18-848 Radio Frequency Machine Learning

Radio Frequency (RF) sensing is a critical component of modern autonomous systems. For example, mmWave radars are ubiquitous in modern vehicles and a key part of autonomous driving systems. Due to their robustness, compact form factor, low cost, and privacy-preserving nature, RF sensing systems are also increasingly used for many indoor and human sensing applications.

To unlock the full potential of these sensors, we naturally turn to machine learning (ML), which has revolutionized how we approach sensing and perception, particularly in the computer vision community. However, to date, there has been relatively little overlap between the RF and ML communities, with relatively limited work on sensing methodologies which integrate RF and machine learning at a more fundamental level. Modern machine learning paradigms such as large-scale foundation models have also yet to be fully explored for the RF domain.

Overview

This research-focused course will provide students with ML and/or RF background the knowledge and skills needed to effectively engage in cross-disciplinary collaborations which push the frontiers of Machine Learning for RF.

The course will feature traditional lectures covering key background, such as the fundamentals of modern FMCW radars and machine learning research methodologies for scaling up ML research for novel domains, as well as research lectures covering the current state-of-the art and guest lectures from academic researchers and industry insiders. Students will also gain hands-on experience working with wireless systems such as mmWave radars and spectrum analyzers, as well as large datasets collected from these systems to develop, train, and apply modern machine learning techniques such as self-supervised learning and scaled-up transformer models. Finally, students will complete a research project applying Machine Learning techniques to a RF system of their choice.

Prerequisites

This course is intended for graduate students and advanced undergraduate students. All students are expected to have some machine learning background (e.g., 10-202, 10-301, or 18-661), be familiar with programming in Python (e.g., to use ML frameworks such as Pytorch), and have at least undergraduate-level linear algebra and probability background.

In addition, students should either have further machine learning or wireless communications background:

• Machine Learning

  Any upper-division or graduate-level ML course, e.g., any 10-6XX or 10-7XX course.

• Wireless Systems

  Any wireless communications course, e.g., 18-750 or 18-452.

Grading

• Lab 1 (15%): Spectrum Analysis

  Capture, analysis, and ML on RF from a spectrum analyzer

• Lab 2 (15%): mmWave Radar

  Data collection, processing, and ML for mmWave radar

• Course Project (50%)

  Individually or in groups of up to 4

  Proposal + Presentation + Final Report

• Paper Presentations (20%)

  Students will read and present 1-2 papers of their choice

Schedule

Subject to change

The schedule may change dramatically, particularly with regards to the timing of guest lectures.

1	Introduction & Logistics
2-3	Wireless Communications A crash course in wireless communications, providing a high level overview for students without wireless background and a review for students with wireless background.
4	Wireless Sensing A crash course in passive wireless sensing
5-6	Intro to Radar Operating principles of radar, and an overview of the different types and applications of radar systems
7	FMCW Radar FMCW radar & FMCW radar data
8	Applied Machine Learning A crash course on modern applied machine learning, from small statistical models to scaled-up foundation models
9	ML for Perception State-of-the-art ML methodology for perception and computer vision
10	ML for RF Systems Some of the key challenges you'll encounter when working with RF data
11	Guest Lecture (TBD)
12	Foundation models Advanced Topics Seminar I — Overview Lecture
13-14	Foundation models Advanced Topics Seminar I — Student Presentations
15	Project Proposals
16	Guest Lecture (TBD)
17	Imaging and simulation Advanced Topics Seminar II — Overview Lecture
18	Guest Lecture (TBD)
19-20	Imaging and simulation Advanced Topics Seminar II — Student Presentations
21	Guest Lecture (TBD)
22	Spectrum sensing Advanced Topics Seminar III — Overview Lecture
23	Spectrum sensing Advanced Topics Seminar III — Student Presentations
24	Guest Lecture (TBD)
25	Guest Lecture (TBD)
26	Project Presentations