Reactive Trajectory Generation and Formation Control for Groups of UAVs in Windy Environments Open Access
Downloadable ContentDownload PDF
Unmanned aerial vehicles (UAVs) continue to become more prolific in academic, commercial, government, and military groups including use by police departments, first responders, infrastructure inspectors, and farmers monitoring agriculture. As UAV operations and missions are expanded, the vehicles must be able to navigate collision-free in windy, gusty, and unknown environments. This dissertation develops an algorithm to minimize the formation error for a group of heterogeneous vehicles maintaining a desired formation in a windy, unknown environment where the wind may exceed the flight controller's closed-loop stability margin requiring reference trajectory replanning. To realize this, a low computational cost 3D spatio-temporal wind model with turbulence and gusting is developed for the unique but correlated wind disturbances experienced by each vehicle in the formation. The vehicle dynamics must reflect the wind's impact on the vehicle motion including thrust variation and blade flapping resulting from induced inflow. These effects have significant impact on the vehicle's stability and maneuverability as the wind increases. To accurately model these effects, the quadrotor propeller dynamics model is compared to wind tunnel data taken for a single motor/propeller assembly. Subsequently, a trajectory generation algorithm is developed in which each vehicle only relies on its on-board sensors and communication with other vehicles to navigate. The algorithm generates smooth trajectories that guarantee the vehicle clears all obstacles and other vehicles by a user-defined clearance radius in finite time given a finite number of bounded obstacles and stationary goal positions. The same trajectory generation algorithm is used by both vehicles that are actively part of the formation and vehicles that are navigating into or out of the formation. This allows the trajectory generation algorithm to be used with high-level planners and low-level motion controllers. The performance of the wind model, vehicle dynamics model, trajectory generation algorithm, and formation control algorithm are demonstrated through simulated scenarios that show a group of vehicles successfully minimizing formation error in environments with and without obstacles and in various wind conditions. These simulations show that the traditional approach of separating path planning from disturbance rejection fails under extreme wind conditions.