ROS2 Implementation of Object Detection and Distance Estimation using Camera and 2D LiDAR Fusion in Autonomous Vehicle

Engineering notes

Key topics: autonomous driving, autonomous vehicle, object detection, lidar, sensor fusion, perception. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

The performance of ‘perception’ is important to enable autonomous vehicles to drive safely and efficiently under all environmental conditions. This paper proposes a methodology for object detection and distance estimation within a ROS2 environment, utilizing the fusion of 2D LiDAR and camera sensors. It highlights potential solutions for future autonomous driving using ROS2, emphasizing how sensor fusion can enhance the reliability and safety of autonomous vehicles. A camera, LiDAR, and vehicle were simulated in Gazebo, allowing for the fusion of sensor data to estimate distances and detect objects. The system includes four integrated nodes. The first, a camera node, publishes image topics. The second, a LiDAR node, publishes data on the location and distance of obstacles. The third, a YOLOv8 node, subscribes to image topics and publishes information about the bounding box, accuracy, class name, and ID. Lastly, the projection node combines data from the previous nodes to detect objects and measure their distances. According to our experimental results, the average error in distance estimation is only $0.75 \%$. We also tested the frequency of topic messages and discovered that, on average, messages related to object detection and distance calculation are issued 33 times per second on NVIDIA Titan X. It is required to optimize the embedded board model and procedures in order to implement this fusion method in practical applications.

Links and sources

• Official / arXiv page

Comments