A Hybrid Explainable AI Framework for Enhancing Trust and Transparency in Autonomous Vehicles

This paper introduces a new hybrid framework for Explainable Artificial Intelligence (XAI) aimed at enhancing the clarity and interpretability of decision-making processes in autonomous vehicles (AVs). A new hybrid framework is developed that integrates symbolic reasoning and Grad-CAM explanations to produce an in-depth understanding of the actions taken by AV. The experimental evaluation of the framework is done with the integration of real-world datasets like Waymo Open, KITTI, and BDD100K datasets along with virtual environments such as CARLA simulator. The outcome is that the framework delivers high fidelity explanations with an average Intersection over Union of 0.81 along with an interpretability score of 8.8 out of 10, which indicates the clarity and practicality of the explanations provided. Apart from that, the proposed framework delivers the real-time performance of 42 frames per second. Therefore, it is appropriately appropriate for safety-critical deployment in autonomous systems. More importantly, user trust is highly amplified as presented with inputs from subject matter experts and end-users. This includes erratic pedestrian behavior. This study is showcasing the hybrid XAI framework, which can make AV technology transparent, safe, and palatable, hence turning into an invaluable asset for applications involving AV in the real world.