Energy-Efficient Cloud-Edge Collaborative Computing: Joint Task Offloading, Resource Allocation, and Service Caching

Mobile Edge Computing (MEC) deploys computing and storage resources from cloud centers to edge servers, thereby meeting the latency requirements of emerging applications and lowering energy consumption. Due to the task diversity, computational services necessary for these tasks must be cached at the edge server to efficiently handle computational tasks. However, due to the constraints of resources in the edge server, strategically caching meaningful computing services becomes crucial. After investigating a three-tier network architecture composed of IoT Mobile Devices (IMDs), one edge server and the cloud server, we propose a cloud-edge collaborative computing strategy that jointly optimizes service caching, resource allocation for communication bandwidth and computing, and task offloading decisions. The goal is to achieve long-term minimization of energy consumption for IMDs while processing tasks under latency constraints. We propose an algorithm employing Twin Delayed Deep Deterministic Policy Gradient (TD3), an advanced reinforcement learning algorithm, to address this Mixed-Integer Non-Linear Programming (MINLP) problem. The simulation results demonstrate that the proposed algorithm surpasses baseline methods by significantly reducing task processing energy consumption across all IMDs, while ensuring compliance with latency constraints.