EAGER: A Cloud-assisted Framework for Improving Pedestrian Safety in Urban Communities using Crowd-sourced Mobile and Wearable Device Data

行人安全仍然是城市社区的一个重要问题。最近的几份报告表明，与交通事故有关的伤亡人数稳步上升，行人分心是这类事故的主要原因之一。主要以独立的方式在用户的智能电话和移动的设备上操作的用于提高行人安全性的现有系统和技术具有若干设计缺陷以及性能和可用性问题，这些设计缺陷和性能以及可用性问题已经阻碍了它们的成功采用和使用。该项目的目标是通过设计准确、有效和可用的工具和技术来改善行人安全，这些工具和技术可以很容易地被城市用户采用。为了实现这一目标，该项目计划追求一个集中的研究议程，涉及新技术和一些探索性和未经测试的想法。作为拟议行人安全框架的一部分，将通过采用即将推出的移动的和可穿戴设备提供的多传感器和异构数据，开发准确且节能的设备上分心检测技术。在这个方向上，监督和半监督学习将用于设计有效的活动分类和分心预测技术，这些技术将使用概念验证实现进行经验评估。与现有的独立方法不同，拟议的框架采用了一种连接社区的方法，以准确地捕捉行人自己的行动以及其他人的行动对他/她的安全的影响。这涉及到隐私保护和云辅助数据分析引擎的设计和实施，以捕获，分析和通知行人即将发生的危险情况，从参与用户获得的众包分心数据。最后，一个全面的性能和可用性评估将进行部署一个大规模的测试平台，涉及参与者从威奇托州立大学（WSU）的校园社区。 该项目的成果，包括计划中的试验台，将对改善WSU校园社区内的行人安全产生重大影响。如果成功，还可以设想在城市或全市范围内进行类似的试验。除了改善行人安全外，该项目还将教育用户和参与者了解技术对行人安全的影响及其在改善行人安全方面的作用。项目成果和结果将通过同行审查的出版物、白色文件和开放源码应用程序传播。从计划的试验台收集的应用程序和匿名数据将适当传播，以促进行人安全技术领域的进一步研究和进步。

Pedestrian safety continues to be a significant concern in urban communities. Several recent reports indicate that injuries and fatalities in pedestrian-related accidents are steadily rising and that pedestrian distraction is one of the leading causes in such accidents. Existing systems and techniques for improving pedestrian safety, which primarily operate on users' smartphones and mobile devices in a stand-alone fashion, have several design drawbacks and performance and usability concerns that have precluded their successful adoption and usage. The goal of this project is to improve pedestrian safety by designing accurate, efficient and usable tools and techniques, which can be easily adopted by urban users. In order to accomplish this goal, this project plans to pursue a focused research agenda involving novel technologies and several exploratory and untested ideas. As part of the proposed pedestrian safety framework, accurate and energy-efficient on-device distraction detection techniques will be developed by employing multi-sensor and heterogeneous data available from upcoming mobile and wearable devices. In this direction, supervised and semi-supervised learning will be used to design efficient activity classification and distraction prediction techniques which will be empirically evaluated using proof-of-concept implementations. Unlike existing stand-alone approaches, the proposed framework employs a connected-community approach to accurately capture the impact of both a pedestrian's own actions, as well as the actions of others, on his/her safety. This involves the design and implementation of a privacy-preserving and cloud-assisted data-analytics engine to capture, analyze and notify pedestrians of impending hazardous situations from the crowd-sourced distraction data obtained from participating users. Finally, a comprehensive performance and usability evaluation will be conducted by deploying a large-scale testbed involving participants from Wichita State University's (WSU) campus community. The project outcomes, including the planned testbed, will have a significant impact on improving pedestrian safety within the WSU campus community. If successful, similar trials at an urban or city-wide scale can also be envisioned. In addition to improving pedestrian safety, this project will educate users and participants on the impact of technology on pedestrian safety and its role in improving the same. Project outcomes and results will be disseminated by means of peer-reviewed publications, white papers and open-source applications. Applications and anonymous data collected from the planned testbed will be appropriately disseminated to facilitate additional research and advances in the area of pedestrian safety technology.