Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reinforcement Learning for Gliding Projectile Guidance and Control

Joel Cahn, Antonin Thomas, Philippe Pastor

TL;DR

The paper develops a six-degree-of-freedom reinforcement-learning framework to control a fixed-wing gliding projectile toward a camera-detected target. It builds a parametric glider model, a strapdown camera projection, and a custom Gymnasium environment to train a PPO agent that optimizes LOS-based guidance while penalizing actuator effort. Results show rapid convergence and improved performance over a classic PID baseline in windy conditions, highlighting RL's potential for autonomous glider navigation. Limitations identified include LOS-guidance weaknesses with moving targets and wind, with future work proposing more robust guidance and real-world validation.

Abstract

This paper presents the development of a control law, which is intended to be implemented on an optical guided glider. This guiding law follows an innovative approach, the reinforcement learning. This control law is used to make navigation more flexible and autonomous in a dynamic environment. The final objective is to track a target detected with the camera and then guide the glider to this point with high precision. Already applied on quad-copter drones, we wish by this study to demonstrate the applicability of reinforcement learning for fixed-wing aircraft on all of its axis.

Reinforcement Learning for Gliding Projectile Guidance and Control

TL;DR

The paper develops a six-degree-of-freedom reinforcement-learning framework to control a fixed-wing gliding projectile toward a camera-detected target. It builds a parametric glider model, a strapdown camera projection, and a custom Gymnasium environment to train a PPO agent that optimizes LOS-based guidance while penalizing actuator effort. Results show rapid convergence and improved performance over a classic PID baseline in windy conditions, highlighting RL's potential for autonomous glider navigation. Limitations identified include LOS-guidance weaknesses with moving targets and wind, with future work proposing more robust guidance and real-world validation.

Abstract

This paper presents the development of a control law, which is intended to be implemented on an optical guided glider. This guiding law follows an innovative approach, the reinforcement learning. This control law is used to make navigation more flexible and autonomous in a dynamic environment. The final objective is to track a target detected with the camera and then guide the glider to this point with high precision. Already applied on quad-copter drones, we wish by this study to demonstrate the applicability of reinforcement learning for fixed-wing aircraft on all of its axis.

Paper Structure

This paper contains 18 sections, 18 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Modeling the glider
  3. Description of the Glider
  4. Coordinate System Definition
  5. 6 DOF Equations
  6. Aerodynamic Forces and Moments
  7. Objective
  8. Reinforcement learning
  9. Observation Space
  10. Action Space
  11. Reward Function
  12. Reinforcement Learning Algorithm
  13. Training Environment
  14. Classic controller
  15. Results and discussion
  16. ...and 3 more sections

Figures (6)

  • Figure 1: Coordinate system used in the simulation
  • Figure 2: Training framework organisation
  • Figure 3: Episodic reward over training
  • Figure 4: Trajectory of the RL agent
  • Figure 5: Circular Error Probable (CEP) no wind
  • ...and 1 more figures