ERI: Toward mmWave Vehicular Communication: A Multisensor Multimodal Deep Data Fusion Approach

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。美国每年有超过3.5万人死于高速公路，全球每年有125万人死于高速公路。要降低如此高的死亡率，需要各方面的努力。其中，网联和自动驾驶汽车（CAV）是唯一可以将这一数字降至接近零的解决方案。顾名思义，CAV涉及两个相互关联的概念：“连接”和“自动化”。近年来，“自动”车辆取得了重大进展，自动驾驶能力达到了5级（完全自动驾驶）。目前的自动控制以一种更孤立的方式实现，通过传感器、预训练的人工智能算法和车载计算机处理来实现。在不久的将来，随着更多测试和部署的自动驾驶汽车，迫切需要提供通过车对车（V2X）通信（如车对车（V2V）和车对基础设施（V2I））交换、传输和收集实时数据的能力，即“连接”的需求。本研究旨在将毫米波（mmWave）频段的高频信号集成到CAV中，从而提供高数据速率和低延迟的可靠车辆连接解决方案。该项目的成果将带来以下影响：1)利用现有自动驾驶汽车的传感数据的变革性无信号方法；2)以本科为主的机构（PUI）的前沿研究经验；3)整合研究和课程开发，为本科生和研究生提供顶点项目；4)为研究社区提供一个包含硬件、软件和数据集的开源平台。毫米波用于自动驾驶汽车面临着毫米波信号传播过程中的高衰减和移动性管理等诸多挑战。现有的解决方案必须启动导频信号来测量信道信息，然后计算出朝向接收端的最佳窄波束，以保证足够的信号功率。这个过程需要大量的开销和时间，因此不适合车载应用。最近的工作研究了集成多传感器输入的可能性，如激光雷达（光探测和测距）和摄像头，传统上用于实现自动驾驶功能，以促进毫米波通信。然而，先前的工作是建立在无线模拟器和3D建模软件上的，缺乏现场测试的测量数据，此外还有对现有设备无法获得的理想波束模式的假设。因此，这种方法在现实场景中的可行性在很大程度上是未知的。为了缩小差距，我们的项目目标是：1)为实验室（室内）和现场（室外）毫米波CAV通信测试开发一个低成本、实时、协作和同步的数据收集平台；2)开发一种先进的深度数据融合的无信号码本/波束选择算法，使基站和车载用户都能以极低的开销选择最佳波束对；3)通过详细的测试计划验证平台和算法的有效性。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 award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).More than 35,000 people die each year on U.S. highways, 1.25 million worldwide. Efforts from all aspects are needed to reduce such a high fatality rate. Among them, connected and automated vehicles (CAV) is the only solution that can bring the number to nearly 0. As the name suggests, CAV involves two interconnected concepts: “connected” and “automated”. Recent years have witnessed significant advances in “automated” vehicles, creating self-driving capability up to Level 5 (fully autonomous). Current autonomous control is implemented in a more isolated way, made possible by sensors, pre-trained AI algorithms, and on-board computer processing within individual vehicle. With more tested and deployed CAVs in the near future, there is a pressing need to provide capability to exchange, transmit, and collect real-time data via vehicle-to-everything (V2X) communications, such as vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I), i.e., the need for “connected”. This research seeks to integrate high frequency signals in millimeter wave (mmWave) band to CAV, thereby providing reliable vehicular connectivity solution with high data rate and low latency. Outcomes from this project will bring following impacts: 1) a transformative signal-less approach utilizing sensory data from existing CAVs; 2) cutting-edge research experience to a primarily undergraduate institution (PUI); 3) integration of research and curriculum development, capstone projects to both undergraduate and graduate students; 4) an open-source platform with hardware, software, and datasets to the research community. mmWave for CAV faces many challenges such as high attenuation during mmWave signal propagation and mobility management. Existing solutions have to initiate pilot signals to measure channel information, then calculate the best narrow beam towards the receiver end to guarantee sufficient signal power. This process takes significant overhead and time, hence not suitable for vehicular applications. Recent works have investigated the possibility to integrate inputs from multisensor, such as LiDAR (Light Detection and Ranging) and camera, traditionally for enabling autonomous driving capability, to facilitate mmWave communications. However, prior work is built from the wireless simulator and 3D modelling software, lacks measurement data from field tests, in addition with assumption on idealized beam patterns that are not available from existing devices. Hence feasibility of such an approach on real-world scenario is largely unknown. To close the gap, our goal for this project is to: 1) develop a low-cost, real-time, cooperative, and synchronized data collection platform for both lab (indoor) and field (outdoor) mmWave CAV communication tests; 2) develop a signal-less codebook/beam selection algorithm with advanced deep data fusion, such that both base station and vehicle user can choose best beam pairs with extremely low overhead; 3) validate the effectiveness of both platform and algorithm with detailed test plans.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.