WESPR: Wind-adaptive Energy-Efficient Safe Perception & Planning for Robust Flight with Quadrotors

Khuzema Habib; Pranav Deshakulkarni Manjunath; Kasra Torshizi; Pratap Tokekar; Troi Williams

cs.ROMar 10, 2026

WESPR: Wind-adaptive Energy-Efficient Safe Perception & Planning for Robust Flight with Quadrotors

Authors: Khuzema Habib, Pranav Deshakulkarni Manjunath, Kasra Torshizi, Pratap Tokekar, Troi Williams

Organizations: University of Maryland, College Park.

Abstract

Local wind conditions strongly influence drone performance: headwinds increase flight time, crosswinds and wind shear hinder agility in cluttered spaces, while tailwinds reduce travel time. Although adaptive controllers can mitigate turbulence, they remain unaware of the surrounding geometry that generates it, preventing proactive avoidance. Existing methods that model how wind interacts with the environment typically rely on computationally expensive fluid dynamics simulations, limiting real-time adaptation to new environments and conditions. To bridge this gap, we present WESPR, a fast framework that predicts how environmental geometry affects local wind conditions, enabling proactive path planning and control adaptation. Our lightweight pipeline integrates geometric perception and local weather data to estimate wind fields, compute cost-efficient paths, and adjust control strategies, all within 10 seconds. We validate WESPR on a Crazyflie drone navigating turbulent obstacle courses. Our results show a 35.6% reduction on average in maximum trajectory deviation and a 24.6% improvement in stability compared to a wind-agnostic adaptive controller.

Figures & tables

Fig. 1: A UAV navigating a turbulent environment. The red path shows the shortest route but crosses stronger winds; the green path follows a wind-aware trajectory through calmer regions.

Fig. 2: The WESPR path generation pipeline as described in Section III .

Fig. 3: Experimental pipeline. We capture an image of the environment’s physical space and the positions of the fans, obstacles, and drone’s start and goal (top left). We convert the image to an occupancy grid with fan positions (1s), and use that for our simulation (4s). Finally, we compute the drone’s path (5s) and, while the drone flies, localizes it (bottom).

Fig. 4: We estimate the state of the Crazyflie drone by fusing odometry from its onboard Flow V2 sensor and position data from the overhead camera via HSV tracking.

Fig. 5: 3D illustration of our three environments with the drone at the start and the red circle as the goal. The red fan represents high-speed, while the white fan represents low-speed. The Base path is shown as a red line, while the WESPR path is green. The velocities shown in each environment are relative values, not absolute.

Distance (m)	Low-Speed (m/s)	High-Speed (m/s)
$0.5$	$4.1$	$5.5$
$1.0$	$3.2$	$3.8$
$1.5$	$1.7$	$2.7$
$3.0$	$0.7$	$1.8$

TABLE I: Fan Speed Measurements over Distance

Fig. 6: Average wind-induced performance metrics across all environments (lower is better).

Env	Metric	Pen.( $\mathcal{P}^{\texttt{Base}{}}_{wind}$ )	Pen.( $\mathcal{P}^{\texttt{WESPR}{}}_{wind}$ )
E1	Max Dev	51.5	20.3
E1	P95 Dev	50.8	18.0
	Frechet	37.8	22.1
E2	Max Dev	76.8	42.8
E2	P95 Dev	67.4	35.1
	Frechet	97.5	50.0

TABLE II: Wind-Induced Penalty (%) by Environment (Best Emphasized)

Env	Metric	Case	Mode	Val.(m)	Traj. Disp.( $\Delta$ ;m)	Rel.Red.( $\eta_{r}$ %)
E1	Max Dev	Base	No Wind	0.386	0.199	58.7%
		Base	Wind	0.483	0.199
		WESPR	No Wind	0.405	0.082
		WESPR	Wind	0.430	0.082
	P95 Dev	Base	No Wind	0.336	0.171	57.9%
		Base	Wind	0.440	0.171

TABLE III: WESPR Quantitative Performance Comparison

Fig. 7: (LEFT) Actual trajectories taken in E2 by the drone (faintly visible with average path in bold), WESPR (Blue) and Base (Red). Base trajectory collides with the obstacles. (RIGHT) Jerk Magnitude along the path. WESPR path is more stable with reduced max and mean jerk compared to Base .

Env	Metric	Base	WESPR	Rel./Reduc.(%)
E1	Path Length (m)	2.785	3.211	+15.29%
	Mean Jerk ( $m/s^{3}$ )	1.94	1.50	+22.64%
	Collision	No	No	–
E2 *	Path Length (m)	3.326	3.484	+4.75%
	Mean Jerk ( $m/s^{3}$ )	1.63	1.23	+24.62%
	Collision	Yes	No	–

TABLE IV: Comprehensive Performance and Dynamics Analysis

Fig. 8: Comparison of FluidX3D and Ansys Fluent configured with matching boundary conditions ( $Re$ , $V_{\text{in}}$ ), across 17 probe locations (white circles) against real-world anemometer measurements.

Fig. 9: A Comparison of Jerk between Bezier vs. A* Trajectory.

Quadrotor Control

Quadrotor control studies how to stabilize, guide, and adapt aerial robots under demanding flight conditions and physical constraints.

See all 117 papers

Real Time Local Wind Inference for Robust Autonomous Navigation
Apr 1, 2026Spencer FolkAerial Robotics Robot Navigation
Mechanical Engineering and Applied Mechanics, University of Pennsylvania
WAND: Learning Robust Navigation under Complex Wind Disturbances and Dense Obstacles for Quadrotors
Oct 8, 2026Zhonghan Tang, Chenhui Li, Shuai Liang +5Disturbance Rejection Robot Navigation
Department of Electronic Engineering and Information Science, University of Science and Technology of China, Hefei 230027, China · Shanghai Artificial Intelligence Laboratory, Shanghai 200232, China · Northwestern Polytechnical University, Xi’an 710072, China +1
Wind-Aware Reinforcement Learning Control of a Small Quadrotor Using Learned Onboard Wind Estimation in Simulated Atmospheric Turbulence
Jul 1, 2026Abdullah Al Tasim, Wei SunRL Control Aerial Robotics
University of Oklahoma, Norman, Oklahoma, 73019

Apr 1, 2026J/K move · Enter open · S save

Real Time Local Wind Inference for Robust Autonomous Navigation

Spencer Folk

Mechanical Engineering and Applied Mechanics, University of Pennsylvania

Aerial Robotics Robot Navigation Spatiotemporal Forecasting

Open paper AskPDF arXiv

This thesis presents a solution that enables aerial robots to reason about surrounding wind flow fields in real time using on board sensors and embedded flight hardware. The core novelty of this research is the fusion of range measurements with sparse in situ wind measurements to predict surrounding flow fields. We aim to address two fundamental questions: first, the sufficiency of topographical data for accurate wind prediction in dense urban environments; and second, the utility of learned wind models for motion planning with an emphasis on energy efficiency and obstacle avoidance. Drawing on tools from deep learning, fluid mechanics, and optimal control, we establish a framework for local wind prediction using navigational LiDAR, and then incorporate local wind model priors into a receding-horizon optimal controller to study how local wind knowledge affects energy use and robustness during autonomous navigation. Through simulated demonstrations in diverse urban wind scenarios we evaluate the predictive capabilities of the wind predictor, and quantify improvements to autonomous urban navigation in terms of crash rates and energy consumption when local wind information is integrated into the motion planning. Sub-scale free flight experiments in an open-air wind tunnel demonstrate that these algorithms can run in real time on an embedded flight computer with sufficient bandwidth for stable control of a small aerial robot. Philosophically, this thesis contributes a new paradigm for localized wind inference and motion planning in unknown windy environments. By enabling robots to rapidly assess local wind conditions without prior environmental knowledge, this research accelerates the introduction of aerial robots into increasingly challenging environments.

Abstract

Quadrotor Control

Explore similar work

Real Time Local Wind Inference for Robust Autonomous Navigation

WAND: Learning Robust Navigation under Complex Wind Disturbances and Dense Obstacles for Quadrotors

Wind-Aware Reinforcement Learning Control of a Small Quadrotor Using Learned Onboard Wind Estimation in Simulated Atmospheric Turbulence