ICG-Restore: Intent-Constrained, Graph-Enhanced LLM Planning with Minimal-Edit Repair for Post-Disaster Emergency Communication Recovery

Post-disaster emergency communication recovery is not merely a link-repair task but a high-level planning problem constrained by service priorities, inter-object dependencies, resource budgets, and time windows. Existing restoration optimization methods generally rely on fully structured inputs, whereas direct large language model (LLM) planning often produces fluent yet operationally infeasible plans due to missing prerequisites, stage-order conflicts, and budget violations. To address this challenge, we propose ICG-Restore, an intent-constrained, graph-enhanced LLM planning framework with rule-consistent minimal-edit repair. The framework compiles natural-language requests, structured network observations, and operational rules into a task-intent object, retrieves task-relevant local context from a heterogeneous scenario graph and a restoration knowledge graph, generates stage-wise restoration candidates, and repairs infeasible plans through bounded edits that preserve the original planning backbone. Feasible candidates are then evaluated and ranked by a safety-aware agent executor in an abstract restoration action space. Experiments on three topology scales, four restoration tasks, and five environmental evolution modes demonstrate that ICG-Restore consistently improves executability, critical-target coverage, and overall recovery quality. Compared with Direct-LLM, it improves CSR, WCTC@5, and CRS by 1.99%, 38.87%, and 24.56%, respectively.