Explainable AI (XAI) for Wireless Communications: UWB Radar for Zone-based Obstacle Detection

The evolution of artificial intelligence (AI) solutions in various fields has raised concerns about managing and understanding these solutions. Ultra-wideband (UWB) radar, known for its precision and reliability, is extensively applied in localization scenarios. However, the increasing use of AI in this field has led to concerns regarding the ambiguity of analytical reasoning in AI solutions. To address these concerns, XAI has been introduced to make these solutions more trustworthy and understandable. This study explores the integration of XAI within UWB radar technology, emphasizing the identification of target zones through zone-based obstacle detection. We investigate the integrated gradient method and its application in a convolutional neural network (CNN) model. A significant portion of the study is dedicated to analyzing the distance and angle for zone-based obstacle detection facilitated by the XAI -empowered CNN model. This analysis aims to provide deep insights into the reasoning mechanics of AI solutions in UWB radar technology, thereby addressing concerns in the field.