Technical
No Paper Title Author, Year Objective Methodology Research Findings
Limitations
Balanced High
Improve real-time Hybrid LIME + interpretability and computational cost
Hybrid LIME–SHAP
1 H.A.T. et al., 2024 interpretability of SHAP feature computational for complex
XAI for AV
AV decisions importance efficiency; improved segmentation
explanation stability models
Taxonomy- Identified model-
Lack of unified
Systematic Review Saranya & Survey XAI models based agnostic and model-
2 evaluation
of XAI Subhashini, 2023 and applications classification of specific explanation
benchmark
XAI methods approaches
End-to-End Learn steering CNN can predict Black-box decision
Bojarski et al., CNN end-to-
3 Learning for Self control directly steering angle directly process not
2016 end learning
Driving from images from camera images interpretable
Detect abnormal Identifies key features Requires high-
Driver Anomaly SHAP +
Chengula et al., driver behaviour such as yawning, head quality labeled
4 Detection using ensemble ML
2023 using explainable pose influencing behavioural
XAI models
ML prediction dataset
Rule-based vs Example-based Not optimized for
Rule vs Example Compare
5 Keane et al., 2021 example-based explanations improve real-time driving
based XAI explanation types
XAI user understanding decisions
No Paper Title Model / Method Dataset Key Findings Research Gap Problem Statement
Identifies feature
Explain object GradCAM, SHAP, High computational
XAI for Object contribution in
6 Hogan, 2025 detection DetDSHAP with complexity for real-
Detection AV bounding box
decisions in AV YOLO time deployment
detection
XAI Safe & Review XAI improves trust, Lack of standard
Rasouli et al., Systematic
7 Trustworthy explainability for debugging and metrics for
2023 literature review
Driving Review safe AV validation explanation quality
Improves
XAI in Improve Decision trees, Trade-off between
transparency in lane
8 Autonomous IJFMR, 2022 transparency in CAM, LRP interpretability and
change and obstacle
Systems AV decisions explanations performance
avoidance decisions
Scene Explain scene Highlights important Sensitive to image
Inception U-Net GradCAM + CNN
9 Understanding segmentation in visual regions noise and lighting
model segmentation
using GradCAM Indian traffic influencing decisions variations
XAI Object Explain YOLO
Identifies critical
Detection Siganos et al., predictions in EigenCAM Limited to visual
10 pixels affecting
Automotive 2025 automotive heatmaps explanation only
detection results
Sector environment
Predict accidents Deep learning + Predicts accident
XAI Accident Karim & Qin, Requires large
11 early with GradCAM 4.57 sec earlier with
Anticipation 2022 labelled video dataset
explanation saliency maps 94% precision
Problem
No Paper Title Model / Method Dataset Key Findings Research Gap
Statement
Align AI Improves
Human
Human Attention Guided attention with Attention maps + interpretability
12 Kim et al., 2022 annotation
XAI human CNN of computer
required
reasoning vision model
Provide Survey of Identifies future
Explainable AI for
Atakishiyev et taxonomy and GradCAM, LIME, directions for Lack of unified XAI
13 Autonomous Driving
al., 2024 framework for SHAP, RL-based trustworthy AV architecture
Overview
XAI AV explanation systems
Explains feature
Explain High inference
Explainable AI for Object YOLO + SHAP KernelSHAP + contribution of
14 bounding box time for SHAP
Detection framework saliency maps detected
predictions methods
objects
Generate Limited
Hybrid Human- SHAP + rule-
Hybrid XAI interpretable Improves model generalization to
15 interpretable DL based
framework explanation for transparency multi-modal
explanation explanation
DL sensors
Improves
Predict crash Dependent on
Early anticipation of traffic Spatio-temporal Attention based situational
16 risk with sensor fusion
accident DL model CNN-LSTM awareness for
explanation quality
ADAS