Acoustic Source Localization for Anti-UAV Based on Machine Learning in Wireless Sensor Networks

Abstract

Unmanned aerial vehicles (UAVs) have developed rapidly and are widely used in many fields. This phenomenon also causes security problems that urgently need to be addressed by anti-UAV technique. The localization of UAV plays an important role in anti-UAV system. An acoustic source localization scheme based on machine learning (ML) in wireless sensor networks is proposed in this study. Five ML algorithms, namely, artificial neural network (ANN), Naive Bayes, decision tree (DT), K nearest neighbors (KNN) and random forest (RF), are designed to estimate the coordinate of a single UAV. The acoustic energy decay model is constructed to simulate the attenuation and distortion caused by the ambient noise and changing surroundings. We use both received signal strength (RSS) based on acoustic energy and the difference of RSS as the input. Our experiments show that ML algorithms perform well except ANN. For ambient noise case, the ones with the input we propose achieve better localization accuracy than those only using RSS. KNN and RF are more suitable and reliable models for localization.