Predicting Disrupted Network Behavior

由于人们在基础设施系统(例如道路系统)内进行大规模的交互，网络建模的区域通常被用来描述所产生的行为的影响。传统上，网络均衡的概念一直被用来模拟许多个人的长期稳态行为，每个人都为了自己的利益而行动。虽然均衡对于基础设施规划和管理至关重要，但它需要几个关键假设，如熟悉度和合理性，这些假设在高压力干扰情况下可能不成立。此研究项目针对发生重大中断、扰乱预期网络状态时的网络行为提出了新的模型。这项研究的主要假设是，个人可以根据他们对干扰的感知以及在途中学习到的信息来改变和调整之前的预期，在不熟悉的情况下，网络用户更重视系统和特定于环境的特征，如路线和道路几何形状、风险偏好和旅行限制(例如，当不熟悉真实的预期成本时，用户可能会选择更长的路径，仅仅是因为它使他们最初更接近目的地)。这项研究将通过心理学实验来识别这些新的个体行为，然后为由此产生的网络影响开发新的数学公式。通过采用前面提到的新的问题特征，可以从根本上开发出针对大规模网络中断的数学系统描述和预测。通过实现卓越的预测能力，通过能够更好地为灾难和疏散可能性做好准备，可以实现实质性的社会改善。此外，通过更好地了解非平衡行为，即使是几乎每天都有实质性的非极端改善也是可以实现的，例如减轻非经常性拥堵和交通事件(这两个领域长期以来都有复杂的交通规划)的影响。许多更广泛的好处也将出现在交通系统之外。随着这项研究解决网络行为的基本问题，许多使用网络模型的领域都可以采用新的行为模型的各个方面。在教育方面，更仔细地考虑用心理行为进行网络建模将带来实质性的好处。此外，两个合作伙伴将开展外展工作，并与德克萨斯大学奥斯汀分校赞助的项目相结合，向学生介绍研究和实践，重点是招收多样化的本科生和研究生。通过这样的计划(包括德克萨斯大学奥斯汀分校的高级学院和美国实习生计划与NSF REU计划相结合)，PI在过去多次成功地招收了如此多样化的学生，并致力于形成密切合作的跨学科研究努力。

As people interact at a large-scale within infrastructure systems (such as roadway systems), the area of network modeling is often employed to characterize the impact of the resulting behavior. Traditionally, the concept of network equilibrium has been employed which models the long-term steady state behavior of many individuals each acting in their own self interest. While equilibrium has been critical for infrastructure planning and management, it requires several key assumptions such as familiarity and rationality which may not hold true in high stress disruptive situations. This research project addresses new models for network behavior when significant disruptions occur which upset the expected network state. The primary hypotheses of this research is that individuals can transform and adapt previous expectations based on their perception of the disruption as well as information learned en-route and that in unfamiliar cases network users place greater weight on system and context-specific characteristics such as route and road geometry, risk preference, and travel constraints (e.g., when unfamiliar with the true expected cost, users may select a longer path simply because it moves them closer to the destination initially). This research will discern these new individual behaviors through psychological experiments and then develop novel mathematical formulations for the resulting network impacts.By adopting the new problem characteristics noted in the previous paragraph, fundamentally new mathematical system descriptions and predictions can be developed for large-scale networks subjected to disruptions. By achieving superior prediction capabilities, substantial societal improvements are achievable by being able to better prepare for disaster and evacuation possibilities. Furthermore, by better understanding non-equilibrium behavior even substantial near-daily non-extreme improvements are achievable such as mitigating the impact of non-recurrent congestion and traffic incidents (both areas which have long complicated transportation planning). Numerous broader benefits will also be seen beyond transportation systems. As this research addresses the fundamental problem of network behavior, numerous fields which employ network models can adopt aspects of the new behavioral models. Educationally, substantial benefits will result from the closer consideration of network modeling with psychological behavior. Further, outreach efforts will be conducted by both Co-PIs and in conjunction with programs sponsored by UT Austin to introduce students to research and practice, with an emphasis on recruiting a diverse mix of undergraduate and graduate students. Through such programs (including the Advanced Institute and US Intern program at UT-Austin combined with the NSF REU program) the PIs have been repeatedly successful in the past in recruiting such a diverse mix of students and are committed to forming a closely cooperative interdisciplinary research effort.