Abstract
Accurate trajectory prediction of concealed targets in complex, interference-laden environments present a formidable challenge for millimeter-wave sensor tracking systems. To address this, we propose a state-of-the-art autonomous prediction framework that integrates an Improved Sequential Variational Mode Decomposition (ISVMD) algorithm with an Extreme Learning Machine (ELM), synergistically optimized by the novel Red-billed Blue Magpie Optimizer (RBMO). The ISVMD enhances signal reconstruction by transforming noisy target echo signals into robust feature sequences, effectively mitigating the impacts of environmental disturbances and intentional concealment. Subsequently, the RBMO-optimized ELM leverages these feature sequences to predict the future trajectories of concealed targets with high precision. The RBMO further refines critical parameters within the ISVMD-ELM pipeline, ensuring adaptability and computational efficiency across diverse scenarios. Experimental validation using real-world data demonstrates that the RBMO-ISVMD-ELM approach surpasses state-of-the-art algorithms in both accuracy and robustness when predicting the trajectories of concealed ground targets, achieving optimal performance metrics under demanding conditions.