Projecte matriculat

Títol: Basic orbital maneuvering using reinforcement learning

---

Director/a: MASDEMONT SOLER, JOSEP JOAQUIM

Departament: MAT

Títol: Basic orbital maneuvering using reinforcement learning

Data inici oferta: 15-07-2024     Data finalització oferta: 15-03-2025

---

Estudis d'assignació del projecte:

MU AEROSPACE S&T 21

Tipus: Individual

Lloc de realització: EETAC

Paraules clau:

Descripció del contingut i pla d'activitats:

The integration of reinforcement learning (RL) with modern computational tools can allow for real-time decision-making and autonomous control. Additionally, RL can be a way to efficiently cope with uncertainties and disturbances in the space environment, such as gravitational perturbations and fuel limitations. The ability to operate with limited prior knowledge makes RL a suitable candidate for missions where pre-existing models may be incomplete or inaccurate. The main purpose of the project is to explore artificial intelligence procedures to assess the efficiency of RL techniques in some aspects of basic orbital control and transfer. In a first approach, a simple Keplerian model can be considered to define the environment, being the main objective to find the simplest optimal policies in order to change orbital elements in a prescribed way. Depending on the results, and the computational capabilities needed for the training process, gravitational and non gravitational perturbations will be also considered.

Orientació a l'estudiant:

Requereix activitats hardware: No

Horari d'atenció a estudiants per a l'assignació de projecte:

---