Collaborative Research: PPoSS: Planning: Scaling Autonomous Vehicle Systems at the Edge: from On-Board Processing to Cloud Infrastructure

该项目的重点是联网和自动驾驶汽车（cav）以及支持其运营的智慧城市基础设施。当前的自动驾驶汽车和智慧城市基础设施系统不能很好地随着应用程序数量的增加而扩展，并且被从嵌入式、路边和边缘设备收集的大量数据所淹没。随着越来越多的基础设施和车辆传感器开始收集数据，需要新的技术和方法来提高这些系统的可扩展性。该项目的新颖之处在于原则、抽象和方法的发展，这些原则、抽象和方法用于设计和实现可扩展的自动驾驶汽车系统，以及支持自动驾驶汽车运行的智慧城市基础设施。该项目的影响在于开发和部署自动驾驶汽车，这将带来更安全、更清洁、更高效的交通方式。该项目开发：(1)理论模型、框架和软件库，以支持异构CAV平台上可扩展并行算法的设计和实现；(2)一个高度可扩展的系统，用于将自动驾驶汽车应用程序随机卸载到云/边缘，该系统将执行自动驾驶汽车上的混合临界调度和任务卸载选择，边缘性能感知车辆路径规划，以及多跳安全和私有卸载；(3)一个可扩展的、安全的实时协同检测系统，在这个系统中，自动驾驶汽车可以利用来自其车载传感器和邻近车辆的传感数据；(4)基于细粒度感知交通数据的可扩展、自适应交通信号和自动驾驶汽车轨迹协调协议，同时解决数据量增加带来的计算需求增加；(5)编程框架，包括库和接口，以促进自动驾驶汽车及其支持的智慧城市基础设施的可扩展应用程序的开发。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。

The focus of this project is on Connected and Autonomous Vehicles (CAVs) and the smart city infrastructure supporting their operation. Current CAV and smart city infrastructure systems do not scale well with the increasing number of applications and are overwhelmed with massive amounts of data collected from embedded, roadside, and edge devices. As more infrastructure and vehicle sensors begin to collect data, novel techniques and methodologies are required to improve the scalability of these systems. The project’s novelties are developments of principles, abstractions, and methodologies for the design and implementation of scalable systems for CAVs and the smart city infrastructure supporting the operation of CAVs. The project's impacts are in the development and deployment of CAVs which will lead to a safer, cleaner, and more efficient transportation. This project develops: (1) theoretical models, frameworks, and software libraries to support the design and implementation of scalable parallel algorithms on heterogeneous CAV platforms; (2) a highly-scalable system for opportunistic offloading of CAV applications to the cloud/edge that will perform on-CAV mixed-criticality scheduling and task-offloading selection, edge-performance-aware vehicle path planning, and multi-hop secure and private offloading; (3) a scalable and secure real-time collaborative detection system in which CAVs leverage sensing data from their on-board sensors and neighboring vehicles; (4) scalable, adaptive traffic-signal and CAV-trajectory coordination protocols via fine-grained sensing of traffic data while addressing the increased computation demands of increased volumes of data; and (5) a programming framework, including libraries and interfaces, that facilitates the development of scalable applications for CAVs and their supporting smart city infrastructure.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.