Long-Range Geomagnetic Navigation with Enhanced Multi-Objective Evolutionary Computation

Extensive research findings indicate that numerous animals possess the ability to undertake long-range migrations utilizing the geomagnetic field without prior knowledge. For unmanned system navigation inspired by animal migrations, this paper proposes an enhanced Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D) that transforms the navigation path planning problem into a multi-objective optimization problem under the stimulation of geomagnetic gradient features. Due to the presence of geomagnetic anomalies and the disappearance of geomagnetic gradients in certain regions, employing the traditional MOEA/D method can result in multiple solutions being scattered along discontinuous segments of the Pareto Front (PF), thus rendering it difficult to determine the heading. To overcome this limitation, this article initially employs Principal Component Analysis (PCA) to reduce the dimensionality of the population generated from geomagnetic data. Subsequently, Density-Based Spatial Clustering of Noisy Data (DBSCAN) is utilized to cluster the groups in the effective region, selecting clusters with higher density as representatives of excellent solutions. Finally, a relationship is established between the optimal solution and the target values to update the decision space. Additionally, a velocity decay coefficient is introduced to stabilize the convergence towards the destination. Simulations of the enhanced MOEA/D and original MOEA/D for navigation are carried out based on the data retrieved from the WMM2020 model. The simulation outcomes verify the practicability and precision of the proposed method.