Deep Reinforcement Learning approach for Small Bodies Shape Reconstruction Enhancement
Margherita Piccinin, Michèle Roberta Lavagna · AIAA Scitech 2020 Forum · 2020
This paper tackles Autonomous Small Bodies Exploration (ASBE), i.e. the problem of mapping an unknown comet or asteroid while autonomously navigating in its proximity. A Deep Reinforcement Learning (DRL) based planning policy is designed to increase mapping efficiency with a smart autonomous selection of the images acquisition times. Such policy is verified with an Image Processing algorithm that generates a point cloud of the body shape model, based on the acquired images, via feature detection and matching.