Player-Aware Resource Management in Cloud Gaming: A Reinforcement Learning Approach to Action Prediction and Network Slice Optimization

Pathan, Muzakkir (2025) Player-Aware Resource Management in Cloud Gaming: A Reinforcement Learning Approach to Action Prediction and Network Slice Optimization. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (954kB) | Preview

Abstract

Cloud gaming often experiences sudden latency spikes during fast in-game actions, where reactive scaling is too slow to help. This study adopts a predict-then-allocate strategy: an edge-resident LSTM predicts imminent combat and a cloud-resident Q-learning agent selects the network slice (Basic/Medium/Premium) within a three-tier edge–fog–cloud architecture. The system is evaluated offline and on AWS against common baselines (Always Premium, Always Medium and a simple threshold policy). The LSTM achieves ~95% offline test accuracy and 78.7% accuracy in AWS deployment, enabling proactive resource allocation decisions under realistic hardware constraints. Compared with Always-Premium, the agent reduces premium-slice usage by ~10% with comparable service quality. The target latency of <50 ms was not achieved, with observed latencies of ~400 ms primarily due to t2.micro instance limitations and inter-service communication overhead rather than algorithmic limitations. Overall, the results indicate that combining action prediction with RL-based allocation can lower cost while preserving experience, and they outline practical upgrades (instance class, co-location, production servers) to close the latency gap.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Makki, Ahmed UNSPECIFIED
Subjects:	T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning G Geography. Anthropology. Recreation > GV Recreation Leisure > Games and Amusements > Online Games
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	30 Mar 2026 13:17
Last Modified:	30 Mar 2026 13:17
URI:	https://norma.ncirl.ie/id/eprint/9252

Actions (login required)

View Item