CS8803 - Mobile Computing and IoT

This course will teach a variety of ideas, concepts, techniques, and algorithms, that are all crucial to understanding and developing mobile systems and applications. The course will begin from first principles and ramp up to real-world systems and technologies. Keywords related to this course includes: wireless sensing, localization, GPS, drones, sensors, motion tracking, acoustics, location privacy, etc.

We will start from absolute basics, and cover important modules from linear algebra and data/signal processing. We will assume that you do not recall anything from prior courses (even if you have taken them). We will discuss the crossover of ideas from these modules to wireless communication and mobile computing. The course is designed with CS students in mind, particularly those inclined towards systems and networking.

Understanding GPS, understanding why indoor positioning still not available ... Location fingerprinting (WiFi, magnetic, BLE), crowd-sourcing, mapping. Unsupervised data-driven learning, clustering, sensor fusion, filtering, simultaneous localization and mapping (SLAM).

Understanding IMU (accelerometer, gyroscope, compass) Can a smartwatch track human gestures and activities? Can embedded IMUs track the motion of a fast-moving baseball? Motion models and filtering techniques

Next generation of wearable devices: finger rings, ear-buds, smart clothing. Rings: Vibration and ultrasound, receiving vibrations (with IMU and microphone), body-channels. Hearables and earables: noise cancellation, bone conduction, motion to speech recovery, binaural sounds, energy optimization (wake-on-speech)

Core challenges in autonomous systems: sensing, computing, communications + actuation. Robotic wireless networks, 5G networks, cell-tower on flying drones, ray-tracing, channel optimization. Cars: LIDAR, RADAR, and vision, sensor fusion, relative map creation.

Why personal, always-ON devices are a major challenge in security and privacy Side channel attacks, inference algorithms, hardware loopholes, sensor data leaks. Case studies: location privacy, password typing, Alexa attacks, IMU fingerprints, acoustic drone attack, clock leaks, etc.

The underlying concepts discussed in this course are taught in class. Students are expected to read research papers about the topic, answer questions about the research paper before class, and engage in classroom discussions. We will be discussing those papers, and students are expected to participate in the class with comments and thoughts about the work. The class is interactive and we make the classroom a fun place. In-person attendance is expected. We will go on field trips in and around campus and talk about mobile computing and IoT technology is different contexts.

This course has homeworks, programming assignments, a midterm, and a final project (no final exam). The project is heavy, starts at the beginning of the class, and students have previously published their works as an outcome of this course (though that is not needed). Examples: Intrusense, Pettrack, Notification Control Demo, uFindMe. See papers or slides on: https://faculty.cc.gatech.edu/~dhekne/Links to an external site.

This course is NOT EASY. In previous years students have worked very hard to score well.

We expect to be done on the final exam day for this class (Dec 5).

Lecture Recordings, Live-stream, Image Release

Location

The class will be held at Sheller 222 every Monday and Wednesday from 2:00pm to 3:15pm location.

About Instructor

Name: Ashutosh Dhekne.
Please call me: Ashutosh or Prof. Dhekne
Pronouns: He/Him/His

Grading (Tentative)

About Generative AI, ChatGPT, etc.

Have you been to the Atlanta Airport? There are horizontal conveyer belts to "walk" in long corridors. Think of AI as those conveyer belts. If you just stand on them (use AI without using your intellect), you will fall behind those who walk on the floor besides you (those who do not use AI but understand the subject). If you want to get ahead, walk on the conveyer belt (use AI to aid your own intellect). That is the principle we use in this class. It is perfectly okay to use ChatGPT to understand a topic better, to ask questions, and gain insights. Use it as if it was a second teacher.
Remember: Learning this subject well will help you for life. The aim is not just to get a good grade in the class, but rather to develop a deep understanding of the subject.
We are piloting an innovative strategy for AI use in this class. I have drafted a list of AI-levels called Guided AI Use Framework (GAUF) which defines various ways in which you could use AI in general. In every assignment, and sometimes parts of the assignment I will make the expected GAUF level clear so that you can remain compliant and not have doubts about your AI usage. GAUF levels could apply to code, written text, proofs, answers to MCQs, etc. Look for GAUF tags in the assignment text. 

Layer	Name	What it Allows	Student Behavior Guidance	Instructor Guidance
L0	No AI	No AI tools at any stage	Think and write/solve/derive/code entirely on your own. Use of traditional tools such as textbooks, online search, asking a TA, etc. is acceptable.	Consider handholding in class or during office hours. Appropriate for all levels of the Bloom’s taxonomy. Great for traditional exams.
L1	AI for Ideation	Use AI for brainstorming only after you have provided initial ideas and seed thoughts, not for final outputs	Use to generate questions, not answers. “create a short quiz on the use of IMUs in mobile computing.” Okay to use AI to point you to good online or offline resources (books)	Encourage use of AI as a second teacher or an additional TA, but not replacement for effort.
L2	AI for Review	Use AI to review your own draft, not to generate it	Write/solve/derive/code yourself first, then optionally refine through AI inputs. E.g.: check if your point comes across clearly by asking AI to summarize your content. Do not copy-paste refined text/code. Only take inspiration from AI suggestions.	Encourage agency and responsibility. Students likely to feel higher level of control and pride in their own work. Allows students to get un-stuck quickly and to expand or make text concise/precise.
L3	AI for Drafting	Use AI to help with drafts, you must revise it critically	You remain the primary author and are responsible for what is presented. You must understand every line (of written text/proof/solution/code) deeply. You must be able to orally answer questions like “why is there a negative sign here?”	Reduces frustration for students struggling to code, and overcomes mind blocks. Deep understanding is expected but difficult to assess. Works best for tasks that may feel mundane to students. E.g., in a class where data-structures are used but not learned for the first time, using AI for creating a linked list is okay, because students are expected to already know how to do that themselves.
L4	AI Unleashed, Responsibly	Freely use AI throughout, but cite tool and version	Use AI to generate entire modules or paragraphs of text. You need to understand what is being done, but at a high level. It is okay to not have answers to “how” and “why” questions which tend to be deeper.	Use with caution! This might seem like making students “job-ready”, but risks allowing students to graduate while remaining very shallow on their understanding of the subject. Can be used for improving content in a dimension orthogonal to the subject at hand. E.g.: making a slide deck’s colors consistent for a student not majoring in design is a task that AI could handle. The student need not know the design aspects of white-space and theory around color palettes.

How to prompt LLMs to use this framework?

Before seeking any help from your favorite LLMs, copy-paste the following prompt. Always annotate your questions the same way as they are in the assignments:

My professor has defined a Guided AI Use Framework (GAUF) for this course. I want to remain compliant with GAUF. So when I ask you questions from assignments, I will tag the questions with their GAUF level as well. Only provide me answers to the extent allowed by the GAUF framework. Here is the full text of the GAUF framework: 

GAUF-Level | Name | What it Allows | Student Behavior Guidance | Instructor Guidance | LLM Guidance
GAUF-L0 | No AI | No AI tools at any stage | Think and write/solve/derive/code entirely on your own. Use of traditional tools such as textbooks, online search, asking a TA, etc. is acceptable. | Consider handholding in class or during office hours. Appropriate for all levels of the Bloom’s taxonomy. Great for traditional exams. | Respond: “This is GAUF-L0. I cannot provide any assistance, as no AI use is allowed.” Do not give hints, brainstorming, or answers.
GAUF-L1 | AI for Ideation | Use AI for brainstorming only after you have provided initial ideas and seed thoughts, not for final outputsUse to generate questions, not answers. “create a short quiz on the use of IMUs in mobile computing.” Okay to use AI to point you to good online or offline resources (books) | Encourage use of AI as a second teacher or an additional TA, but not replacement for effort. | If no initial ideas are provided, ask the student for them first. Only after receiving them, give brainstorming suggestions, resource pointers, and possible directions — not a final answer. Avoid polished responses.
GAUF-L2 | AI for Review | Use AI to review your own draft, not to generate it | Write/solve/derive/code yourself first, then optionally refine through AI inputs. E.g.: check if your point comes across clearly by asking AI to summarize your content. Do not copy-paste refined text/code. Only take inspiration from AI suggestions. | Encourage agency and responsibility. Students likely to feel higher level of control and pride in their own work. Allows students to get un-stuck quickly and to expand or make text concise/precise. | Ask the student to paste their draft. Give review comments, suggestions for clarity, completeness, or style. Avoid writing a fresh solution — focus on improving their work.
GAUF-L3 | AI for Drafting | Use AI to help with drafts, you must revise it critically| You remain the primary author and are responsible for what is presented. You must understand every line (of written text/proof/solution/code) deeply. You must be able to orally answer questions like “why is there a negative sign here?” | Reduces frustration for students struggling to code, and overcomes mind blocks. Deep understanding is expected but difficult to assess. Works best for tasks that may feel mundane to students. E.g., in a class where data-structures are used but not learned for the first time, using AI for creating a linked list is okay, because students are expected to already know how to do that themselves. | Provide a draft solution, but include explicit prompts for the student to check, revise, and explain each step. Do not present it as final — mark it as a starting point.
GAUF-L4 | AI Unleashed, Responsibly | Freely use AI throughout, but cite tool and version | Use AI to generate entire modules or paragraphs of text. You need to understand what is being done, but at a high level. It is okay to not have answers to “how” and “why” questions which tend to be deeper. | Use with caution! This might seem like making students “job-ready”, but risks allowing students to graduate while remaining very shallow on their understanding of the subject. Can be used for improving content in a dimension orthogonal to the subject at hand. E.g.: making a slide deck’s colors consistent for a student not majoring in design is an example of a task that AI could handle. The student need not know the design aspects of white-space and theory around color palettes. | Provide complete solutions as requested. Cite “Generated with [tool name & version, date]” in the output. Encourage minimal verification steps.

 

Integrity

You are here to learn. Participate with integrity. Cheating will not be tolerated. We will refer you to the Office of Student Integrity. 

Georgia Tech honor code for students:

I commit to uphold the ideals of honor and integrity by refusing to betray the trust bestowed upon me as a member of the Georgia Tech community.

Join the Microsoft Teams Team:

Team Name: CS8803-MCI-FA2025

Team Code: 10mudk0

Tentative Syllabus

Date	Topic	Material	Paper Reviews/ToDo/Deadlines
08/18/2025 (Mon)	Introduction		How to read a research paper? [link]
08/20/2025 (Wed)	Project Ideas	Classroom discussion of the semester wide projects. Pick one or bring your own project lides	Install the right programming environment for assignments
08/25/2025 (Mon)	Outdoor Localization	Basics of localization. GPS	Programming Assignment 1 - GPS
08/27/2025 (Wed)	Indoor Localization	RF localization using received signal strength, introduction to ranging	read RADAR paper [link]
09/01/2025 (Mon)	No Class	No Class  (Links to an exterDownload Android Code
09/03/2025 (Wed)	Ranging Basics	Discussion on ranging, demo with UWB, Static anchors, moving anchors, interchangeable tag-anchor roles ownload Matlab Sound CodDownload Matlab Signal Animation	PA1 Due, PA2 - UWB Ranging read Decawave tutorial [link] read TrackIO [link]
09/08/2025 (Mon)	TDoA	TDoA Basics, Highly scalable localization	PA3 - UWB Localization read PnPLoc [link] Project Title and Group Due
09/10/2025 (Wed)	Cisco Tour	We will meet in Coda
09/15/2025 (Mon)	Whiteboard Explanations	Some demos, free question-answer session on UWB localization	PA2 Due. Possibly Online!
09/17/2025 (Wed)	Linear Algebra Matrices, Active Passive ranging	Location spoofing	Possibly Online! read UnSpoof [link]
09/22/2025 (Mon)	IMU Sensors	Inertial sensors on mobile phones, demo	PA3 Due PA4 - IMU Orientation
09/24/2025 (Wed)	IMU Based Localization	UnLoc, Apps for step counter	read UnLoc [link] Onl
09/29/2025 (Mon)	IMU+UWB	Tracking a cricket ball. Demo on embedded IMU Download Slides1	read iBall
10/01/2025 (Wed)	On-body IMU and UWB	Visual signals recorded	read ViSig
10/6/2025 (Mon)	No Class	Use this free time wisely! -- Fall Break --	Working on project is a good idea  (Links to an external site.
10/8/2025 (Wed)	Project Progress Day	Discuss with the TAs. Not optional	Project Check-in has associated points.
10/13/2025 (Mon)	Signals	wnload SlidesLecture Video	PA4 Due PA5 - Signal Processing Casual    :
10/15/2025 (Wed)	Signals
10/20/2025 (Mon)	Signals		read Foresight [link]
10/22/2025 (Wed)	Material Identification	LiquID, other follow-up works	read LiquID [link] PA 6 - Wireless sensing
10/27/2025 (Mon)	X-ray vision	See through wall, X-ray vision for Augmented Reality	read WiVi [link] read X-AR [link]
10/29/2025 (Wed)	Wireless Communication	Location Based Caching	PA5 Due read Foresight [link]
11/03/2025 (Mon)	Authentication	2FA using UWB, Stealing passwords using a smartwatch	read UWB-Auth [link]
11/05/2025 (Wed)	Acoustics	Side Channel Attacks	Read: Backdoor [link]
11/10/2024 (Mon)	Impact Far and Wide	How the under-developed and developing world benefits from mobile computing. Slow internet, improvised technology JavaRosa, IIAB, LittleBeats	PA6 Due
11/12/2024 (Wed)	Batteryless Computing	Full-stack level thinking for a new paradigm	Read Protean [link]
11/17/2024 (Mon)	Underwater Computing	Challenges, Current State of the Art	Read paper [link]
11/19/2024 (Wed)	Industrial IoT	Introduction to Zigbee, sensors, actuators, and use-cases
11/24/2024 (Mon)	Sustainability and Mobile Computing	Environmental Impact, Solutions through Mobile Computing, Solutions through IoT
11/26/2024 (Wed)	No Class	No Class ---Thanksgiving Break ---
12/01/2025 (Mon)	Kendeda Building Tour	Meet at Kendeda Building

 

Office Hours

Rahul Bulusu (TA)	TBD	Klaus (Research Wing) 3124
Patherya, Kausar (TA)	TBD	Klaus (Research Wing) 3124
Ashutosh Dhekne (Instructor)	In person and online - Wed 9 am	Klaus (Research Wing) 3312