Autonomous Wingman Flying Strategies Using a Learning Embedded Control Approach

Unmanned systems have grown significantly in popularity since the turn of the century. The advanced development of old and emergent technologies has ushered in an era of capable and accessible unmanned solutions. Increased processing power, memory density, component miniaturization, and energy storage capabilities all lend themselves to smaller systems without the need for a human onboard. Many industries have found applications for inexpensive and portable systems to reach heights, depths, and everything in between that were previously inaccessible without significant funding. Agriculture, defense, and entertainment are just a few of the industries who have been quick to acquire and apply these systems. With the increased demand for unmanned vehicles comes an equivalent need for control systems to utilize these vehicles in useful ways. One such requirement, and the topic of this thesis, is the implementation and optimization of autonomously controlled formation flight for fixed wing aircraft.

Traveling in formation is particularly useful in situations where multiple systems are transiting between locations. Whether this is the shipment of freight, passenger transportation, or multiple-angle videography, the ability to have a single system with one or many automated followers simplifies the control of multi-agent scenarios. This research proposes 2-dimensional, nonlinear control algorithms and expands the use of virtual target points to stabilize flight. Additionally, the application of virtual targets to anticipate a variable path is optimized through machine learning techniques. The result is a control system capable of following a leader while maintaining a relative position across a wide range of maneuvers.