Content area
Abstract
In recent years, numerous coordinated unmanned aerial vehicle systems have been extensively studied and developed for a variety of applications. Researchers aimed to expand the size of the collective group by incorporating diverse agents. However, due to connectivity constraints, a hierarchical control structure has been proposed for the autonomous operation of large unmanned aerial vehicle systems. A significant limitation of this hierarchical control structure is leader failure, which may lead to cascading failures within the entire group. The primary objective of this dissertation is to introduce a distributed, fully autonomous, and efficient leader selection algorithm supported by swarm intelligence. The proposed methodology is formalized within the framework of the best-of-n model. This dissertation comprises four key sections. Firstly, an in-depth analysis of swarm intelligence and behavior has been conducted. Building upon this analysis, a novel swarm intelligence algorithm, termed ”Tree Search Optimization Algorithm (TSOA),” is proposed. Utilizing the TSOA, a new leader selection algorithm is introduced, along with an automatic method for capturing fallen UAVs, which serves as the triggering event for the algorithm. Concurrently, a task allocation algorithm is put forth for large UAV systems based on the C-Net Protocol. The swarm intelligence analysis includes 21 algorithms and compares them against 30 benchmark tests. As a result, CSO and CHOA-B were identified as more universal, and 10 swarm intelligence algorithms were selected for comparison with TSOA. TSOA was compared with classical methods and five CEC benchmark suites, performing better on 45 out of 62 tests and 80% better than other swarm intelligence algorithms according to Wilcoxon’s rank test. The TSOA leader selection algorithm was tested against GHS, Raft, and Random methods and was determined to be over 90% accurate, as well as time and memory efficient. Considering the quality of the results, instead of focusing solely on the cost of the solution, the leader selection algorithm should be made more efficient for use in large-scale applications. The proposed automatic fallen UAV capturing method is 89.47% faster and more resourceful. Simultaneously, the proposed task allocation method is also quite applicable for larger multi-UAV systems. This dissertation introduces a fully autonomous, distributed, time and memory-efficient leader selection algorithm, improving mission reliability and safety to support many large UAV applications.
