Balancing Safety, Efficiency, and Sustainability: A Survey of Modeling Tools for Autonomous Vehicles Interactions

Autonomous vehicles (AVs) promise substantial gains in road safety, traffic efficiency, and environmental sustainability, yet these objectives often impose conflicting requirements on modeling approaches. Selecting appropriate models is therefore essential to achieve a balanced and reliable AV decision-making process in mixed-traffic environments. This survey systematically reviews 1) traditional rule-based and physics-driven interaction models, 2) data-driven and learning-based approaches, and 3) hybrid methods that integrate interpretable safety logic with adaptive intelligence. It also examines simulation platforms and real-world validation strategies as part of the AV development pipeline. The analysis shows that traditional approaches provide interpretability and formal safety assurances but lack adaptability; learning-based models capture complex, context-dependent behaviors but present challenges in verification and deployment; hybrid methods have emerged as promising candidates for jointly addressing safety, efficiency, and sustainability; and the persistent simulation-to-reality gap remains a major barrier to trustworthy operation. To organize and unify these developments, the paper proposes a three-tier unified taxonomy that links modeling paradigms to their dominant objectives and embeds simulation and real-world validation as an operational tier. The taxonomy structures the survey and provides a coherent lens for synthesizing methods, assessing trade-offs, and guiding future AV interaction modeling research.