The aim of this work is to develop an application for autonomous landing. We exploit the properties of Deep Reinforcement Learning and Transfer Learning, in order to tackle the problem of planetary landing on unknown or barely-known extra-terrestrial environments by learning good-performing policies, which are transferable from the training environment to other, new environments, without losing optimality. To this end, we model a real-physics simulator, by means of the Bullet/PyBullet library, composed by a lander, defined through the standard ROS/URDF framework and realistic 3D terrain models, for which we adapt official NASA 3D meshes, reconstructed from the data retrieved during missions. Where such model were not available, we reconstruct the terrain from mission imagery - generally SAR imagery. In this setup, we train a Deep Reinforcement Learning model - using DDPG - to autonomous land on the lunar environment. Moreover, we perform transfer learning on the Mars and Titan environment. While still preliminary, our results show that DDPG can learn a good landing policy, which can be transferred to other environments.
Autonomous Planetary Landing via Deep Reinforcement Learning and Transfer Learning / Ciabatti, Giulia; Daftry, Shreyansh; Capobianco, Roberto. - (2021), pp. 2031-2038. (Intervento presentato al convegno 1st Workshop on AI for Space at CVPR 2021 tenutosi a Virtual) [10.1109/CVPRW53098.2021.00231].
Autonomous Planetary Landing via Deep Reinforcement Learning and Transfer Learning
Ciabatti, Giulia
Primo
;Capobianco, Roberto
Ultimo
Supervision
2021
Abstract
The aim of this work is to develop an application for autonomous landing. We exploit the properties of Deep Reinforcement Learning and Transfer Learning, in order to tackle the problem of planetary landing on unknown or barely-known extra-terrestrial environments by learning good-performing policies, which are transferable from the training environment to other, new environments, without losing optimality. To this end, we model a real-physics simulator, by means of the Bullet/PyBullet library, composed by a lander, defined through the standard ROS/URDF framework and realistic 3D terrain models, for which we adapt official NASA 3D meshes, reconstructed from the data retrieved during missions. Where such model were not available, we reconstruct the terrain from mission imagery - generally SAR imagery. In this setup, we train a Deep Reinforcement Learning model - using DDPG - to autonomous land on the lunar environment. Moreover, we perform transfer learning on the Mars and Titan environment. While still preliminary, our results show that DDPG can learn a good landing policy, which can be transferred to other environments.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.