Jamming Efficacy Analysis of Chaff using AI/ML

Abstract

Chaff is a Passive Electronic Countermeasure technology that plays a pivotal role in war scenarios. It can be used as a passive jammer to shield a war platform effectively. It can mimic the platform’s radar cross section (RCS) signature and act as a deceptive pseudo target. This manuscript presents an analysis of the jamming efficacy of chaff cloud. For this, three feature extraction and four AI/ML classification methods were employed, assuming that the Moving Target Indicator (MTI) and Doppler capabilities of the tracking radar are off. The effect of three different chaff deployment locations on its jamming performance has been analysed to determine the best possible deployment location. In the measurement setup, the range profile of the cloud is measured in the presence of a target. The classification performances of the extracted feature vectors are evaluated using the Support Vector Machine (SVM), Unsupervised Distance Classification (UDC), Naïve Bayes (NB) and Decision Tree (DT). A maximum 58.33 % decrease in recognition rate was observed with the introduction of chaff cloud when UDC and SVM approaches are employed and chaff is deployed from 90°. Noise has been introduced to closely predict the actual, practical performance of chaff in an actual deployment environment. The recognition rates fall less than 8.33 % for SVM and NB when AWGN (Artificial White Gaussian Noise) is 1.2 times. Based on these results, the jamming efficacy and the optimized tactical strategy of the chaff cloud are proposed. Better Out of the four AI/ML approaches, UDC and DT exhibit the best jamming performance and SVM exhibits the best anti-jamming performance. Any discrepancy and chances of overfitting can be avoided using a larger dataset with more features.