Abstract: To address the challenges of multi-scale collaborative modeling and insufficient interaction continuity in existing traffic simulation systems, this paper proposes a unified multi-scale traffic simulation framework integrating micro-, meso-, and macro-level modeling. The framework follows a core workflow of data-driven processing, scale classification, strategy execution, trajectory control, and simulation feedback, and establishes a modular architecture for multi-source information fusion and behavior-strategy decoupling.The framework establishes a mesoscale bridging layer that enables bidirectional state propagation between individual vehicle dynamics and network-level flow evolution, thereby mitigating scale discontinuities. A social-value–driven group Monte Carlo Tree Search (MCTS) mechanism is further incorporated to support cooperative multi-agent decision-making under heterogeneous interaction patterns. At the macroscopic level, a dynamic, time-activated Origin–Destination (OD) generation module is constructed to capture temporal variations in travel demand and to regulate system-wide flow input adaptively.Comprehensive experiments demonstrate that the proposed framework achieves substantial improvements in both behavioral plausibility and scenario adaptiveness while maintaining computational scalability. Compared with state-of-the-art baselines, the method reduces conflict occurrence by 19.8% and increases network throughput by 34.6%, highlighting its effectiveness as a robust modeling paradigm for intelligent traffic coordination and large-scale dynamic simulation. 

Key words: multi-scale, collective Monte Carlo Tree Search (MCTS) -based cooperative strategy, microscopic trajectory planning, time-varying Origin-Destination (OD) generation mechanism, multi-agent

摘要： 针对交通仿真系统中多尺度协同建模困难与交互连续性不足的问题，设计了一种融合微观、介观与宏观的多尺度交通仿真框架。该框架以“数据驱动-尺度分类-策略执行-轨迹控制-仿真反馈”为核心逻辑，构建模块化建模体系，实现多源信息融合与行为策略解耦。通过建立介观桥梁层，打通微观行为与宏观趋势之间的双向信息通路，并融入基于社会价值导向的群体蒙特卡洛树搜索（MCTS）协同决策机制；同时，在宏观层构建动态OD（Origin-Destination）矩阵生成系统，赋予仿真系统对交通流时序演变的自适应调控能力。实验结果表明，本文方法在保证仿真效率的同时，显著提升系统行为合理性与场景适应力，与传统的单一尺度模型和无协同策略的单车决策模型相比，冲突率降低19.8%、通行率提升34.6%，为复杂交通场景下的智能协同与资源优化提供了有效的技术路径。

关键词: 多尺度, 群体蒙特卡洛树搜索协同策略, 微观轨迹规划, 时变OD生成机制, 多智能体