CS 494: Edge Computing Systems

Edge computing brings computation out of the cloud and onto small, distributed devices physically located where data is generated. Students will design and implement complete edge pipelines that integrate embedded processors, sensors, communication protocols, and on-device machine learning.

Course Overview

This course blends multiple computer science disciplines:

  • Embedded Systems: Programming resource-constrained devices
  • Networking: Low-power communication protocols (MQTT, BLE, LoRa)
  • Systems Design: Real-time processing and deployment strategies
  • Machine Learning: TinyML and TensorFlow Lite inference at the edge
  • Software Engineering: GitHub-based collaboration and documentation

The course is entirely project-based, featuring guest speakers from both academia and industry. Teams will build and field-test working systems addressing real-world constraints such as latency, privacy, and power efficiency.

Course Material

There is no required textbook. All readings will be provided as PDFs or online resources from:

  • EdgeSys, TinyML Research Symposium, ACM IoT, various conference proceedings
  • Vendor documentation (e.g., Raspberry Pi, NVIDIA Jetson/Orin/Thor)
  • Open-source TinyML and TensorFlow Lite tutorials

Grading

  • 25% Homework (Preparation Assignments)
  • 20% Milestone 1 (Prototype)
  • 20% Milestone 2 (Field Test)
  • 30% Final Project (Demo, Repository, Documentation)
  • 5% Participation and Peer Review

Schedule

Week Topic
1 Introduction to Edge Computing

Course overview, edge vs. cloud vs. HPC, ecosystem landscape.
2 Edge Hardware and Embedded Development

Raspberry Pi, NVIDIA Jetson/Orin/Thor, resource-constrained devices.
3 Sensor Integration and Data Acquisition

Interfacing sensors, data collection pipelines, signal conditioning.
4 Edge Networking (MQTT, BLE, LoRa)

Low-power communication protocols, message brokers, pub/sub patterns.
5 Lightweight ML Inference

TinyML, TensorFlow Lite, on-device inference under constraints.
6 Power and Bandwidth Optimization

Energy budgets, duty cycling, compression, data reduction at the edge.
7 Milestone 1: Prototype

Teams present working prototypes for peer review.
8 Edge-Cloud Integration and APIs

Hybrid architectures, REST/gRPC, data synchronization strategies.
9 Containerization and Update Strategies

Docker on edge, OTA updates, fleet management.
10 Security, Privacy, and Ethics in Edge AI

Threat models, federated learning, differential privacy, ethical deployment.
11 Milestone 2: Field Test

Teams deploy systems and present field test results.
12 Performance Evaluation and Optimization

Latency profiling, throughput, reliability metrics.
13 Guest Speakers — Industry and Research

Practitioners from academia and industry share real-world edge deployments.
14 Final Project Demo and Peer Review

Teams demo working systems, peer evaluation, course wrap-up.