Yantao Lu

Aug 28, 2024 | mdpi.com | Yilan Li |Xing Fan |Shiqi Sun |Yantao Lu

1. IntroductionIn the context of simultaneous localization and mapping (SLAM) of unmanned vehicles tasks, deep neural networks (DNNs) have achieved remarkable performance in providing proficient and adaptable solutions for perception modules. These modules encompass multiple computer vision sub-tasks, including object detection, semantic segmentation, and agent localization. However, recent studies have highlighted that DNNs can be vulnerable to specific perturbation patterns during inference.