| Abstract [eng] |
With the growing use of unmanned aerial vehicles in various industrial fields, search and military operations, the issue of autonomous swarm control is becoming increasingly relevant. Increasingly, there is a desire to ensure that a group of drones can coordinate actions as independently as possible and carry out missions without direct human intervention. This master's thesis examines the problem of developing swarm control for unmanned aerial vehicles. The theoretical part of the work reviews the main characteristics of unmanned aerial vehicle swarms and the technologies of their control systems. Analyzing the models of “Boids”, “Vicsek”, potential fields, leader-follower, PSO, ACO and SOMA optimization methods, it was found that none of them are suitable for ensuring the desired performance of the swarm. The stability of the “Boids” model decreases when swarms use a large number of agents. “Vicsek” model is often too simple, so its operation relies on the use of other models. Local minimums created by the potential fields model can stop the swarm when the swarm finds positional balance, and the leader-follower model can break if the only leader of the swarm fails. For these reasons, the need to use a hybrid control system model arises. The methodological part analyzes the “Fuzzy” mathematical model, which is applied to the basic control of the swarm movement. The model's ability to handle uncertain data systems without the need to precisely define the entire swarm environment is well suited to perform the desired algorithm. The swarm algorithm is composed of four distinct phases, during which the swarm is formed into a formation which is designed for optimal search pattern, launched for a large-scale search, grouped around a newly elected leader once an object is found, and set for collision with the found object. A three-component motion control system is used for phase motion, which includes separation, alignment, and coherence components. This algorithm is tested by running two different simulations in the “MATLAB” program, where the object is either found or not. The simulations demonstrate the capabilities of the developed algorithm to effectively and stably control swarms in close and wide formations and to perform rapid and coordinated actions in close environments. The shortcomings of the used system are also reviewed. Although the developed algorithm implements the desired goals of the swarm, to use the algorithm in real life scenarios, it is necessary to evaluate its current operation in two-dimensional space and evaluate communication interference and real-world data transmission problems. |
