Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Wind Estimation in Unmanned Aerial Vehicles with Causal Machine Learning

Abdulaziz Alwalan, Miguel Arana-Catania

TL;DR

The paper tackles wind estimation for UAVs using only trajectory data, addressing sensor-free wind awareness. It applies a causal curiosity framework that combines $DTW$-based time-series classification, $k$-means clustering, and a causal reward to identify wind conditions. It analyzes constant, shear, and turbulence winds and studies how wind speed, range similarity, environment count, thrust-change frequency, and wind direction affect identifiability, with a Cross-Entropy Method ($CEM$) used to optimize thrust sequences. The results show high classification performance and point to practical uses in trajectory planning and energy-efficient UAV operation under challenging weather.

Abstract

In this work we demonstrate the possibility of estimating the wind environment of a UAV without specialised sensors, using only the UAV's trajectory, applying a causal machine learning approach. We implement the causal curiosity method which combines machine learning times series classification and clustering with a causal framework. We analyse three distinct wind environments: constant wind, shear wind, and turbulence, and explore different optimisation strategies for optimal UAV manoeuvres to estimate the wind conditions. The proposed approach can be used to design optimal trajectories in challenging weather conditions, and to avoid specialised sensors that add to the UAV's weight and compromise its functionality.

Wind Estimation in Unmanned Aerial Vehicles with Causal Machine Learning

TL;DR

The paper tackles wind estimation for UAVs using only trajectory data, addressing sensor-free wind awareness. It applies a causal curiosity framework that combines -based time-series classification, -means clustering, and a causal reward to identify wind conditions. It analyzes constant, shear, and turbulence winds and studies how wind speed, range similarity, environment count, thrust-change frequency, and wind direction affect identifiability, with a Cross-Entropy Method () used to optimize thrust sequences. The results show high classification performance and point to practical uses in trajectory planning and energy-efficient UAV operation under challenging weather.

Abstract

In this work we demonstrate the possibility of estimating the wind environment of a UAV without specialised sensors, using only the UAV's trajectory, applying a causal machine learning approach. We implement the causal curiosity method which combines machine learning times series classification and clustering with a causal framework. We analyse three distinct wind environments: constant wind, shear wind, and turbulence, and explore different optimisation strategies for optimal UAV manoeuvres to estimate the wind conditions. The proposed approach can be used to design optimal trajectories in challenging weather conditions, and to avoid specialised sensors that add to the UAV's weight and compromise its functionality.
Paper Structure (16 sections, 6 equations, 9 figures, 10 tables)

This paper contains 16 sections, 6 equations, 9 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. UAVs wind conditions identification methods
  4. Causal curiosity in machine learning
  5. System model
  6. Methodology
  7. Experiments
  8. Results
  9. Effect of Wind speed on classification
  10. Effect of range similarity on classification
  11. Effect of number of environments on classification
  12. Effect of number of thrust changes on classification
  13. Effect of wind direction on classification
  14. Cross-entropy method search strategy for thrust actions sequence
  15. Discussion
  16. ...and 1 more sections

Figures (9)

  • Figure 1: UAV position and wind velocities through the trajectory.
  • Figure 2: System flow charts using different optimal action search strategies.
  • Figure 3: Example trajectories for a single loop and scores for all of them.
  • Figure 4: Silhouette score vs loop number for the wind speed experiments.
  • Figure 5: Silhouette score vs the number of environments for the range similarity experiments. A decrease in the number of the environment is equivalent to an increase in the difference between the wind speed ranges of the two wind conditions.
  • ...and 4 more figures