Scholarlink

<p>As artificial intelligence (AI) advances in autonomous driving, autonomous navigation systems should be integrated into educational platforms to provide students practical experience in robotics and machine learning. In this paper, we propose the detailed study of autonomous driving within the Webots simulation platform, specifically designed for student experiments. We incorporate advanced techniques such as obstacle detection, precise lane following, and intelligent traffic light recognition based on an improved YOLOv8 architecture for robust detection of road signs and traffic lights. We utilized the Feature Pyramid Network (FPN) and Path Aggregation Network (PAN) which are integral to YOLOv8’s multi-scale feature extraction and enhance their efficiency by replacing standard convolutions with depthwise separable convolution in the neck of network. This modification reduces the computational overhead while preserving feature representation quality, leading to faster inference speeds and improved small-object detection accuracy, thus improving autonomous navigation performance in simulated scenarios. To consider the performance of our proposed model, we compare it against widely used and well-established object detection models on KITTI and RF100 datasets. The experimental results demonstrate that the proposed model exhibits a good advantage in the field of autonomous driving and boosting student engagement and preparing them for future careers in AI and robotics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced autonomous driving within Webots simulation for student experiments

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Emmanuel Mumba,
  • Abubakar Sulaiman Gezawa,
  • Chibiao Liu

摘要

As artificial intelligence (AI) advances in autonomous driving, autonomous navigation systems should be integrated into educational platforms to provide students practical experience in robotics and machine learning. In this paper, we propose the detailed study of autonomous driving within the Webots simulation platform, specifically designed for student experiments. We incorporate advanced techniques such as obstacle detection, precise lane following, and intelligent traffic light recognition based on an improved YOLOv8 architecture for robust detection of road signs and traffic lights. We utilized the Feature Pyramid Network (FPN) and Path Aggregation Network (PAN) which are integral to YOLOv8’s multi-scale feature extraction and enhance their efficiency by replacing standard convolutions with depthwise separable convolution in the neck of network. This modification reduces the computational overhead while preserving feature representation quality, leading to faster inference speeds and improved small-object detection accuracy, thus improving autonomous navigation performance in simulated scenarios. To consider the performance of our proposed model, we compare it against widely used and well-established object detection models on KITTI and RF100 datasets. The experimental results demonstrate that the proposed model exhibits a good advantage in the field of autonomous driving and boosting student engagement and preparing them for future careers in AI and robotics.