SBIR Phase I: Information fusion-driven adaptive corridor-wide traffic signal re-timing

这个小型企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力将通过实施完全自适应的交通信号重新定时解决方案来显著减少交通延误，碰撞和死亡。最近的国家交通信号报告卡给了一个不及格的等级D+交通信号操作在美国。这些不及格的成绩是尽管机构每年花费约20亿美元的信号操作，维护和资本改善。如果美国将其信号支持在“A”级，公众将看到：（i）交通延误减少15-40%，旅行时间节省高达25%，停车减少10- 40%;（ii）燃料消耗减少10%或更多，导致全国每年节省近1700亿加仑的汽车燃料;及（iii）减少高达22%的有害排放物。从科学和技术的角度来看，这一努力将成为人工智能和机器学习的一个有影响力的成功故事。由于小企业是爱荷华州州立大学创业工厂的产品，该项目预计将涉及学生寻找行业经验的项目，导致他们更全面的教育。这个小型企业创新研究（SBIR）第一阶段项目将开发和展示一个完全自适应的交通信号重新定时解决方案的概念验证。这项工作的关键智力价值将是开发深度学习模型，从一系列异构信息源中提取抽象特征，以执行特征级融合。在特征提取后，拟议的解决方案将使用可扩展的深度强化学习模型来获得重新定时决策。强化学习过程将帮助系统适应不同时间尺度上不断变化的交通场景，而无需大量的人工干预。该解决方案将灵活适用于车载和基于云的计算，具体取决于此类平台的可用性。总体而言，拟议的系统将减少实施时间以及资本和维护支出。这些优势将鼓励美国各地和国际上的城市采取这样的重新定时战略，并将大大改变这个市场的当前景观。这个奖项反映了NSF的法定使命，并已被认为是值得支持的评估使用基金会的知识价值和更广泛的影响审查标准。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project will result from a significant reduction in traffic delays, crashes, and fatalities by implementing a fully adaptive traffic signal re-timing solution. The most recent National Traffic Signal Report Card gave a failing grade of D+ to traffic signal operations in the United States. These failing grades are despite the fact that agencies spend approximately $2 billion every year on signal operation, maintenance, and capital improvements. If the US supports its signals at an "A" level, the public would see: (i) a 15-40% reduction in traffic delay, travel time savings up to 25%, and a 10-40 % reductions in stops; (ii) a 10 % or more reduction in fuel consumption resulting in nationwide savings of almost 170 billion gallons of motor fuels per year; and (iii) up to 22% reduction in harmful emissions. From a science and technology perspective, this effort will be an impactful success story for artificial intelligence and machine learning. As the small business is a product of Iowa State University start-up factory, the project is expected to involve students looking for industry experiences in the project leading to a more comprehensive education for them. This Small Business Innovation Research (SBIR) Phase I project will develop and demonstrate proof-of-concept of a fully adaptive traffic signal re-timing solution. The key intellectual merit of this effort will be developing deep learning models to extract abstract features from a range of heterogeneous information sources to perform feature-level fusion. Upon feature extraction, the proposed solution will use scalable deep reinforcement learning models to obtain re-timing decisions. The reinforcement learning process will help the system adapt to changing traffic scenarios at different time-scales without the need for significant manual interventions. The solution will be flexible for both onboard and cloud-based computing, depending on the availability of such platforms. Overall, the proposed system will reduce implementation time and capital and maintenance expenditures. These advantages will encourage cities around the US and internationally to adopt such a re-timing strategy and will dramatically transform the current landscape of this market.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.