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

TL;DR

This work reframes coverage optimization as an expensive mixed-variable optimization problem (EMVOP) and introduces RI-SHM, a RankNet-inspired surrogate-assisted hybrid metaheuristic. RI-SHM integrates a RankNet-based pairwise global surrogate for robust global ranking, a surrogate-assisted local Estimation of Distribution Algorithm (EDA) for exploitation, and a fitness diversity-driven switching strategy to balance exploration and exploitation. Empirical results on large-scale 3-D coverage scenarios (up to $D=300$, $|Q|>1800$ targets) show RI-SHM outperforms state-of-the-art SAEAs by up to 56.5% in solution quality and maintains favorable runtimes, with ablation analyses highlighting the complementary value of global and local components. The approach advances scalable, high-fidelity coverage optimization by leveraging pairwise ranking surrogates, online adaptations, and diversity-aware strategy switching, offering practical impact for sensor placement and regional surveillance under complex visibility and angular constraints.

Abstract

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 reveal that the 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 computational cost and complexity, evolving the MVOP into an Expensive Mixed-Variable Optimization Problem (EMVOP). While canonical Evolutionary Algorithms have yielded promising results, their reliance on numerous fitness evaluations 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), an extension of our previous work. RI-SHM integrates three key components: (1) a RankNet-based pairwise global surrogate that innovatively predicts rankings between pairs of individuals, bypassing the challenges of fitness estimation in discontinuous solution space; (2) a surrogate-assisted local Estimation of Distribution Algorithm that enhances local exploitation and helps escape from local optima; and (3) a fitness diversity-driven switching strategy that dynamically balances exploration and exploitation. Experiments demonstrate that our algorithm can effectively handle large-scale coverage optimization tasks of up to 300 dimensions and more than 1,800 targets within desirable runtime. Compared to state-of-the-art algorithms for EMVOPs, RI-SHM consistently outperforms them by up to 56.5$\%$ across all tested instances.

Figures (12)

• Figure 1: A high-fidelity realistic coverage optimization scenario in our previous work wu2024mixed. The upper heatmap of Fig. \ref{['fig:coverage_intro']} represents the demand distribution, while Fig. \ref{['fig:Resolution_a']} and \ref{['fig:Resolution_b']} show the distribution of demand points within the deployment area under low and high discretization resolutions, respectively. It is obvious that the computational costs are expensive as the pairs of supply-demand points mount up cordeau2019benders. For example, solving a case with 100 candidate sites, 300 dimensions, and 30,000 targets requires over 280 hours.
• Figure 2: The generic framework of the proposed RI-SHM consists of three key components: a RankNet-based pairwise global surrogate model, a surrogate-assisted local EDA, and a fitness diversity-driven switching strategy.
• Figure 3: The network structure of global surrogate model.
• Figure 4: Diagram of voting-based preselection strategy.
• Figure 5: Diagram of multiple plateau characteristics in our problem. Solutions with similar fitness values may differ significantly in the solution space.