Partner Institution: National Natural Science Foundation of China general project

Project Information: 2026.1-2029.12

Project Description: The reliable operation of Connected and Automated Vehicles (CAVs) in complex dynamic traffic environments hinges on the deep integration of explainable trajectory prediction, safe collaborative planning, and interactive adaptation capabilities. However, existing approaches still exhibit significant limitations in the physical interpretability of trajectory prediction, social compliance of multi-vehicle collaborative planning, and robustness of interactions in non-stationary mixed traffic scenarios. To address these challenges, this study proposes a comprehensive decision-making optimization framework for CAVs in complex dynamic environments:

Explainable Trajectory Prediction: A hybrid prediction model integrating Social Artificial Intelligence (Social AI) and control-theoretic principles is designed, which embeds vehicle dynamics constraints and interaction-aware attention mechanisms to enhance the physical plausibility and behavioral interpretability of predicted trajectories;

Collaborative Safe Planning: A game-theoretic and Inverse Reinforcement Learning (IRL) co-optimization framework is developed to quantify human drivers’ social normative preferences, thereby generating planning strategies compliant with traffic regulations and compatible with multi-vehicle intentions;

Interactive Adaptation Enhancement: An adaptive interaction decision system based on the Condensation of Variables (CoV) framework and Immersion and Invariance (I&I) control is proposed, validated through a high-fidelity simulation platform that enables real-time adjustment of interaction strategies in unstructured roads and mixed traffic flow environments.