Data-driven unmanned aerial vehicle wind disturbance model online wind disturbance estimation method

Abstract

The invention discloses a data-driven unmanned aerial vehicle wind disturbance model online wind disturbance estimation method, which does not need a complex and tedious wind disturbance theoretical model in practical application, adopts a data-driven strategy, and obtains a group of neural network characteristic quantities capable of displaying wind disturbance characteristics through offline training. Huge calculation amount of online updating of neural network parameters is avoided; the least square algorithm is used for carrying out linear combination on the neural network characteristic quantity, online real-time feedback data can be used, parameters of linear combination can be adjusted online, and a self-adaptive estimation effect is achieved. According to the method, the concept of multi-task learning is utilized, the neural network features are used as a sharing layer, the linear combination of the features is used as a top layer, the neural network feature training process with a large calculation amount is performed offline, and the least square algorithm with a small calculation amount is operated online, so that the strong representation capability of the neural network can be fully utilized, and the robustness of the neural network is improved. And the online calculated amount can meet the requirement of real-time operation.

Description

technical field [0001] The invention relates to the technical field of UAV anti-disturbance control, in particular to a data-driven UAV wind disturbance model online wind disturbance estimation method. Background technique [0002] In recent years, due to the wide application of UAVs in exploration, search and rescue, space imaging, transportation and other scenarios, research on UAV control has received great attention from academia and industry. It is worth noting that in practical applications, wind disturbances in the natural environment are almost ubiquitous, causing the uncertainty and complexity of the UAV system. influences. However, the general motion control of UAV systems usually does not design an anti-disturbance control system for wind disturbance scenarios, but simply introduces an integral link in the controller to suppress possible disturbances, because the integral takes time to offset the disturbance. , so the disturbance suppression process is very slow...

AI Technical Summary

Problems solved by technology

It is worth noting that in practical applications, the wind disturbance in the natural environment is almost everywhere, causing the uncertainty and complexity of the UAV system. If it is not properly handled, it will seriously affect the movement of the UAV influences

However, the general UAV system motion control usually does not design the anti-disturbance control system for the wind disturbance scene, but directly and simply introduces the integration link in the controller to suppress the possible disturbance, because the integration takes time to offset the disturbance , so the process of disturbance suppression is very slow. During this process, the UAV will greatly deviate from the set target point

Images