• The survey provides a systematic taxonomy that categorizes physical adversarial attacks and defenses into real-world, simulator-based, and digital-world scenarios, offering a structured understanding of the threat landscape.
• It analyzes adversarial vulnerabilities across different sensor modalities—camera, LiDAR, and multifusion—highlighting the unique challenges each sensor type faces in autonomous driving systems.
• Defense mechanisms are classified into input image preprocessing, adversarial example detection, and model enhancement, covering the full spectrum of countermeasures for DNN-based systems.
• The paper identifies open challenges and future research directions for enhancing the robustness and safety of autonomous driving systems against physical adversarial threats.