CAREER: Tackling Congestion in Smart Cities via Data-Driven Optimization-Based Control of Connected and Automated Vehicles

该学院早期职业发展计划（Career）奖支持基础研究，这些研究需要使用来自许多联网车辆的实时高频数据来自动控制驾驶和路线决策。这项研究本质上是跨学科的，因为它需要理解各个领域的研究和教育方法，包括交通系统、运筹学、控制理论、车辆动力学、数据科学和计算机科学。该项目将通过课程编制和参与研究，从多学科的角度对下一代科学家和工程师进行培训。PI将与高中教师和学生密切合作，针对代表性不足的少数族裔，通过面向第一代大学生的暑期实习项目和使用一套在线工具和案例研究的青年发展指导项目。该项目旨在为联网和自动驾驶车辆提供实时、在线、可预测的最优驾驶和路线决策。最终，该项目将有助于解决智能城市的交通拥堵问题。这项研究将填补将实时高频互联车辆数据集成到在线鲁棒自动驾驶和路线控制系统中的知识空白。具体而言，本研究将(1)通过“预测他人的预测”和在线学习的概念，提出基于数据驱动优化的模型预测控制（MPC）模型，用于交通不确定性下的协同互联和自动驾驶；(2)结合n人动态路由博弈和MPC模型，研究参与者和容量不确定条件下的非合作连通和自动路由问题；(3)通过优化排阵，平衡协同自动驾驶和非协同自动路由，提高网络效率。为了解决大规模网络中的计算可扩展性问题，本研究还将探索虚拟游戏类型算法来解决n人动态路由博弈，以及利用分布式优化来解决排编队问题中的计算问题。将使用联网和自动驾驶车辆进行模拟分析和道路测试，以分析模型的可行性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Faculty Early Career Development Program (CAREER) award supports fundamental research needed to use real-time high-frequency data from many connected vehicles to automatically control driving and routing decisions. This research is interdisciplinary in nature as it requires the understanding of research and education approaches in various fields including transportation systems, operations research, control theory, vehicle dynamics, data science, and computer science. The project will contribute to training the next generation of scientists and engineers in multidisciplinary perspectives through curriculum development and research involvement. The PI will work closely with high school teachers and students, targeting underrepresented minorities, through a Summer Internship Program for first-generation college students and a Youth Development Mentoring Program using a set of online tools and case studies. The project is to provide real-time, online, and predictive robust optimal driving and routing decisions to connected and automated vehicles. Ultimately, the project will help address traffic congestion in smart cities. This research will fill knowledge gaps in the integration of real-time high-frequency connected vehicle data into online robust automated driving and routing control systems. Specifically, this research will (1) suggest data-driven optimization-based model predictive control (MPC) models through the concepts of "forecasting the forecasts of others" and online learning for cooperative connected and automated driving under traffic uncertainty; (2) integrate n-person dynamic routing games and MPC models for non-cooperative connected and automated routing under player and capacity uncertainties; and (3) balance cooperative automated driving and non-cooperative automated routing through optimizing platoon formation to increase network efficiency. To address computational scalability issues in large scale networks, this research will also explore fictitious play type algorithms to solve the n-person dynamic routing games, as well as leverage distributed optimization to address computational issues in platoon formation problems. Simulation analysis and road testing will be conducted for feasibility analysis of the models using a fleet of connected and automated vehicles.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.