Artificial Intelligence Techniques for Deep Recursive Self-Attention Modules: MANET based Integrated Sensor System for Disaster Detection and Communication in Hazardous Environments: Trends and Challenges

Abstract

Disaster detection and communication in hazardous environments remain critical challenges due to dynamic conditions, infrastructure damage, and limited connectivity. Mobile Ad Hoc Networks (MANETs) have emerged as a viable solution for enabling decentralized communication in such environments. Recent advancements in artificial intelligence, particularly deep learning techniques such as convolutional neural networks, autoencoders, and attention-based architectures, have significantly enhanced disaster detection and response systems. This paper presents a comprehensive review of artificial intelligence techniques, focusing on deep recursive self-attention modules integrated with MANET-based sensor systems for disaster detection and communication. Recent studies demonstrate that attention-based deep learning models improve feature extraction by focusing on critical regions and contextual information, thereby enhancing disaster detection accuracy. For instance, attention mechanisms enable models to prioritize relevant features and suppress irrelevant data, leading to improved performance in remote sensing and image processing applications. Additionally, deep learning-based object detection and image enhancement techniques, such as YOLO and autoencoder-based methods, have significantly improved rescue operations in disaster scenarios by enabling real-time detection and monitoring. Despite these advancements, challenges such as high computational complexity, energy constraints, network instability, and data security remain unresolved. This review highlights key trends, challenges, and future directions in integrating deep recursive self-attention models with MANET-based disaster management systems.