Abstract Manager

Autonomous driving is expected to revolutionize road traffic decreasing current externalities, especially accidents and congestion. Industries, academia, and governments have been working on autonomous driving for years and significant progress has been made in recent years. However, having truly safe automated driving will depend on defining the comprehensive list of overlapping conditions an Autonomous vehicle (AV) might encounter, what's called the Operational Design Domain (ODD).

Recent progress in this field has enabled the fully automated operation of AVs in normal weather conditions. In other words, an AV would have all knowledge in an ideal world. With its sensors, communication tools, and computational capability, it could anticipate any danger on the road and take no chances. However, the performance of AVs degrades drastically in adverse weather conditions like snowy, rainy, and foggy weather. The technical restrictions of current driving automation systems limit their ability based on operational circumstances. This is acknowledged in the SAE automation levels, which employ the idea of the ODD to specify the circumstances in which a specific driving automation system is intended to operate. The evaluation of an automated system to ascertain its ODD does not yet have a clear standard or a methodical methodology. Inappropriate usage of automation outside of the ODD has also been a significant contributor to accidents involving driving automation systems.

In this work, we study different approaches and architectural designs to achieve maximum functionality of AVs in adverse weather based on ODD management. We compare various sensors' performance in different weather conditions, review approaches to handle the system failures, and show how defining an appropriate ODD can be beneficial in this field.