.png){kind=icon}
返回
Wind estimation by multirotor dynamic state measurement and machine learning models
.png){kind=icon}
PRE
AI
DOI：10.1016/j.measurement.2022.111331.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
We present multiple models for the problem of wind estimation by a multirotor drone in low altitude atmospheric turbulence. Data is collected by test flights with an instrumented drone in proximity to an anemometer for training, validation, and testing. Three machine-learning models are developed: a long-shortterm-memory (LSTM) neural-network, an artificial neural-network, and a Gaussian-process-regression. These models are developed with variations in the inputs, considering the addition of drag estimations by known equations of motion and motor speeds, both of which improved performance. The LSTM model performed the best reaching 0.34 m/s root-mean-square-error in validation. The similarity between all model's performance without optimization indicates that we may be approaching the limits of accuracy of the experimental ground truth-a single anemometer that cannot be exactly co-located with the drone. The performance demonstrated so far suggests that the method may be useful in pollutant plume characterization, preliminary wind surveys, and other applications.
Keyword:
Wind estimation
Multirotor drone
Machine-learning
Long-short-term-memory
Neural-network
Gaussian-process-regression
期刊
IF:
5.6
论文数:
2.1W
被引数:
5.4W
机构
引用论文
A UAV-Based Eddy Covariance System for Measurement of Mass and Energy Exchange of the Ecosystem: Preliminary Results.png)
SENSORS
OAAI
SENSORSIF3.5
OAAI分享收藏Two Supervised Machine Learning Approaches for Wind Velocity Estimation Using Multi-Rotor Copter Attitude MeasurementsSENSORSOAAI
SENSORSIF3.5
OAAI分享收藏
查看更多内容