-world and reproducible conditions. In this way, we are creating important foundations for automated mobility that can be deployed more safely and reliably in everyday applications in the future,” says Georg Schwab, Managing Director, AVL Software and Functions.
Technical specifications- Project: ROADVIEW (Robust Automated Driving in Extreme Weather)
- Project partners: AVL + 14 partners (industry & academia)
- Sensor focus: LiDAR, cameras (perception and reliability assessment under adverse weather)
- LiDAR outlier filtering: 3D detection and removal of weather-induced measurement points
- Perception improvements observed: semantic segmentation ≈ +5%; object detection up to ≈ +2%
- Real-world data: collected across all four seasons (rain, fog, snow)
- Controlled testing site: AVL Mobility and Sensor Center, Roding
- Weather chamber capabilities: rainfall simulation up to 120 mm/h; dense fog simulation (~7 m visibility); variable lighting including complete darkness
- Tested LiDAR temperature range: -10 °C to +50 °C
- Connectivity approaches: V2X integration to provide infrastructure-sourced environment and weather information to vehicles
- Localization approaches: LiDAR-based matching against high-resolution, semantically enriched maps for GNSS-independent positioning