- Benefits of edge computing: latency, b/w, energy, data safety and privacy
- Challenges: programmability, naming, data abstraction, service management, security and privacy, optimization metrics
- Fog computing benefits: mobility support, geo-distribution, location-awareness, and low latency
- Applications:
- AR and real-time video analytics
- content delivery and caching
- mobile big data analytics
- Issues:
- fog networking with heterogeneous devices
- QoS(connectivity, reliability, capacity, delay)
- programmability
- computation offloading (dynamic nodes, dynamic wireless network, dynamic resources in the fog)
- accouting and monitoring
- provisioning and management
- security and privacy
A Brief History of Cloud Offload: A Personal Journey from Odyssey Through Cyber Foraging to Cloudlets
- Divide processing between edge and the cloud
- Key frame selection among a cluster of cameras(Early discard)
- Content-aware traffic scheduling: new metric of ops(object per second)
- Hybrid surveillance-access network: use edge nodes as access points to counter-costs
- Key frame detection, called Trigger frame in the work
- Active Cache: use tracking on the mobile device and oject detection on a backend server to achieve real-time object recognition
- Use Google Glass to detect volume and speaking rates in public speaking to provide feedback in real-time
- Find users more pleased with sparse delivery of recommendation of words than continuous streams of information presented with bars and plots.
- Use a computational offload approach as well. Audio is transmitted and processed on a backend server.
- Remove sensitive information from camera feeds with user-drawn and gesture-based region
- Doesn't use computational offload
- HMD VR that eschews all real-time scene rendering
- Aggresively precomputes and caches all possible images
Personalized, Wearable Control of a Head-mounted Display for Users with Upper Body Motor Impairments (CHI)
- Investigated how to use wearables to provide unobtrusive, always-available, and glanceable vocabulary support for people with aphasia
- Used Google glass and computational offload to identify # of people, age, and gender distribution to help the visually impaired do small talk.
Sensor-assisted face recognition system on smart glass via multi-view sparse representation classification (IPSN)
- On-device face recognition using both the camera and IMUs
- Analog convolutional image sensor that performs layers of a convolutional neural network in the analog domain before quantization
ZOE: A Cloud-less Dialog-enabled Continuous Sensing Wearable Exploiting Heterogeneous Computation (MobiSys)
- A custom-built wearable device that supports deep sensor inferences on continuously sensed data
- A speech dialog system to interface with user
- Without cloud or smartphone support
- Heterogeneous platform to realize a three-tier low-power processor hierarchy