Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Continuous World Coverage Path Planning for Fixed-Wing UAVs using Deep Reinforcement Learning

Mirco Theile, Andres R. Zapata Rodriguez, Marco Caccamo, Alberto L. Sangiovanni-Vincentelli

TL;DR

This work tackles continuous-world coverage path planning for fixed-wing UAVs with the goal of minimizing power while achieving complete target-zone coverage under nonholonomic and safety constraints. It introduces a SCMDP-based framework solved via action-mapping Soft Actor-Critic (AM-SAC) and a self-adaptive curriculum, supported by a set-based neural network with Rectangle Feature Extractors and Cross-Attention to handle variable NFZ/TZ sets. Trajectories are generated as curvature-constrained quartic Bézier curves to ensure smooth, autopilot-friendly paths, with a feasibility model filtering infeasible actions before the objective policy selects among feasible options. Experiments on procedurally generated and hand-crafted maps show energy-efficient, scalable learning and generalization to new layouts, highlighting the practical potential for real-world fixed-wing CPP under continuous dynamics.

Abstract

Unmanned Aerial Vehicle (UAV) Coverage Path Planning (CPP) is critical for applications such as precision agriculture and search and rescue. While traditional methods rely on discrete grid-based representations, real-world UAV operations require power-efficient continuous motion planning. We formulate the UAV CPP problem in a continuous environment, minimizing power consumption while ensuring complete coverage. Our approach models the environment with variable-size axis-aligned rectangles and UAV motion with curvature-constrained Bézier curves. We train a reinforcement learning agent using an action-mapping-based Soft Actor-Critic (AM-SAC) algorithm employing a self-adaptive curriculum. Experiments on both procedurally generated and hand-crafted scenarios demonstrate the effectiveness of our method in learning energy-efficient coverage strategies.

Continuous World Coverage Path Planning for Fixed-Wing UAVs using Deep Reinforcement Learning

TL;DR

This work tackles continuous-world coverage path planning for fixed-wing UAVs with the goal of minimizing power while achieving complete target-zone coverage under nonholonomic and safety constraints. It introduces a SCMDP-based framework solved via action-mapping Soft Actor-Critic (AM-SAC) and a self-adaptive curriculum, supported by a set-based neural network with Rectangle Feature Extractors and Cross-Attention to handle variable NFZ/TZ sets. Trajectories are generated as curvature-constrained quartic Bézier curves to ensure smooth, autopilot-friendly paths, with a feasibility model filtering infeasible actions before the objective policy selects among feasible options. Experiments on procedurally generated and hand-crafted maps show energy-efficient, scalable learning and generalization to new layouts, highlighting the practical potential for real-world fixed-wing CPP under continuous dynamics.

Abstract

Unmanned Aerial Vehicle (UAV) Coverage Path Planning (CPP) is critical for applications such as precision agriculture and search and rescue. While traditional methods rely on discrete grid-based representations, real-world UAV operations require power-efficient continuous motion planning. We formulate the UAV CPP problem in a continuous environment, minimizing power consumption while ensuring complete coverage. Our approach models the environment with variable-size axis-aligned rectangles and UAV motion with curvature-constrained Bézier curves. We train a reinforcement learning agent using an action-mapping-based Soft Actor-Critic (AM-SAC) algorithm employing a self-adaptive curriculum. Experiments on both procedurally generated and hand-crafted scenarios demonstrate the effectiveness of our method in learning energy-efficient coverage strategies.
Paper Structure (29 sections, 24 equations, 8 figures, 1 table)

This paper contains 29 sections, 24 equations, 8 figures, 1 table.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. State-wise Constrained Markov Decision Process
  4. Reinforcement Learning
  5. Action Mapping
  6. Bézier Curves
  7. Problem Formulation
  8. Environment
  9. Fixed-Wing UAV
  10. Evolution of the Target Zones
  11. Power Modeling
  12. Optimization Objective
  13. Methodology
  14. SCMDP Formulation
  15. Observation Space
  16. ...and 14 more sections

Figures (8)

  • Figure 1: Example state of a fixed-wing UAV in a continuous world CPP problem on the left, showing the covered area, trajectory, and field of view, with a legend on the right.
  • Figure 2: Illustration of the TZ observation, showing five TZs in different colors and the rectangles that comprise them. Circles in the center indicate the centroids of the zones. Measurements in the green zone show the rectangle descriptors within a zone, while the ones for the blue zone indicate the TZ descriptors.
  • Figure 3: Visualization of the transition between consecutive quartic Bézier curves, highlighting the control points for the next spline (green) and the direction and normal vector of the UAV.
  • Figure 4: Neural network architecture and components.
  • Figure 5: Training curve of one agent using the curriculum, showing the increase of episodic difficulty dependent on the progress, using a log scale for training steps.
  • ...and 3 more figures