Enhanced DQN in Task Offloading Across Multi-Tier Computing Networks

Abstract

Effective task workload decisions can enhance the utilization of network and computing resources in edge computing, thereby reducing the timeout rate of time-sensitive tasks and decreasing the average task processing time. We introduced an abstracted multi-tier computing network environment that closely resembles real-world conditions compared to other studies. DQN, or Deep Q-Network, is a reinforcement learning algorithm that leverages deep neural networks to optimize decision-making in sequential decision tasks. We employed a decision-making strategy utilizing deep reinforcement learning, presenting an enhanced DQN model incorporating advanced techniques. Our validation demonstrated its superior performance compared to baseline strategies.