Abstract:The rapid advancement of autonomous vehicle (AV) technology has introduced significant challenges in ensuring transportation security and reliability. Traditional AI models for anomaly detection in AVs are often opaque, posing difficulties in understanding and trusting their decision making processes. This paper proposes a novel feature ensemble framework that integrates multiple Explainable AI (XAI) methods: SHAP, LIME, and DALEX with various AI models to enhance both anomaly detection and interpretability. By fusing top features identified by these XAI methods across six diverse AI models (Decision Trees, Random Forests, Deep Neural Networks, K Nearest Neighbors, Support Vector Machines, and AdaBoost), the framework creates a robust and comprehensive set of features critical for detecting anomalies. These feature sets, produced by our feature ensemble framework, are evaluated using independent classifiers (CatBoost, Logistic Regression, and LightGBM) to ensure unbiased performance. We evaluated our feature ensemble approach on two popular autonomous driving datasets (VeReMi and Sensor) datasets. Our feature ensemble technique demonstrates improved accuracy, robustness, and transparency of AI models, contributing to safer and more trustworthy autonomous driving systems.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the accuracy and interpretability of anomaly detection in autonomous driving systems. Specifically, the paper addresses the following key challenges: 1. **Incomplete feature importance assessment**: Individual Explainable AI (XAI) methods usually can only provide limited insights into feature importance and are unable to comprehensively capture all the factors that influence the decisions of anomaly detection models. 2. **Lack of consensus among XAI methods**: Different XAI methods may produce conflicting explanations, making it difficult to reach a consistent understanding of model behavior. 3. **Under - utilization of multiple AI models**: Relying on a single AI model limits the robustness of anomaly detection because different models may exhibit different strengths in data interpretation. 4. **Challenges in feature selection optimization**: Effective anomaly detection requires the identification of the most relevant features, and using a single XAI method may be limited and difficult to achieve this goal. To solve these problems, the paper proposes an XAI - based feature integration framework that combines multiple XAI methods (SHAP, LIME, and DALEX) with multiple AI models (decision tree, random forest, deep neural network, K - nearest neighbor, support vector machine, and AdaBoost) to enhance the anomaly detection ability and interpretability of autonomous driving systems. By fusing the top - ranked features identified by these XAI methods, the framework creates a more powerful and comprehensive feature set for detecting anomalies. These feature sets are then evaluated using independent classifiers (CatBoost, logistic regression, and LightGBM) to ensure unbiased performance. The main contributions of the paper include: - Proposing a new feature integration method that combines three XAI methods, SHAP, LIME, and DALEX, to enhance feature importance analysis. - Applying this feature integration method to six different AI models to identify common important features in these models and improve the robustness of the results. - Applying this feature integration method on two popular autonomous driving datasets (VeReMi and Sensor), which helps to understand the key features of autonomous vehicle safety and anomaly detection. - Using independent classifiers (CatBoost, logistic regression, and LightGBM) to evaluate the effectiveness of the fused feature set, demonstrating performance improvement and practical application value. - Making the source code publicly available for the research community to use. Through these contributions, this research aims to develop more reliable, interpretable, and safe autonomous driving systems and bridge the gap between high - performance anomaly detection and the identification of explanatory features.

XAI-based Feature Ensemble for Enhanced Anomaly Detection in Autonomous Driving Systems

On Evaluating Black-Box Explainable AI Methods for Enhancing Anomaly Detection in Autonomous Driving Systems

DAnoScenE: a driving anomaly scenario extraction framework for autonomous vehicles in urban streets

Enhancing the Safety of Autonomous Vehicles: Semi-Supervised Anomaly Detection With Overhead Fisheye Perspective

A Novel Hybrid XAI Solution for Autonomous Vehicles: Real-Time Interpretability Through LIME–SHAP Integration

Scene Semantic Anomaly Detection of Multi-Sensor in Autonomous Driving

XAI-IoT: An Explainable AI Framework for Enhancing Anomaly Detection in IoT Systems

A Multimodal Data-Driven Approach for Driving Risk Assessment

Toward interpretable anomaly detection for autonomous vehicles with denoising variational transformer

An Anomaly Behavior Analysis Framework for Securing Autonomous Vehicle Perception

Advancing Autonomous Vehicle Safety: Machine Learning to Predict Sensor-Related Accident Severity

Utilizing XAI Technique to Improve Autoencoder based Model for Computer Network Anomaly Detection with Shapley Additive Explanation(SHAP)

Anomaly Detection for Robust Autonomous Navigation.

Enhanced Scene Understanding and Situation Awareness for Autonomous Vehicles Based on Semantic Segmentation

S2DEVFMAP: Self-Supervised Learning Framework with Dual Ensemble Voting Fusion for Maximizing Anomaly Prediction in Timeseries

An Explainable Ensemble-based Intrusion Detection System for Software-Defined Vehicle Ad-hoc Networks

Lightweight Object Detection Ensemble Framework for Autonomous Vehicles in Challenging Weather Conditions

SDAC: A Multimodal Synthetic Dataset for Anomaly and Corner Case Detection in Autonomous Driving

Graph Neural Network Based Abnormal Perception Information Reconstruction and Robust Autonomous Navigation

Anomaly Detection Using XGBoost Ensemble of Deep Neural Network Models

An advanced accident avoidance system based on imbalance-control ensemble and deep learning fusion design