Recent Advances in Hybrid Transformer based Gated Graph Attention Capsule Network Design for Preventing Attack in Radar Target Detection: A Systematic Review

Abstract

Radar target detection is essential in modern surveillance, defence systems, autonomous navigation, and intelligent sensing applications. However, challenges such as noise, clutter, jamming attacks, and adversarial interference significantly reduce detection accuracy and system reliability. Traditional signal processing and machine learning techniques often struggle to model complex spatial-temporal dependencies in radar data, particularly in dynamic and hostile environments. Deep learning-based approaches have emerged as effective solutions by enabling automatic feature extraction and adaptive decision-making. Hybrid transformer-based architectures have gained considerable attention due to their ability to capture long-range dependencies and global contextual relationships through self-attention mechanisms. The integration of transformers with convolutional neural networks further enhances detection performance by combining local feature extraction with global context modelling. Additionally, graph attention networks improve radar sensing by modelling relational dependencies among interconnected radar nodes, making them highly effective in multi-sensor and heterogeneous environments. These graph-based approaches enhance feature propagation and improve robustness under low signal-to-noise ratio conditions. Capsule networks further strengthen radar detection frameworks by preserving spatial hierarchies and improving resistance to noise, distortions, and adversarial attacks. The combination of transformers, graph attention mechanisms, and capsule networks provides a powerful and intelligent framework for developing robust, accurate, and attack-resistant radar target detection systems.