A RankNet-Inspired Surrogate-Assisted Hybrid Metaheuristic for Expensive Coverage Optimization

Coverage optimization generally involves deploying a set of facilities to best satisfy the demands of specified points, with broad applications in fields such as location science and sensor networks. Recent applications have revealed that subset site selection, coupled with continuous angular parameter optimization, can be formulated as mixed-variable optimization problems (MVOPs). Meanwhile, high-fidelity discretization and visibility analysis significantly increase the computational cost and complexity, turning such MVOPs into expensive MVOPs (EMVOPs). While canonical evolutionary algorithms (EAs) have yielded promising results, their reliance on numerous fitness evaluations (FEs) is too costly for our problem. Furthermore, most surrogate-assisted methods face limitations due to their reliance on regression-based models. To address these issues, we propose the RankNet-inspired surrogate-assisted hybrid metaheuristic (RI-SHM), which extends our previous standalone optimizer into a surrogate-assisted global–local framework for expensive realistic coverage optimization. RI-SHM integrates three key components: 1) a RankNet-based pairwise global surrogate that predicts rankings between pairs of individuals, avoiding the difficulty of direct fitness estimation in discontinuous solution spaces; 2) a surrogate-assisted local estimation of distribution algorithm (EDA) that enhances local exploitation and helps escape from local optima; and 3) a fitness diversity-driven switching strategy that dynamically balances exploration and exploitation. The experiments demonstrate that our algorithm can effectively handle large-scale coverage optimization tasks of up to 300 dimensions and more than 1800 targets within a desirable runtime. Compared with state-of-the-art (SOTA) algorithms for EMVOPs, RI-SHM achieves improvements of up to 56.5% across all tested instances.