Adaptive morphing of wing and tail for stable, resilient, and energy-efficient flight of avian-informed drones

TL;DR

This work addresses control for avian-informed drones with morphing wings and tails, where large, coupled actuation DoFs pose control challenges. It introduces a body-rate control framework that uses a dynamics-informed mapping to drive all actuators, achieving stable flight even under disturbances and partial actuator loss. Through in-flight Bayesian optimization, the authors identify morph configurations that reduce energy consumption, delivering up to 11.5% savings at 8 m/s and 4.4% at 12 m/s, with configurations echoing avian flight strategies. The approach enables autonomous, energy-efficient, and robust flight in diverse wind conditions, laying groundwork for practical deployment of morphing aerial vehicles.

Abstract

Avian-informed drones feature morphing wing and tail surfaces, enhancing agility and adaptability in flight. Despite their large potential, realising their full capabilities remains challenging due to the lack of generalized control strategies accommodating their large degrees of freedom and cross-coupling effects between their control surfaces. Here we propose a new body-rate controller for avian-informed drones that uses all available actuators to control the motion of the drone. The method exhibits robustness against physical perturbations, turbulent airflow, and even loss of certain actuators mid-flight. Furthermore, wing and tail morphing is leveraged to enhance energy efficiency at 8m/s, 10m/s and 12m/s using in-flight Bayesian optimization. The resulting morphing configurations yield significant gains across all three speeds of up to 11.5% compared to non-morphing configurations and display a strong resemblance to avian flight at different speeds. This research lays the groundwork for the development of autonomous avian-informed drones that operate under diverse wind conditions, emphasizing the role of morphing in improving energy efficiency.

Figures (9)

• Figure 1: Overview. a The avian-informed drone LisEagle in free flight, with a wing span of 1.52m and a ready-to-fly mass of 752g. b It features eight degrees of freedom, with interdependent actuation between the wing sweep and twist and between the tail sweep and elevator position. c Wind tunnel experiments determine aerodynamic coefficients at various actuator positions, angle of attack, and angle of side slip. The interpolated data formulates a dynamics model, enabling the computation of the sensitivity matrix which yields a mapping matrix, that scales the importance $i$ of each control surface proportional to its influence on the body rate. d A cascaded control architecture produces reference body rates and thrust, with the latter directly applied to the drone. The mapping matrix (green) projects measured and reference body rates into the actuator space ($p_{\text{meas}}, p_{\text{ref}}$), enabling the use of Proportional--Integral--Derivative control with a feedforward term $F$. e An additional control loop (blue) using Bayesian optimization to minimize energy consumption. The configuration of the drone, described by $c_{\text{tail}}, c_{\text{sweep}}, c_{\text{twist}}$, modifies the centre positions of the tail sweep, symmetric wing sweep, and symmetric wing twist, influencing the lift-to-drag ratio and hence energetic consumption.
• Figure 2: Control commands respond to physical perturbations in flight. The LisEagle is disturbed at its extremities using a rod to induce torque in the roll, yaw, and pitch axis. Perturbations are categorized as soft or hard based on the force applied during the disturbance. To enhance readability, control commands are depicted transparently for scenarios in which their impact is minimal.
• Figure 3: Controller is robust to turbulent airflow. Turbulence is induced over the right wing of the LisEagle by using a bar to asymmetrically disturb the airflow, visualized through white smoke.
• Figure 4: Actuator loss is simulated by fixing specific control commands during flight. Position and angular velocity errors are analysed, with each column representing an experiment conducted at three different speeds throughout 20s. Unstable behaviour is denoted by a cross.
• Figure 5: Leveraging morphing to increase energy efficiency.a Overlay of different morphing configurations in flight. b Visualization of the BO algorithm exploring $20$ different configurations (coloured dots) per flight. Each experiment starts with the initial configuration $i$ followed by three model-based guesses $g$ to initialize the BO, which determines the next $16$ configurations. The best configuration $b$ corresponds to the lowest measured energy consumption at each velocity. The colour of each dot represents the respective energy consumption measurement divided by the airspeed (energy consumed per distance flown). The white crosses indicate unstable configurations (angular velocities $>$0.5rad/s or position error $>$13cm). c Illustration of the initial and best configurations and their resulting angle of attack of the wing $\alpha$ and pitch angles $\Theta$. d Comparison of the energy efficiency between the initial and best configurations at 8m/s, 10m/s and 12m/s, measured over 10s at 5Hz. The level of significance is indicated by the number of stars.