The School of Computer Science is pleased to present…
Hybrid POMDP for Mobile ITS with Co-optimized Pedagogical Effectiveness and Mobile Performance
MSc Thesis Proposal by: Vimanga Umange
Date: May 4th 2026
Time: 3PM – 5PM
Location: Teams Meeting
Abstract:
Mobile intelligent tutoring systems (ITS) aim to deliver personalized, engaging learning experiences, yet real-time adaptive avatar-based tutors remain constrained by limited computational resources, latency, and network variability on mobile devices. Existing approaches often optimize either pedagogical effectiveness or system performance in isolation, reducing their practicality in dynamic mobile environments. This research proposes a hybrid cloud–edge adaptive tutoring framework that jointly optimizes instructional quality and system efficiency. The approach models tutoring as a Partially Observable Markov Decision Process (POMDP), integrating learner state, contextual conditions, and device constraints to enable adaptive decisions such as computational offloading and interface selection (text, 2D, or 3D avatars). Unlike prior systems that decouple pedagogy from system-level adaptation, the proposed framework ensures that real-time performance optimizations preserve core instructional principles, including continuous student modeling and step-level feedback. The system is designed to operate using locally available and context-aware signals, improving scalability and deployment feasibility without reliance on heavy infrastructure.
Keywords: Intelligent Tutoring Systems, Mobile Learning, POMDP, QoL, QoE, Adaptive Avatars
Thesis Committee:
Internal Reader: Dr. Jessica Chen
External Reader: Dr. Muhammad Asaduzzaman
Advisor: Dr. Xiaobu Yuan
Proposal not open to public
