Improved metaheuristic object detection and multi objective path planning models for task scheduling of autonomous multi agent aerial systems for search and rescue missions in flood isolated zones

Abstract

This doctoral dissertation explores the domain of autonomous multi-agent aerial newlinesystems, focusing on enhancing their capabilities for search and rescue operations in newlineflood-isolated regions. The research concentrates on two pivotal elements: survivor newlinedetection and multi-objective path planning. The survivor identification methodology newlineinvolves three key stages pre-processing, feature extraction, and object recognition. newlineVideos are transformed into image frames, and an enhanced region-based convolution newlineneural network is employed for object detection. Improved features such as Local Gabor newlineBinary Pattern, traditional residual networks, and Scale-Invariant Feature Transform are newlineextracted from these images. In the object recognition phase, the African Vulture newlineUpdated Honey Badger Optimization is introduced to optimize weight parameters in newlinehybrid neural networks like bi-directional gated recurrent unit and long short term newlinememory. This hybrid deep learning model significantly enhances accuracy and newlinerobustness in identifying critical elements within dynamic and challenging newlineenvironments, such as flood-stricken areas. newlineAdditionally, the research introduces a Multi-Objective Path Planning newlineframework aimed at optimizing safety levels and travel distances for autonomous multiagent aerial systems. The best flight path is determined optimally between neighbouring newlineacquisition points through a new hybrid optimization model the Deer Hunter Updated newlineWhale Optimization, a conceptual fusion of standard Deer Hunting Optimization newlineAlgorithm and Whale Optimization Algorithm. The proposed collision-free optimal newlinepath framework, referred to as the Multi-Objective Path Planning with the Deer Hunter newlineUpdated Whale Optimization model, is compared with existing models in terms of newlinesafety levels and distance. The integrated optimization model, Deer Hunter Updated newlineWhale Optimization demonstrates superior performance in collision-free optimal path newlineplanning, particularly in three dimensional cluttered environments.