10–14 Nov 2025
Office of Grants and Research
Africa/Accra timezone

A GAME THEORY- REINFORCEMENT LEARNING APPROACH TO COMPUTING RESOURCE ALLOCATION AND SCHEDULING

12 Nov 2025, 11:30
15m
Office of Grants and Research

Office of Grants and Research

Oral Presentation Emerging Technologies, Artificial Intelligence, and Engineering Innovations

Speaker

Daniella Apeadu

Description

Task scheduling in computing faces the dual challenge of fair resource allocation and thermal imbalance, which can lead to overuse of some resources, overheating, throttling, and hardware degradation. Traditional schedulers, such as First-Come First-Served (FCFS), focus on throughput but ignore fairness in load distribution and temperature, while basic Reinforcement Learning (Rl) approaches aim for adaptive thermal balance yet do not prioritize balanced load distribution. To address these limitations, this thesis proposes a cooperative game–theoretic hybrid model that integrates the Nash Bargaining Solution with Reinforcement Learning to achieve Pareto-optimal task allocations that are both fair and thermally aware. The efficacy of the model is evaluated using the Black-Scholes workload and compared against FCFS and a pure RL scheduler. Experimental results demonstrate that the hybrid scheduler achieves lower average core temperatures, reduced thermal variance, and more balanced utilization than the baseline algorithms, offering improved long-term energy efficiency and hardware longevity without sacrificing fairness

Primary author

Presentation materials