Aerodynamic Variables and Loads Estimation Using Bio-Inspired Distributed Sensing

Abstract

Conventional control systems for autonomous aircraft use a small number of precise sensors encoding centre of mass motion and generally are setup for flight regimes where rigid body assumptions and linear flight dynamics models are valid. Flying animals in contrast take advantage of highly non-linear structural dynamics and aerodynamics to achieve efficient and robust flight control. It appears that the distributed arrays of flow and force sensors found in flying animals play a key roll in enabling their remarkable flight control. This paper presents current research using a wing model instrumented with distributed arrays of load and flow sensors to provide estimates of a range of aerodynamic and load related variables. The characteristics of instrumentation on the wing model, as well as those of a 1-DOF pitch rig are described and characterisation experiments carried out in a closed circuit low turbulence wind tunnel are presented. The results from these experiments show that a wealth of information can be extracted from the pressure and strain signals, including the state of the flow around the wing, and rate dependent non-linear structural and aerodynamic behaviour over a wide range of angles of attack, including well into the stall region. Using the signals from the distributed array Artificial Neural Networks were trained to provide estimates of angle of attack, airspeed, drag, lift and pitching moment. The networks were able to accurately estimate α (RMS error 0.15°), airspeed (RMS error 0.15 m/s-1.25% Full-Scale-Error (FSE)), drag (RMS error 0.33 N-1.78%FSE), lift (RMS error 0.57 N-0.60%FSE) and pitching moment (RMS error 0.03 N m-5.00%FSE). These estimators provided good estimates even in the stall region when the distributed array pressure and strain signals became unsteady. Future applications based on distributed sensing could include enhanced flight control systems that directly use measurements of aerodynamic states and loads, allowing for increase manoeuvrability and improved control of UAVs with high degrees of freedom such as highly flexible or morphing wings. Nomenclature Roman Symbols D Aerodynamic drag force, N L Aerodynamic lift force, N M Aerodynamic pitching moment, N m c Wing model mean aerodynamic chord, m q Wing model pitch rate,°/s b Wing model span, m S Wing model reference surface, m 2 V Wind speed, m/s Greek Symbols

Simulation of a Machine Learning Based Controller for a Fixed-Wing UAV with Distributed Sensors

“…Miniaturised airflow sensor systems based on arrays of distributed sensors have been designed to overcome these limitations on small and micro UAVs, e.g. [2][3][4][5]. These arrays are inspired by the flow sensor arrays found in birds [6], bats [7] and insects [8].…”

“…In addition, the outputs of the sensors showed that both the controlled and natural gusts can be detected and that properties which are not encoded by the inertial measurement unit (IMU) can be measured or calculated. Following this, a wing model was instrumented with both distributed pressure and strain sensors and the angle of attack, airspeed, drag, lift and pitch moment were estimated using ANNs [5]. Measurements captured detachment of the flow and non-linear behaviours such as hysteresis and rate dependent effects.…”

Artificial Neural Network-Based Flight Control Using Distributed Sensors on Fixed-Wing Unmanned Aerial Vehicles

“…These systems have utilized a number of different types of sensors such as diaphragm based pressure sensors [18], hot film sensors [19] and artificial hair sensors [20]. Studies of these systems in wind tunnel testing [21], as well as flight testing [22], have shown the potential of these systems for measuring a range of aerodynamic parameters.…”

“…Initial work has shown the potential for this type of sensor information to improve flight control, with the potential for faster responses [23] and the use of a physics based control approach [24]. Previous studies have looked at the advantages of each system in isolation in separate aircraft [25], the potential advantages of utilizing both distributed airflow and load information in combination with Artificial Neural Networks (ANN) to obtain estimates of the aerodynamic variables and loads [21], as well as the use of pressure-based distributed control systems for flight control [26] and gust alleviation [27].…”

“…The experimental platform was originally designed as a wing-only model. It was used to characterize the airflow and load signals from the distributed array and the signals were then used to obtain estimates of the aerodynamic variables and loads in a low turbulence wind tunnel [21]. The platform was modified to be used in a larger 2.13 m × 1.52 m (7 ft × 5 ft) wind tunnel, which allowed a larger α range (at least ±20°).…”