Planning, Control, and Estimation for Diverse Multi-UAS Missions

Unmanned Aircraft Systems (UAS) are being used for a variety of single vehicle missions such as surveillance, inspection, and payload delivery. Teams of UAS can perform these same missions more efficiently and can pursue novel cooperative missions not possible with a single vehicle. However, this comes at the cost of increased system complexity and introduces the challenge of safe team coordination. This motivates our research to pursue four diverse multi-UAS mission configurations. We propose novel methods to command and control teams of UAS with the majority supported with full-scale experimental validation. To support all experiments conducted in this thesis, an experimental test bed consisting of a custom, open-source quadrotor, flight controller, and supporting infrastructure is developed with designs and code shared publicly. This thesis offers four specific contributions to enable the deployment of UAS teams in missions with potential to benefit society. First, an existing formation control method, continuum deformation, is experimentally validated with observed tracking errors and delays informing the theory. Required inter-vehicle separation constraints are defined and applied to the real system. A global minimum separation bound based on a local controller error bound is derived to guarantee safety during real-world flights. Second, a novel heads-up haptic pushing interface is developed that enables a user to move a heavy payload carried by multiple small UAS through a crowded cluttered environment. Real-time estimation of user applied force via an instrumented payload updates virtual dynamics of an admittance controller to guide the system. This capability will support resilient and safe package delivery to untrained consumers and can assist in delivering medical and survival supplies in disaster relief scenarios. Third, computationally efficient planning methods are developed to support wildfire mapping over a large area by a team of UAS. A state machine is used to handle multi-vehicle task allocation between exploration (coverage) and exploitation (line following). Efficiency gains are achieved by separating the 3D problem into 2D lateral and 1D terrain avoidance sub-problems. This work offers a first step towards mapping increasingly severe wildfire threats that cause significant damage and claim the lives of hundreds of people every year. Fourth, a generalized path planner is developed to manage a deformable small UAS formation capable of rotating, expanding, contracting, and shearing. Provably sufficient inter-agent collision avoidance constraints are leveraged to efficiently plan safe trajectories in large-scale complex but static environments. An integrated guidance and control module onboard each small UAS tracks the designed trajectory while avoiding pop-up obstacles and vehicle failures by following an ideal fluid flow field airspace template. We are only beginning to imagine and prototype the missions made possible with teams of UAS. This thesis provides problem formulations, solution methods, and experimental realizations for four multi-UAS mission configurations. Countless other missions can be pursued that will operate safely and can improve quality of life. Continued research is needed to achieve this including full system integration with onboard sensing and collaboration with key stakeholders for each mission that deeply understand the problem spaces. Nonetheless, this thesis provides a solid foundation for future researchers to build upon.PhDRoboticsUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/174372/1/mmroma_1.pd