High-performance wireless perception using deep learning and mems devices

Guan, Junfeng

Permalink

https://hdl.handle.net/2142/116231 Copy

Description

Title

High-performance wireless perception using deep learning and mems devices

Author(s)

Guan, Junfeng

Issue Date

2022-07-15

Director of Research (if dissertation) or Advisor (if thesis)

Al-Hassanieh, Haitham

Doctoral Committee Chair(s)

Al-Hassanieh, Haitham

Committee Member(s)

• Roy Choudhury, Romit
• Patel, Sanjay
• Valdes Garcia, Alberto

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Wireless Perception
• Radar Imaging
• Joint Communication and Sensing
• Deep Learning
• MEMS

Abstract

Recent years have witnessed much interest in expanding the use of wireless networks beyond their traditional use for communications to providing new perception solutions, such as sensing, imaging, and localization. The vision is, the wireless perception functionalities of the next generation wireless networks are going to create a digital twin of the physical world. This thesis introduces new software and hardware primitives that advance wireless technologies towards achieving the vision of ubiquitous perception in next-generation wireless networks. The software primitive we introduce is AI-enhanced wireless imaging, where we leverage recent advances in deep neural networks to extract the underlying perceptual and contextual information of the environment from raw wireless images. We demonstrate the applications of AI-enhanced wireless imaging in self-driving car perception, where we develop systems to achieve millimeter-wave radar-based high-resolution imaging and accurate object detection. The hardware primitive we introduce is the first of its kind Micro-Electro-Mechanical System (MEMS) filter hardware, which we leverage to enable joint communication and high-performance sensing in next-generation wireless networks. Towards this end, we first present a spectrum sensing scheme that can efficiently sense wideband spectra with high time resolution. This system can be used to enable dynamic spectrum sharing between perception and communication services in future wireless networks for them to coexist. We also exploit reusing communication signals for perception. We develop an accurate Internet-of-Things (IoT) self-localization system that simply overhears ambient 5G communications signals without any coordination with the base stations in 5G cellular networks.

Graduation Semester

2022-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/116231

Copyright and License Information

Copyright 2022 Junfeng Guan

Owning Collections