I am Zhuangdi Xu (Andy), PhD student
at Georgia Institute of Technology
doing research in Edge computing
and visual data analysis.

My current research focuses on two topics: 1) Smart Camera System at the Edge of the Network, 2) Explore Visual Database Management System and 3) TPU resource management at Edge. Other than computer science, I have broad interests including mathematics, biology and quantum computing. I am also a competitive bridge (card game) player.

• CV

Publications:

• Zhuangdi, Xu, Kakkar, Gaurav Tarlok, J. Arulraj and U. Ramachandran. "EVA: A Symbolic Approach to Accelerating Exploratory Video Analytics with Materialized Views." ACM SIGMOD/PODS International Conference on Management of Data. SIGMOD ’22 (2022)
• Xu, Zhuangdi, S. ShahHarshil and U. Ramachandran. “Coral-Pie: A Geo-Distributed Edge-compute Solution for Space-Time Vehicle Tracking.” In 21st International Middleware Conference Middleware ’20 (2020). pdf presentation
• Xu, Zhuangdi, Harshit Gupta and U. Ramachandran. “STTR: A System for Tracking All Vehicles All the Time At the Edge of the Network.” DEBS ’18 (2018). pdf
• Xu, Zhuangdi, Sayan Sinha, S. ShahHarshil and U. Ramachandran. “Space-Time Vehicle Tracking at the Edge of the Network.” Proceedings of the 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges (2019). pdf
• Gupta, Harshit, Zhuangdi Xu and U. Ramachandran. “DataFog: Towards a Holistic Data Management Platform for the IoT Age at the Network Edge.” HotEdge (2018). pdf

Project

Space-Time Tracking of Vehicles for Smart Camera Surveillance

Our application is targeted to track all vehicles across live camera streams in real time. The result is stored as trajectories of vehicles in a graph database for later queries such as “where did the suspicious vehicle come from and head to?” Learn more...

Repo: Github

Coral-Pie: A Geo-Distributed Edge-compute Solution for Space-Time Vehicle Tracking

As a continuation of our previous work, we build a space-time vehicle tracking system on realistic cost-effective edge devices (i.e., Raspberry PI and EdgeTPU). We explore the necessary pipeline architecture and horizontal communication management. Learn more in Middleware 2020 presentation

Repo: Github

Reuse in Visual Database Management System

Derived data from the same video (e.g., bounding boxes) can be useful for multiple applications. We explore the intermediate results materialization and reuse opportunities across queries in a visual data management system (i.e., think MySQL for videos) called EVA.

Repo: Github

TPU resource management at Edge

TPU resources are usually underutilized even in a low-cost edge cluster. To improve the resource utilization of a MicroEdge Cluster (built on Raspberry Pis and EdgeTPU), we design a mechanism for TPU multiplexing, and extend the Kubernetes scheduler to orchestrate TPU resources.