Title: Adaptively Serving XR Workloads from Resource-constrained Edge  

Date: Monday, July 14, 2025
Time: 10:00 AM EST 
Location (Hybrid): KACB 3402 and MS Teams (link)  

Jin Heo
Ph.D. Candidate
School of Computer Science
College of Computing 
Georgia Institute of Technology

Committee: 
Dr. Ada Gavrilovska (Advisor) - School of Computer Science, Georgia Institute of Technology
Dr. Alexey Tumanov - School of Computer Science, Georgia Institute of Technology
Dr. Ashutosh Dhekne - School of Computer Science, Georgia Institute of Technology
Dr. Ketan Bhardwaj - School of Computer Science, Georgia Institute of Technology
Dr. Klara Nahrstedt - Siebel School of Computing and Data Science, University of Illinois Urbana-Champaign

Abstract: 
Extended reality (XR) applications demand high computational power for perception and 3D rendering, often exceeding the capabilities of modern mobile devices. Edge computing offers a solution by offloading intensive tasks to servers at the network's edge. However, the resource constraints of edge infrastructures present significant challenges in delivering high-quality, multi-user XR experiences. Current static workload distribution methods often prove suboptimal due to the diverse deployment environments of XR applications. Additionally, existing compression methods for emerging sensor data compromise the benefits of edge assistance by failing to consider real-time XR constraints.

To deliver high-quality, multi-user mobile XR experiences via edge computing, an XR serving system at the edge server must be adaptive. It needs to coordinate the distribution and optimization of XR workloads, alongside data transfer, using quality metrics that accurately capture end-user experiences. To address these challenges in edge-assisted XR, this thesis presents four key contributions: (1) FleXR, a system enabling flexible distribution of XR workloads, designed and built to provide runtime adaptation in workload distribution without requiring modifications to XR functionalities, which is essential for fully realizing the benefits of edge assistance across various deployment contexts; (2) FLiCR, a fast and lightweight LiDAR point cloud compression method specifically designed for enabling edge-assisted online perceptions; (3) Stimpack, a system that adaptively optimizes 3D graphics rendering qualities by considering both user-side visual quality and server-side rendering cost, preventing wasted computing resources and enhancing the edge server's capacity to serve multi-user rendering workloads; and (4) Morphis, a context-aware adaptive perception framework for serving XR, which achieves its goal by predicting the user-perceived effectiveness of perception results, dynamically adjusting service levels to improve resource utilization and to efficiently scale to multiple users.