Data-driven Development Framework for ADAS and Automation for Marine Applications

Abstract

The maritime industry is increasingly adopting Advanced Driver Assistance Systems (ADAS) and automation, drawing inspiration from progress in the automotive sector. Applying these technologies to marine environments, however, introduces unique challenges such as sensor limitations, unpredictable conditions, and the lack of standardized validation methods.

This thesis presents a data-driven framework to support the testing and validation of marine ADAS/AD (Autonomous Driving) systems. The framework focuses on data logging, structured data handling, and the use of Key Performance Indicators (KPIs) as objective measures of performance. In collaboration with Volvo Penta and Volvo GTT (Group Trucks Technology), a proof of concept was developed around two representative features, with relevant KPIs defined. The framework was partially implemented on a test vessel equipped with LiDAR (Light Detection and Ranging) and cameras for perception, a Dynamic Positioning System (DPS) for positioning, and a high-bandwidth logger to capture raw sensor data during real operations.

While post-test data ingestion, KPI calculation, and KPI-driven refinement were not completed within the scope of this thesis, these stages are outlined as future extensions. The work provides a foundation for a systematic, data-driven development methodology in the marine ADAS/AD domain, bridging the gap between conceptual design and a fully operational validation pipeline.