Distributed Acceleration Sensing for Small UAS Wind Gust Estimation

The use of small unmanned aircraft systems (sUAS) has grown exponentially over the past decade for applications such as military reconnaissance, search and rescue, 3D mapping and surveying, and meteorological and atmospheric research. Because of the smaller build and weight of these aircraft in comparison to their manned counterparts, turbulence and other atmospheric disturbances have larger adverse effects on sUAS control and stability, such as deviations in flight trajectory, diminished aerodynamic performance, and reduced endurance. Turbulence models exist as a means of simulating the response of an aircraft to gusty conditions; however, existing models do not scale well for sUAS. In order to create a turbulence model that is more representative of gusts a small aircraft might encounter at low altitudes, inertial wind velocity measurements must be taken and analyzed from these aircraft.

Current methods of wind measurement include anemometers and Pitot probes. These devices need calibration and can be expensive, heavy, and difficult to manufacture. A method of gust state estimation using distributed acceleration measurements is proposed in this study that removes the difficulties imposed by other wind probes. By precisely placing multiple accelerometers distributed away from the center of gravity, an aircraft’s translational and rotational acceleration states can be estimated. These acceleration states, which are proportional to the forces and moments acting on an aircraft, contain inherent information about wind gust disturbances. By feeding acceleration information through a state observer, wind gust velocity states can be estimated in addition to aircraft states.

The main contribution of this work is the development of a gust state observer that uses distributed acceleration measurements to provide estimates of gust velocities. To test the observer, an aircraft simulation was conducted using MATLAB and Simulink in which the aircraft was buffeted by translational and angular wind velocities for 15 seconds while flying at the trim condition. The aircraft model used was a linear model for the Tempest UAS and wind gusts were generated using the Dryden Turbulence Model. All aircraft velocity and acceleration states and control inputs were assumed known. Initial conditions for the gust velocities in the observer were chosen at random. The gust velocity estimates showed convergence to the simulated gust velocities over time and showed less error than traditional Kalman Filter estimates that were made without acceleration input. Future work includes testing this observer with actual flight data and comparing to Pitot probe measurements.