Twin Delayed DDPG (TD3)-Based Edge Server Selection for 5G-Enabled Industrial and C-ITS Applications

In the 3GPP-defined edge computing architecture, User Equipment (UE) interacts with the Edge Enabler Layer to request application services. The Edge Enabler Client (EEC) forwards these requests to the Edge Enabler Server (EES), which leverages the 3GPP core network for UE analytics, such as location and Quality-of-Service (QoS) requirements, to identify the optimal Edge Application Server (EAS). The Edge Configuration Server (ECS) then provisions the configurations for connecting the UE to the selected EAS. While this architecture provides a robust framework, finding the most appropriate EAS for task execution remains a critical challenge in heterogeneous environments where Cooperative Intelligent Transportation Systems (C-ITS) and industrial 5G applications coexist. The diverse nature of applications, such as latency-critical safety tasks in C-ITS and throughput-intensive industrial applications like real-time robotics, makes it essential to intelligently map tasks to edge servers. Additionally, uneven task distribution and server overloading necessitate mechanisms for dynamic load sharing between edge servers in the Edge Data Network (EDN). This paper proposes an AI-driven framework for optimal edge server selection and collaborative load management in 5G-enabled C-ITS and industrial environments. Twin Delayed Deep Deterministic Gradient Policy (TD3) based framework is proposed to dynamically assign tasks to the most suitable EAS while considering constraints such as low latency, high throughput, computation resources, and bandwidth resources. Furthermore, it introduces an edge node-sharing mechanism to offload tasks from overloaded servers to neighboring nodes, ensuring balanced resource utilization and seamless service delivery.