Bio-inspired Wake Tracking for Aircraft Formation Flight Based on Reinforcement Learning

Abstract

Formation flight is known for improving the overall aerodynamic efficiency of a pair of aircraft. This work presents a bio-inspired tracking strategy that correctly positions the follower in the wake of the leader in order to optimize the benefits of formation flight, i.e. optimize apparent drag reduction. A simplified aerodynamic model based on Prandtl's lifting line theory enables to compute the 6 degrees-of-freedom dynamics of the follower under the influence of a leader wake. Those dynamics are controlled over time thanks to a speed-control autopilot, while a neural network modifies the speed target according to the follower dynamics. Indeed, the follower dynamic measurements are used to train the neural network through the reinforcement learning framework, drawing inspiration from the trial and error of animal learning. The use of the follower dynamics as unique control measurements ensures a total independence of the tracking method from direct measurements of the leader wake position. Simulations demonstrate the good performances of this new reinforcement learned tracking method when dealing with a simple two-aircraft formation as well as with more challenging configurations involving a larger flock.