Safety risks in Advanced driver Assistance Systems (ADAS): A systematic review of causation and analysis methodologies.

Abstract

Advanced Driver Assistance Systems (ADAS) have achieved rapid technological advancement, yet their widespread adoption has introduced new and multifaceted safety challenges. This review provides a systematic synthesis of ADAS-related accident scenarios and risk factors. Based on multinational accident databases and detailed crash reports, collision types are classified into three major categories: vehicle-to-vehicle, vehicle-to-pedestrian, and vehicle-to-non-motorized interactions. The underlying causes of these accidents are traced to sensor performance degradation under adverse weather, algorithmic limitations in dynamic traffic environments, and excessive driver dependence on automated functions. Furthermore, this review examines mainstream analytical approaches, including statistical modeling, system-theoretic frameworks, and AI-driven techniques, and evaluates their applicability in risk identification and mitigation. Finally, a forward-looking framework is proposed to advance multi-sensor data fusion and privacy-preserving governance, strengthen dynamic scene modeling and open scenario databases, and incorporate human factors into causal inference for proactive risk management. The findings are expected to enrich the methodological foundation of ADAS accident analysis and support future improvements in intelligent vehicle safety.