SDD: Spectral Clustering and Double Deep Q-Network Based Edge Server Deployment Strategy

Xia OuHongbing Cheng^*

• zhejiang university of technology, 浙江 杭州, China

With the rapid development of 5G technology, Mobile Edge Computing (MEC) has become a critical component of next-generation network infrastructures. The efficient deployment of edge servers (ESs) plays a vital role in enhancing the quality of service (QoS). Although various deployment strategies have been proposed, most existing methods struggle in large-scale and high-density scenarios such as smart cities. In these settings, user distributions are highly heterogeneous and task loads vary significantly, which makes maintaining adaptability and optimal performance difficult. To address this problem, we propose a Spectral Clustering and double deep Q-Network-based edge server deployment method (SDD). Our goals are to reduce service delay and improve system load balance. First, we use spectral clustering to extract spatial features such as base station locations. Based on node similarity, we divide base stations into several clusters and determine candidate deployment centers. Then, we build a reinforcement learning environment guided by the clustering structure. A double deep Q-Network (DDQN) framework is used to jointly optimize server deployment and task load distribution. This process improves the overall deployment efficiency and service quality. We conduct large-scale experiments using a real base station dataset from the Shanghai Telecom Bureau. Our method is compared with several baselines, including Random, Improved Top-K, K-means, and ESL. The results show that our method performs better in both delay reduction and load balancing.