Assessment of EMI Effects on UAV Data Links

To assess the survivability of an unmanned aerial vehicle (UAV) in a complex electromagnetic environment, a novel method for assessing electromagnetic interference (EMI) threats to a UAV is introduced. A dataset of loss-of-lock thresholds for the UAV data link was generated through EMI injection tests. A genetic algorithm (GA)-optimized extreme gradient boosting (XGBoost) was then applied to efficiently predict these loss-of-lock thresholds. And Shapley additive explanations (SHAP) were used to measure the importance of features. Compared with K-nearest neighbor, support vector machine, decision tree, and XGBoost, GA-XGBoost shows better prediction accuracy and overall performance. Based on this, a GA-XGBoost-based assessment method was proposed to classify EMI effects into four levels using a three-level effect index. Finally, the EMI effect levels and SHAP results were used to formulate targeted anti-interference strategies. The proposed method can help to improve the anti-interference performance of UAVs.