Research on TD3-Based Offloading Strategies for Complex Tasks in MEC Systems

The rapid growth of smart medical devices, smart home devices, and vehicular equipment has significantly increased the demand for computing power in mobile devices. These devices handle large-scale data processing and complex computational tasks but face inherent limitations in resources and performance. Mobile Edge Computing (MEC) addresses these challenges by providing IT service environments and cloud computing capabilities at the network edge, reducing latency, enhancing network efficiency, and optimizing user experience. However, MEC faces limitations such as high deployment and maintenance costs, limited application types, and an unstable operating environment. This paper proposes a Twin Delayed Deep Deterministic Policy Gradient (TD3)-based MEC multi-objective hierarchical complex task offloading algorithm. The algorithm aims to adaptively allocate computational and network resources, reduce average service time, balance resource utilization, and minimize task delay rates. Key contributions include integrating MEC with SDN and deep reinforcement learning, detailed modeling of complex tasks using Directed Acyclic Graphs (DAG), and defining a hierarchical multi-objective reward function. This mechanism optimizes multiple objectives simultaneously, enhancing decision-making. Experiments in a simulated MEC environment show that the TD3-based algorithm outperforms existing methods in learning speed, average service time, task delay rate, and server load balancing. This study provides a novel solution for complex task offloading in MEC environments and offers valuable insights for future MEC system design.