Multi-agent Collaborative Adaptive Cruise Control Based On Reinforcement Learning

Evaluation of Drivers’ Interaction Ability at Social Scenarios: A Process-Based Framework

“…Constructing a model that accurately describes the interactive process of drivers is essential for evaluating interaction behavior. There are different kinds of methods to model the human driving behavior, such as reinforcement learning (Peng et al 2021;Liu et al 2023a) with attention mechanism (Liu, Sun, and Qi 2023;Liu et al 2024c), which is a powerful data-driven method, and game theory methods (Hang et al 2020;Liu et al 2024b). Compared with data-driven methods, game-based methods have the advantages in stability and interpretability, which have been widely used for behavior modeling in recent years (Wang et al 2022;Liu et al 2024b).…”