Novel raw Synthetic Aperture Radar data processing for rapid detection of road conditions

Purpose and goal

We will assess the feasibility of near–real-time processing of raw Synthetic Aperture Radar (SAR) data using artificial intelligence (AI) models to detect hazardous road conditions. This study will outline future steps towards developing a novel methodology to support the transport and infrastructure, as well as relevant public entities and authorities. Satellite SAR data are suitable for this goal because of their wide swaths and consistent revisit cycles, independent of weather and daylight.

Expected effects and result

AI-driven onboard SAR analytics directly contribute to enable Swedish road operators and actors, such as Trafikverket, the automotive industry, and civil preparedness (MCF), to deliver faster actionable insights, support autonomous vehicle systems, and strengthen emergency preparedness. In the long term, an integrated system could reduce the risk of accidents caused by environmental factors, lower maintenance costs, and improve operational efficiency.

Planned approach and implementation

The specific SAR radar wavelength for which data is accessible to will determine the set of use cases we can test the proof-of-concept AI model on (WP2). Based on this investigation we will collect a dataset for initial tests on base AI model (WP3). We will evaluate existing raw-echo AI models from related applications and determine whether to adapt a proven architecture or design a new one tailored to road-condition inference (WP4).