Dimension-Partitioned Variable-Length Genetic Algorithm for Layered Protection Optimization of High-Value Assets

Abstract

To address the challenge of mass approaches by unidentified Unmanned Surface Vehicle (USV) swarms toward maritime High-Value Assets (HVAs), this paper investigates the resource optimization of a layered protection system. Given that the optimal number of protection layers is uncertain, resource optimization presents a complex variable-dimensional nonlinear problem. To solve this, this paper proposes a Dimension-Partitioned Variable-Length Genetic Algorithm (DP-VLGA). The algorithm employs a hybrid variable-length encoding strategy to map protection layers, spatial layout, and Unmanned Aerial Vehicle(UAV) resource allocation onto variable-length chromosomes, enabling the adaptive evolution of protection configurations. To overcome premature convergence common in variable-dimensional search spaces, a dimension-partitioned solution pool management mechanism is introduced. Furthermore, an adaptive alignment-based crossover operator is developed to ensures physical validity during gene exchange across different dimensions. Experimental results demonstrate that DP-VLGA adaptively generates optimal protection configurations under varying approach intensities. Compared to other variable-dimensional optimization algorithms, the proposed method exhibits significant advantages in search accuracy and convergence stability, achieving higher control probabilities with lower resource consumption.