Deep Q-Network-based Task Offloading for Multi-Tier Computing Networks
Xiangrui Yang, Yongpeng Wu, Yang Yang, Wenjun Zhang · 2024
With the performance metrics of user tasks become increasingly personalized, allocating network resources for real-time offloading to enhance user satisfaction poses a significant challenge. Unlike existing models, grounded in real-world scenarios, we propose a multi-tier computing network characterized by heterogeneous resources. Subsequently, we model the task offloading problem as a Markov Decision Process and propose a task offloading algorithm based on Deep Q-Network (DQN). The computing network can leverage task characteristics and performance metrics to offload tasks onto the most suitable computing nodes for parallel processing. Simulation results demonstrate that the proposed algorithm can better utilize network resources, particularly outperforming other algorithms in terms of user satisfaction ratio, storage usage, and system overhead.