A robust two-stage edge cloud resource allocation under edge–edge collaboration

Edge cloud computing brings computational resources closer to end users, enabling rapid-response services for emerging applications. Although this paradigm mitigates the limitations of centralized architectures, two main challenges persist. First, accurately estimating regional service demands is complicated by both slow-evolving (long-term) and sudden (short-term) patterns. Second, individual nodes typically have limited capacity, rendering them vulnerable to demand surges. In this paper, we investigate an edge–edge collaboration mode where nearby nodes collectively share tasks to bolster system capacity and reliability. We propose a two-stage planning model that integrates strategic decisions (node location and capacity configuration) with operational implementation (service receiving and demand forwarding). By explicitly modeling both service-to-node and inter-node network delays, our approach captures realistic latency considerations in collaboration decisions. To handle uncertain demand, we develop an adaptive robust optimization model that sequentially applies a data-driven distributional approach to characterize long-term demand patterns, and a budget-based uncertainty set to capture short-term fluctuations. We devise an efficient nested column-and-constraint generation algorithm to solve the resulting mixed-integer problem with nonlinear interactions. Numerical experiments, grounded in real-world operating cases from a two-city deployment by an edge service provider, demonstrate that our approach achieves superior performance under various uncertainty scenarios compared to strategies reflective of current service provider practices. Our planning model provides a practical solution for robust and flexible edge cloud resource allocation.