Utilizing Deep Reinforcement Learning for Task Offloading and Resource Management in MEC-Enabled 6G Networks

Abstract

Mobile edge computing (MEC) is an important emerging technology in recent years that offers energyconsuming services to mobile operators, who need a lot of storage and computing capabilities. Nevertheless, user devices (UDs) usually have limited resources, and satisfying such requirements can be problematic. In this work, the issue of computationally intensive services and the scarce resources of UDs are mentioned, and a strategy of partial task offloading and partial resource allocation is proposed to maximize job completion within reasonable time limits and minimize energy consumption. The approach uses the deep reinforcement learning algorithm, namely the deep deterministic policy gradient (DDPG) method, to define the optimization task as a Markov decision process (MDP). The performance outcomes show that the suggested method can greatly decrease energy consumption and raise the job completion rate compared to conventional techniques. Our approach minimizes overall energy consumption while maximizing job completion efficiency. Simulation results reveal that the proposed method reduces energy usage by approximately 84.96% compared to traditional approaches and enables a task completion rate of 95.9% under optimal conditions with a 16 GHz link bandwidth. Additionally, our method achieves a 42.08% increase in overall rewards. It outperforms existing strategies in energy efficiency and task execution speed, demonstrating its effectiveness in managing resources in dynamic 6G environments. Finally, this contribution proves that it is possible to efficiently conduct resource management in MEC-enabled 6G networks with the help of DDPG.