CAREER: Interactive and Explainable AI for Next-Gen Transportation Systems Management

这项学院早期职业发展(Career)补助金建立了一个综合研究和教育计划，以重新设计现有的人工智能(AI)算法，以促进关于整体交通问题的多任务学习，同时消除模型可解释性的瓶颈。该研究计划将产生一个开源平台，该平台托管快速、可扩展并能够在不同交通问题上执行多任务的模型。开发的平台有可能从根本上创造新的数据源，以加快在不同交通方式中安全引入自动驾驶和互联汽车技术。此外，还采用了探究式学习等教育战略，以解决劳动力发展方面的差距和系统失衡问题，这些问题继续使人们对有色人种长期持有的偏见观点永久化。还将通过开放源码软件开发、互动讲习班和课程开发创造独特的外联和教育机会。该项目的主要内容包括：i)研究新的边缘和分布式云计算方法，以开发一个专门为在大数据上部署人工智能算法而构建的海量并行数据平台；ii)研究用于部署多用途机器学习算法的新体系结构，这将使决策支持系统的开发成为可能，该系统具有通过语音、文本或网络交互回答问题并不断向人类专家学习的能力，使其能够随着时间的推移改进其性能；以及iii)开发一个框架，通过将基于规则的学习集成到多功能模型的构建块中，在人工智能生命周期的每个阶段(培训、测试、部署)注入可解释性-这将提供对黑盒人工智能算法的洞察和可见性，允许人类在人工智能过程中发挥积极作用，并在需要时纠正路线。在追求这些目标的过程中，这个职业项目从深度学习和自适应计算理论中获得灵感，以设计和部署端到端交通系统管理解决方案。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势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 (CAREER) grant establishes an integrated research and education plan to re-design existing artificial intelligence (AI) algorithms to facilitate multi-task learning on holistic transportation problems, while removing bottlenecks to model interpretability. The research plan will result in an open-source platform that hosts models that are fast, scalable and able to multi-task on different transportation problems. The platform developed has the potential to create fundamentally new sources of data to speed up the safe introduction of autonomous and connective vehicle technologies across different transportation modes. In addition, educational strategies such as inquiry-based learning are used to address gaps in workforce development and systemic imbalances that continue to perpetuate long-held deficit views of communities of color. Unique opportunities for outreach and education will also be created through open-source software development, interactive workshops, and curriculum development. The main thrusts of the project include the following: i) investigate novel edge and distributed cloud computing methods to develop a massive, parallel data platform purposefully built for deploying AI algorithms on big data; ii) investigate new architectures for deploying multipurpose machine learning algorithms, which will enable the development of a decision support system that is interactive and collaborative - with abilities to respond to questions via voice, text, or web interactions and continuously learn from a human expert to enable it to improve its performance over time; and iii) develop a framework for infusing explainability at every stage of the AI lifecycle (training, testing, deployment) by integrating rule-based learning into the building blocks of the multipurpose model – this will provide insight and visibility into black-box AI algorithms, allowing humans to play an active part of the AI process and course-correct when needed. In pursuing these objectives, this CAREER project draws inspiration from theories of deep learning and adaptive computing to design and deploy an end-to-end transportation systems management solution.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.