Risk perception and quantitative assessment for intelligent vehicles: a systematic review of methodologies, policy implications, and transportation systems integration

Study	Focus area	Scope	Limitation addressed by this study
Xiong et al.^[4]	Driving risk assessment in multi-vehicle interaction	Specific scenarios (interactions)	Limited to specific interaction scenarios; does not cover broader risk architecture or decision-making ethics.
Pei & Hou^[5]	Safety assessment of urban arterial traffic flow	Traffic flow level	Focuses on traffic flow management rather than vehicle-level risk perception mechanisms.
Cheng et al.^[6]	Driving safety risk progress (identification, prediction, warning)	Technical methodologies	Lacks integration with transportation policy and ethical governance.
This review	Risk perception, quantitative assessment, policy, and decision-making	Systematic: 'human-vehicle-road' + policy integration	Proposes a holistic framework integrating technical standards with practical transport applications and ethical considerations.

Table 1.

Comparison between this review and existing literature.

Assessment dimension	Typical methods	Key indicators/models	Strengths	Limitations	Ref.
Risk architecture	'Human-vehicle-road' synergistic framework	Driving risk field, neural domains	Holistic perspective: Integrates multi-source factors	High modeling complexity: Difficult to quantify interactions	Sun et al.^[7]; Lisowski^[10]
Risk identification and prediction	Multi-source data fusion	LiDAR, radar, camera fusion; LSTM, transformer	High real-time capability: High prediction accuracy in structured environments	Sensitive to sensor noise: Poor generalization in unstructured scenarios	Zhang et al.^[14]; Han et al.^[19]
Quantitative risk evaluation	Spatiotemporal risk modeling	TTC, THW, Spatial-Temporal Risk Field (STRF)	Visualizable risk distribution: Good guidance for path planning	Parameter tuning relies on experience: Static environment assumptions	Ahmad et al.^[18]; Zhang & Guo^[24]
Risk decision-making	Reinforcement learning and game theory	DRL (SAC, DDPG), dynamic game, MPC	Adaptive decision-making: Handles multi-vehicle interactions	'Black box' nature: Ethical trade-offs hard to encode	Fang et al.^[35]; Yin et al.^[43]

Table 2.

Summary of risk assessment methodologies for intelligent vehicles.

Category	Key challenges	Impact on deployment	Potential solutions/future directions
Data governance	Data silos: Fragmentation between traffic management agencies and OEMs: Heterogeneous data standards.	Hinders cross-departmental collaboration: Limits system-wide optimization: Inefficient incident management.	Establish unified data standards (e.g., European Mobility Data Space); Utilize Blockchain for secure sharing; Build integrated traffic big data centers.
Ethical trade-offs	Fairness Constraints: Conflicts between pedestrian vs passenger protection: Algorithmic bias in risk quantification.	Raises legal disputes: Erodes public trust: Hinders social acceptance of IVs.	Develop ethical guidelines aligned with legal frameworks; Incorporate fairness metrics into algorithm design; Adopt transparent 'utilitarian' or 'random' decision protocols.
Model generalization	Environmental Adaptability: 'One-size- fits-all' models fail in unstructured roads (rural lanes) or corner cases.	Safety risks in complex/mixed traffic scenarios: Limitation of simulation validation.	Context-aware risk modeling; Scenario-based testing (ISO 21448); Integration of Edge AI for real-time local processing.

Table 3.

Summary of technical bottlenecks and potential solutions in IV risk assessment.