摘要
This study introduces a novel bi-level defender–attacker model (BDAM) designed to address real-world homeland defense scenarios. Building on the shortest path network interdiction problem (SPNIP), BDAM incorporates dual defense operations—node interdiction and edge destruction—while explicitly modeling the defender's supply support. Unlike conventional SPNIP formulations, BDAM jointly considers the attacker's path disruption and the defender's logistical requirements, ensuring that all interdicted nodes are supported by available supply nodes without exceeding their capacity. To solve this NP-hard problem, a hybrid metaheuristic algorithm named Improved Simplified Swarm Optimization with Dijkstra (iSSOD) is proposed. The method integrates a population-based SSO framework with a randomized repair mechanism to ensure feasibility and an entropy-guided local search to enhance exploitation. The attacker's optimal response is computed efficiently using Dijkstra's algorithm, embedded within the defender's fitness evaluation. The experimental results on 36 artificial instances demonstrate that iSSOD consistently outperforms several benchmark evolutionary algorithms, providing high-quality solutions through a defense-aware, supply-constrained optimization framework. Furthermore, a real-world case study based on geographic data validates the model's applicability under realistic defense conditions. © 2025 Elsevier B.V.