Overview: This seminar-format class explores opportunities and challenges associated with edge computing, the diffusion of centralized cloud computing functionality to include resource-constrained systems in physical proximity to the users, such as cloudlets, mobile phones, and smart gateways.
The course surveys recent advances in edge computing and its role in enabling the next generation of the Internet of Things and the smart cities of the future. Students will learn the strengths and the limitations of edge computing systems, and will explore a range of algorithm and system adaptation techniques for developing edge-specific platforms, algorithms, and applications. Students will complete an individual or a team-based research project, theory-oriented or applied.
Lecture times: Mondays, Wednesdays 13:45 PM – 15:00 PM. This course involves interactive discussions of research papers and is unfortunately not suitable for asynchronous learning.
Professor office hours: Wednesdays 10:00 AM – noon, or by appointment
- Quizzes: 10%
- Research paper presentation: 25%
- Research project: 50%
- Participation in class discussions: 15%
A brief sampling of topics we will discuss:
- Evolution of distributed systems: from mainframes to cloud to edge
- Multi-tier distributed system architectures, in the Internet of Things and in the 5G vision
- Hardware and software edge computing platforms
- Taking machine learning out of datacenters:
- Federated learning
- Reinforcement learning on the edge
- Edge computing aiding advanced interactive applications:
- Edge-supported smart vehicles and drones
- Augmented and virtual reality
- Function allocation and resource management in multi-tier fog computing systems
- Security and privacy in edge computing
Research project information:
A research project in edge computing is an important component of this class; it corresponds to 50% of the overall course grade.
Research projects can be done individually or in groups of three. Projects can be theory-oriented or applied. Project teams need to be established by February 4th; the project proposal outlining the key ideas of the proposed work and the plan of action needs to be submitted by February 15th.
Students are particularly encouraged to come up with projects that can improve some specific element of Duke University student, staff, or visitor experience. Bonus points will be given for connecting the project to Duke Blue Devils, Duke Lemur Center, or other specific and unique element of life at Duke. Bonus points will also be given for projects that aim to address the unique challenges of campus life under COVID-19 restrictions.
- Federated learning, by Yichen (Ethan) Ruan, Carnegie Mellon University
- Edge computing infrastructure, by Dr. Gopika Premsankar, University of Western Ontario
- Mobile visual computing, by Dr. Yuhao Zhu, University of Rochester
- Trustworthiness in ground support networks for high-scale UAV operations, by Chuck Byers, Valqari and the Industrial Internet Consortium.
Research papers covered in this class’s student presentations:
- Distributed Deep Neural Networks over the Cloud, the Edge and End Devices, IEEE ICDCS’17 [ web link ]
- An Empirical Study of Latency in an Emerging Class of Edge Computing Applications for Wearable Cognitive Assistance, ACM Symposium on Edge Computing’17 [ web link ]
- Real-Time Video Analytics: The Killer App for Edge Computing, IEEE Computer, 2017 [ web link ]
- Dependability in Edge Computing, Communications of the ACM, 2020 [ web link ]
- PMF: A Privacy-preserving Human Mobility Prediction Framework via Federated Learning, ACM IMWUT’20 [ web link ]
- Edge-SLAM: Edge-assisted Visual Simultaneous Localization and Mapping, ACM MobiSys’20 [ web link ]
- Cloud Computing and Grid Computing 360-Degree Compared, 2008 Grid Computing Environments Workshop [ web link ]
- DeltaVR: Achieving High-performance Mobile VR Dynamics through Pixel Reuse, IEEE IPSN’19 [ web link ]
- Think Locally, Act Globally: Federated Learning with Local and Global Representations, 2020 [ web link ]
- Overcoming Noisy and Irrelevant Data in Federated Learning, 2020 [ web link ]
- SDN Controller Placement at the Edge: Optimizing Delay and Overheads, IEEE INFOCOM’18 [ web link ]
- Communication-Efficient Learning of Deep Networks from Decentralized Data, 2017 [ web link ]
- GROOT: A Real-time Streaming System of High-fidelity Volumetric Videos, ACM MobiCom’20 [ web link ]
- Orbital Edge Computing: Nanosatellite Constellations as a New Class of Computer System, ACM ASPLOS’20 [ web link ]
- DeepDecision: A Mobile Deep Learning Framework for Edge Video Analytics, IEEE INFOCOM’18 [ web link ]
- Learning Context-Aware Policies from Multiple Smart Homes via Federated Multi-Task Learning, ACM/IEEE IoTDI’20 [ web link ]