Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural-based Control for CubeSat Docking Maneuvers

Matteo Stoisa, Federica Paganelli Azza, Luca Romanelli, Mattia Varile

TL;DR

This paper presents an innovative approach employing Artificial Neural Networks trained through Reinforcement Learning for autonomous spacecraft guidance and control during the final phase of the rendezvous maneuver, offering insights into future mission expectations.

Abstract

Autonomous Rendezvous and Docking (RVD) have been extensively studied in recent years, addressing the stringent requirements of spacecraft dynamics variations and the limitations of GNC systems. This paper presents an innovative approach employing Artificial Neural Networks (ANN) trained through Reinforcement Learning (RL) for autonomous spacecraft guidance and control during the final phase of the rendezvous maneuver. The proposed strategy is easily implementable onboard and offers fast adaptability and robustness to disturbances by learning control policies from experience rather than relying on predefined models. Extensive Monte Carlo simulations within a relevant environment are conducted in 6DoF settings to validate our approach, along with hardware tests that demonstrate deployment feasibility. Our findings highlight the efficacy of RL in assuring the adaptability and efficiency of spacecraft RVD, offering insights into future mission expectations.

Neural-based Control for CubeSat Docking Maneuvers

TL;DR

This paper presents an innovative approach employing Artificial Neural Networks trained through Reinforcement Learning for autonomous spacecraft guidance and control during the final phase of the rendezvous maneuver, offering insights into future mission expectations.

Abstract

Autonomous Rendezvous and Docking (RVD) have been extensively studied in recent years, addressing the stringent requirements of spacecraft dynamics variations and the limitations of GNC systems. This paper presents an innovative approach employing Artificial Neural Networks (ANN) trained through Reinforcement Learning (RL) for autonomous spacecraft guidance and control during the final phase of the rendezvous maneuver. The proposed strategy is easily implementable onboard and offers fast adaptability and robustness to disturbances by learning control policies from experience rather than relying on predefined models. Extensive Monte Carlo simulations within a relevant environment are conducted in 6DoF settings to validate our approach, along with hardware tests that demonstrate deployment feasibility. Our findings highlight the efficacy of RL in assuring the adaptability and efficiency of spacecraft RVD, offering insights into future mission expectations.

Paper Structure

This paper contains 10 sections, 5 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Deep Reinforcement Learning for control
  3. Problem formalization
  4. Methodology
  5. Spacecraft model
  6. Reinforcement Learning setup
  7. Simulations
  8. Results
  9. Hardware tests
  10. Discussion

Figures (3)

  • Figure 1: Observations (back) and real values (front) of position coordinates during a docking maneuver using $\alpha=0.1$.
  • Figure 2: Visualization of 10 maneuver trajectories using $\alpha=0.05$. Other than the safety cone and its axis, the sphere where initial conditions are drawn is displayed.
  • Figure 3: Terminal position and velocity values using $\alpha=0.05$. Regarding distances reached along the Y-axis, the -0.005$m$ margin represents the minimal distance from the target to consider the docking satisfied.