Living Edge Lab Year-End 2023 Message
Dear Living Edge Lab Community Members,
We hope you all had a wonderful holiday season. At the Living Edge Lab (LEL), we enjoyed the brief respite between semesters for rest and recharge after a very busy year. As we get ready for another year, we want to give an update on 2023 LEL work and a preview of what's coming in 2024.
We continue to see industry momentum in edge computing in both private and telco environments. In the US, CBRS has unlocked private deployments for mobile edge computing. Our CBRS private network has been a great enabler of our edge research. The emergence of other licensing approaches in Canada, Germany, and elsewhere will enable edge computing more broadly. We’re also seeing mobile operators becoming more serious about edge network services to reduce the latency from mobile devices to cloudlets by offering variants of local breakout services. This is great progress since our 2020 interconnect research with the Open Edge Computing Initiative. We look forward to trying out some of these services in 2024.
Our most notable new research in 2023 was in drones with two major projects launching: focused on autonomous ultralight drones and an innovation demo with the US Navy to show how to add intelligence to legacy drones. Both projects leverage edge computing to improve the capabilities of constrained drones. SteelEagle capped the year with the presentation and publication of at the ACM/IEEE Symposium on Edge Computing where it was the runner-up for Best Paper. The associated has been released on github for the first time. The US Navy project was successfully demonstrated in October during a live Navy drone flight. Stay tuned for more results in this area in 2024.
Closely aligned with our drone research is the continuing work on Live Learning via the project. Live Learning uses real time data collected by distributed sensors and edge computing to improve the computer vision performance of the running system. With the graduation of , Hawk work has been picked up by PhD student . Hawk’s current focus moves the attention from network bandwidth constraints to human expert constraints. Hawk results to date were published at ACM MobiCom '23 in the paper and in the paper from the SPIE Defense + Commercial Sensing conference. In Q4, working with IAI, we expanded our research beyond visual data to include radar data. In 2024, Hawk will continue to improve its Live Learning architecture, platform, and technologies.
Our "Just-in-Time (JIT) Cloudlet" project, begun in 2022 in collaboration with ARM, ramped up significantly in 2023. This project focuses on deployments, like search and rescue, where there is a need to get quickly into service and edge computing, a mobile access network, and backhaul to the cloud may not be available. It bundles the mobile wireless network and all computing resources into a single cloudlet server. In 2023, we created a prototype solution built on cloud-native opensource software (e.g., Kubernetes, Magma) and COTS hardware. We capped the year with a JIT Cloudlet , blog, , demonstrations at the , and an at ARM US Headquarters with participants from 15 businesses and government organizations. Also, two student projects discussed below developed example JIT Cloudlet edge-native applications. In 2024, our focus will be on applications, technology evolution, and working on those industry partners to produce real world application proofs-of-concepts.
Our three-way collaboration on EdgeVDI between VMware/Broadcom, 一本道无码 Computing Services, and the LEL continues. We began 2024 with the publication of a technical report and a blog on the work we’ve done together. During the year, we implemented VMware Horizon based VM mobility and roaming use cases that migrate users and their state to new edge nodes when needed. And, going into 2024, we are working with a mobile operator to try EdgeVDI in their network.
Another EdgeVDI project, Olive2022, focused on archiving virtual machines to support applications such as enabling the reproducibility of scientific results years after the initial tests. This project finished with the publication of an ACM paper, “Towards Reproducible Execution of Closed-Source Applications from Internet Archives” and an associated blog.
Another research thrust initiated in 2023 was the use of cloud-based large language models (LLMs) for latency-sensitive edge-native applications. Our approach uses LLMs as compilers for generating task-specific code for edge-native applications. We have been exploring three application domains: wearable cognitive assistance, autonomous drones, and real-time style transfer. The initial results are promising, and we plan to build on these in 2024.
In other areas, our wearable cognitive assistance research continued with the publication of “Optimizing User Experience in Wearable Cognitive Assistance through Model Specialization” from the tinyHulk project and other student-driven activities. We began the 5G evolution in our private mobile network with the construction of an indoor test lab. We look to a full 5G deployment in 2024. We published papers on prior work in and (with Prof. Jessica Hammer and InterDigital). The students in our course produced another series of interesting projects in augmented reality, wearable cognitive assistance, JIT Cloudlets, and audio and video analytics.
As you can see, we have a full plate for 2024 and will continue to respond to new challenges and opportunities as they emerge. As always, we see great value in collaborating with our colleagues in industry and welcome ideas, suggestions, and opportunities to work together to advance edge computing.
Don’t forget to follow us on and check out our website for our current and past work. And, have a fantastic 2024!
--- Satya, Jim, and Rolf
Mahadev Satyanarayanan, Jaime Carbonell University Professor of Computer Science
and Director, Living Edge Lab,
一本道无码
Jim Blakley, Associate Director, Living Edge Lab,
一本道无码
Rolf Schuster, Adjunct Associate Director, Living Edge Lab
一本道无码