Prediction of Cable Crosstalk for USV Based on LIGHTGBM

Abstract

Unmanned surface vehicle (USV) is an intelligent surface platform. It is used to perform dangerous or unsuitable for manual operations. USV contains complex types of electronic equipment. It causes worse electromagnetic environment. In order to reveal the relationship between cable harness layout and electromagnetic compatibility in USV, this paper propose a regression prediction model based on Lightgbm. A physical model is established for crosstalk between parallel transmission lines in metal confined spaces in USV. The above model is trained by parameters such as the distance between the cable bundles, the height from the ground, the voltage of the interference source. The accuracy of the prediction model is verified by comparing various prediction models. It can be found that comparing with BP neural network model and the decision tree model, the prediction model based on LIGHTGBM have a higher accuracy. The accuracy reaches 99%, the MSE reaches 10e-6, and the MAPE reaches 0.312. It can predict cable crosstalk for USV in high accuracy and avoid the complex operation of traditional numerical calculation method, which can provide good guidance for design of USV.