Skip to main content
Skip to article control options
No Access

Robust Neurocontrol for Autonomous Dynamic Soaring

AIAA 2022-1417
Session: Learning, Reasoning, and Data Driven Systems I
Published Online:

View Video Presentation:

The application of dynamic soaring techniques on small unmanned aerial vehicles (SUAVs) aims to exploit naturally occurring wind gradients to increase flight endurance. However, considering the limited computational resources onboard SUAVs and the imperfect nature of real-world environments and physical systems, there is a practical need to design simple and robust control systems. As such, this paper presents a neuroevolutionary strategy for generating efficient neurocontrollers that exhibit generalized and robust behavior. The Neuroevolution of Augmenting Topologies (NEAT) algorithm is applied to evolve networks in a way that preserves simplicity while maximizing performance. Simulated flight tests in stochastic environments show that resulting controllers can perform dynamic soaring for a range of initial conditions and time-varying parameters. Flight trajectories and robustness metrics are presented and compared for a small autonomous aircraft operating in such environments.