Visual aided formation control of unmanned aerial vehicle swarms

In this thesis, formation maintenance control of quadcopter swarms utilizing an onboard monocular camera and cubic marker in MATLAB and ROS/Gazebo simulation is studied. The proposed controller is a distributed controller that does not require any communication between the agents. The controller is designed to maintain the formation shape and the desired distance between the agents. The controller is tested in MATLAB and ROS/Gazebo simulation environments. The results show that the controller is able to maintain the formation shape and the desired distance between the agents. In this study, to build our swarm system, two different control structures, low-level and high-level, are utilized. While low-level control is for controlling vehicle dynamics, high-level control is for producing swarm behavior. The high-level controller is designed as a fully distributed controller that does not need any communication methodology or centralized computing platform. The high-level controller is designed based on single integrator dynamics, and the outputs of it are fed as desired position or velocity, depending on the high-level controller design. In low-level controller design, there are two different design methodologies that are preferred based on the simulation environment. The first one is feedback linearization for the tests in MATLAB. The second one is PX4 software in the loop for the tests in Gazebo. The results that are obtained from our swarm system show that our control method is able to maintain the formation shape and the desired distance between the quadcopters in simulation environments. These results are promising for real-world applications in areas that have communication and navigation problems. When the method is evaluated, it can be said that it can either be utilized as a primary control methodology for appropriate quadcopters or can also be utilized as an auxiliary system that helps the primary system. In addition, if the swarm system encounters any problems, our method can take over system control and safely continue to accomplish the task.