Abstract: Many studies on traffic network evacuation have been conducted, but none have examined traffic network evacuation using rational inattention theory. With the randomness of evacuation network traffic state and the limited capacity of traveler information processing, this paper internalizes the cost of traffic state information and develops a bilevel evacuation network optimization model based on rational inattention theory. The upper model minimizes the total evacuation time of the system, with the decision variable being whether a link is unidirectional. The lower model is the user equilibrium model for rational inattention travelers. The DPSO-MSA hybrid heuristic algorithm is designed. The upper model is solved using the particle swarm optimization algorithm, and the resulting one-way strategy is passed down to the lower model, which is solved using the MSA algorithm. An example is provided to demonstrate the model's effectiveness. It is found that the optimal one-way strategy outperforms both the non-one-way strategy and the whole one-way strategy. The designed algorithm can quickly identify the key links of the evacuation network. For the whole evacuation system, more information obtained by the traveler may not lead to better performance. The results can provide a reference for the formulation of evacuation strategy

Key words: urban traffic, bi-level programming, DPSO-MSA algorithm, emergency evacuation, rational inattention

摘要： 研究交通网络疏散问题的文献较多，但鲜有基于理性疏忽理论来分析交通网络疏散问题的。本文考虑疏散网络交通状态的随机性和出行者信息处理能力的有限性，将交通状态信息成本内生化，建立基于理性疏忽理论的疏散网络双层优化模型。上层以系统总疏散时间最小为优化目标，将路段是否单行作为决策变量，下层建立基于理性疏忽理论的用户均衡模型。设计离散粒子群优化算法与逐次平均法相结合的启发式混合算法(DPSO-MSA)，上层采用粒子群算法求解，将上层得到的单行策略传递给下层，下层模型采用MSA方法求解，将得到的路段交通量返回给上层。并通过算例验证模型的有效性。研究发现，最优单行策略要优于非单行策略和全单行策略，设计的算法可以快速识别疏散网络的关键路段。对于整个疏散系统而言，出行者获取的信息并不是越多越好。研究结果可以为疏散策略的制定提供参考依据。

关键词: 城市交通, 双层规划, DPSO-MSA算法, 应急疏散, 理性疏忽