Periodic Reporting for period 1 - MEW (Multirotor Estimation of Wind)
Berichtszeitraum: 2024-06-17 bis 2026-06-16
The MEW (Multirotor Estimation of Wind) project addresses the measurement of unpredictable impact of wind forces on multirotor systems, which is a critical barrier to the widespread adoption of drones. Wind disturbances cause instability, degraded image quality, and safety concerns, limiting applications of drones in sectors such as infrastructure inspection, agriculture, delivery, and industrial monitoring. The EU drone market is projected to exceed €10 billion annually by 2035, making the improvement of reliability under challenging real-world conditions both economically and strategically important.
The project uses the following techniques to develop new wind estimation solutions:
1. A novel Acoustic Resonance Sensor (ARS) for direct wind velocity measurements.
2. An Equivariant Filter (EqF) for state estimation that takes advantage of the geometry inherent to multirotor vehicles.
3. Reduced Order Modelling (ROM) to decribe wind dynamics clearly and effectively for application.
The MEW project will help drones to operate safely and effectively in windy environments, unlocking new industrial applications and addressing societal concerns about drone safety.
The project uses the following techniques to develop new wind estimation solutions:
1. A novel Acoustic Resonance Sensor (ARS) for direct wind velocity measurements.
2. An Equivariant Filter (EqF) for state estimation that takes advantage of the geometry inherent to multirotor vehicles.
3. Reduced Order Modelling (ROM) to decribe wind dynamics clearly and effectively for application.
The MEW project will help drones to operate safely and effectively in windy environments, unlocking new industrial applications and addressing societal concerns about drone safety.
The response of an Acoustic Resonance Sensor (ARS) to various wind conditions was completed in the laboratory, and equations for an Equivariant Filter (EqF) were derived for the estimation of multirotor and wind states. Substantial custom software was built for the ARS to be able to interface with standard robotics software tools such as the Robotics Operating System (ROS). The response of the ARS was successfully and accurately modelled as an additive Gaussian process with time lag. An experimental setup was developed with a large fan to produce a known wind that could be measured by the ARS and compared to its reported wind speed. The model of the ARS response was extended to describe the sensor behaviour at different angles to the incoming wind. This was done by modifying the experimental setup to allow for two servo motors to change the angle between the sensor and the wind autonomously and continuously. Further software was also created to control these motors and record their states over time.
This project proposed a novel random walk stochastic model for the Acoustic Resonance Sensor (ARS), which was validated in laboratory experiments with wind measured at a standard deviation of 0.28 m/s. This was achieved at a wind speed of 4.8 m/s, therefore the average prediction error was indeed less than 20% of the maximum wind speed. Additionally, the sensor lag was identified as 0.14 seconds. This model is a key outcome beyond the state of the art of wind estimation in aerial robotics. Potential impacts include significantly improved flight of multirotor systems in challenging high-wind conditions.