HAC-LSTM: Mobility-Aware Joint Service Placement and Task Offloading Scheme in MEC

Mobile Edge Computing (MEC) is a promising paradigm that deploys cloud-based services at the network edge to process computationally intensive tasks with low service delay. However, optimizing MEC systems remains challenging due to the random task arrival patterns of mobile edge devices and the necessity for dynamic service placement. Consequently, decisions regarding service placement significantly influence task offloading choices, and vice versa. This interdependence has been relatively overlooked, limiting the overall performance of MEC systems. To address this challenge, we formulate the cooperative service placement and task offloading problem as a Partially Observable Markov Decision Process (POMDP). We propose a simple yet efficient approach named HAC-LSTM, which integrates Long Short-Term Memory (LSTM) networks with a hybrid actor-critic (HAC) algorithm. HAC-LSTM employs an LSTM-based state encoder to effectively extract hidden information and manage partial future information of the MEC system. Subsequently, the HAC algorithm leverages hybrid actors—a discrete actor for service placement and a continuous actor for task offloading—to make optimized joint decisions. We evaluate our approach through extensive experiments using real-world mobility datasets and varying key parameters. The results demonstrate that HACLSTM outperforms baseline algorithms by significantly minimizing average delay, thereby enhancing the overall efficiency and responsiveness of MEC systems.