Scholarlink

<p>In the context of smart cities Internet of Things (IoT), enabling resilient and real-time urban services, such as intelligent transportation and public safety monitoring, depends on scalable data handling at the network edge. These services are increasingly developed using microservice architectures deployed within Mobile Edge Computing (MEC) environments. However, efficiently orchestrating these microservices to process vast streams of urban data under multiple real-world constraints is a critical challenge. Existing research often overlooks practical constraints like the energy consumption of edge nodes and service reliability, which are vital for sustainable and trustworthy urban operations. Existing microservice orchestration approaches in Mobile Edge Computing mainly focus on latency optimization, while energy consumption and service reliability are often insufficiently considered as strict constraints. This paper addresses the microservice orchestration problem to enable smart city applications. We construct a novel joint optimization model that, beyond traditional resource limits, introduces a total system energy consumption cap and an end-to-end service reliability floor as hard constraints, directly reflecting the operational demands of urban services. For latency analysis, crucial for real-time data processing, we employ Jackson’s open queuing network theory to accurately model request flows. To solve this complex multi-constraint optimization problem, we propose a solution based on the Soft Actor-Critic (SAC) deep reinforcement learning algorithm. By maximizing policy entropy, our method enhances exploration to find stable and efficient deployment and routing strategies within the complex solution space. We have designed a state space, action space, and a composite reward function tailored to guide the agent toward minimizing latency while satisfying all urban service requirements. Simulation results demonstrate that our proposed method significantly reduces response latency while strictly adhering to energy and reliability constraints, ensuring the scalable and robust enablement of smart city.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing edge microservices deployment and routing for smart city IoT using the soft actor-critic algorithm

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Hao Yang,
  • Xiaodong Zhang,
  • Zhipeng Li,
  • Yixuan Chen,
  • Li Chen,
  • Xiaoxia Kou

摘要

In the context of smart cities Internet of Things (IoT), enabling resilient and real-time urban services, such as intelligent transportation and public safety monitoring, depends on scalable data handling at the network edge. These services are increasingly developed using microservice architectures deployed within Mobile Edge Computing (MEC) environments. However, efficiently orchestrating these microservices to process vast streams of urban data under multiple real-world constraints is a critical challenge. Existing research often overlooks practical constraints like the energy consumption of edge nodes and service reliability, which are vital for sustainable and trustworthy urban operations. Existing microservice orchestration approaches in Mobile Edge Computing mainly focus on latency optimization, while energy consumption and service reliability are often insufficiently considered as strict constraints. This paper addresses the microservice orchestration problem to enable smart city applications. We construct a novel joint optimization model that, beyond traditional resource limits, introduces a total system energy consumption cap and an end-to-end service reliability floor as hard constraints, directly reflecting the operational demands of urban services. For latency analysis, crucial for real-time data processing, we employ Jackson’s open queuing network theory to accurately model request flows. To solve this complex multi-constraint optimization problem, we propose a solution based on the Soft Actor-Critic (SAC) deep reinforcement learning algorithm. By maximizing policy entropy, our method enhances exploration to find stable and efficient deployment and routing strategies within the complex solution space. We have designed a state space, action space, and a composite reward function tailored to guide the agent toward minimizing latency while satisfying all urban service requirements. Simulation results demonstrate that our proposed method significantly reduces response latency while strictly adhering to energy and reliability constraints, ensuring the scalable and robust enablement of smart city.