Deep Learning and Optimization Approaches in Deep Recursive Self-Attention Modules: MANET-Based Integrated Sensor System for Disaster Detection and Communication in Hazardous Environments: A Review

Abstract

Disaster detection and communication in hazardous environments require intelligent, adaptive, and infrastructure-independent systems capable of real-time monitoring and decision-making. Mobile Ad Hoc Networks (MANETs), combined with integrated sensor systems, provide a decentralized framework for disaster management. This paper presents a comprehensive review of deep learning and optimization approaches, focusing on deep recursive self-attention modules for enhancing MANET-based disaster detection systems. Deep learning techniques such as convolutional neural networks, recurrent neural networks, and autoencoders have been widely applied for analysing sensor data and detecting anomalies. Recent advancements in self-attention and transformer-based architectures have significantly improved the ability to capture long-range dependencies in dynamic environments. Recursive self-attention modules further enhance prediction accuracy by iteratively refining feature representations. Additionally, optimization techniques such as reinforcement learning and energy-efficient routing algorithms improve network performance and resource utilization in MANET systems. Studies indicate that deep learning-based approaches achieve high accuracy in detecting anomalies and security threats, with some models reaching up to 99.5% detection accuracy in MANET environments. Furthermore, integrated sensor systems using machine learning models demonstrate strong performance in disaster detection tasks, achieving high correlation and prediction accuracy. Despite these advancements, challenges such as energy efficiency, computational complexity, and security vulnerabilities remain. This review highlights current trends, identifies research gaps, and proposes future directions for developing intelligent and scalable disaster management systems.