A Systematic Review of Influences and Test Methodologies: Evaluating the Impact of Adverse Weather Conditions on Perception Systems for Intelligent Vehicles

Abstract

Intelligent vehicles (IVs) are the strategic high ground of future automotive technology, with perception systems (PSs) acting as critical intermediaries between IVs and the driving environment. Adverse weather conditions (AWCs) substantially degrade PSs’ performance, elevating IV safety risks. To evaluate AWCs’ impact on PSs systematically, this article surveys the existing literature from two perspectives: AWCs’ influences and corresponding test methodologies. First, this article examines how four AWCs (rain, fog, snow, and unconventional lighting) influence PSs (LiDAR, radar, camera, and ultrasonic sensors), respectively, emphasizing original signal interference mechanisms. Second, virtual and real-world test methodologies are discussed. The virtual test methods can be summarized as X-in-the-loop (XIL), including software-in-the-loop (SIL), hardware-in-the-loop (HIL), and vehicle-in-the-loop (VIL) techniques. Specifically, this article reviews 12 simulation software capable of simulating AWCs, analyzes four types of HIL test methods and associated platforms, and explores VIL studies addressing AWCs testing. For real-world testing, 11 globally recognized closed test sites with AWCs generation capabilities are summarized. Also, 28 natural driving datasets for testing PSs in AWCs are introduced. Finally, emerging research trends and future directions are discussed. Findings indicate PSs frequently reveal design vulnerabilities under AWCs. Thus, accurate, quantitative, and efficient evaluation methods are essential for IV advancement and deployment. Future research on PS evaluation under AWCs is expected to advance through integrated multimodal sensor simulation, generative modeling for realistic AWC scene synthesis, automated edge-case scenario creation, and enrichment of naturalistic datasets with AWC annotations.