Proposal — Autonomous Navigation of Small Aircraft Through Complex Terrain

Current autonomous control systems for small, remote-controlled, fixed-wing aircraft lack the capabilities needed for navigation of complex environments. This limits their use in areas with unpredictable and intricate terrain, such as forests.

The goal of this project is to enable autonomous navigation of an RC plane in complex environments. I hope to create an AI that can navigate to a designated destination while avoiding obstacles and maintaining stability. This has potential applications in search and rescue, environmental monitoring, surveying, and disaster response.

To accomplish this, I will begin by designing a realistic model of a plane in Fusion 360. I will then use a CFD tool such as OpenFOAM or SU2 to determine aerodynamic data about the plane, which will be used to train a surrogate model for Unity simulations. The project will implement EKF SLAM (Extended Kalman Filter Simultaneous Localization And Mapping) for real-time mapping of terrain, combined with RL algorithms like SAC and TD3 for flight control. I will use Unity for the simulated environment in which I will train the AI.