文献类型：会议论文

英文题名：Emergency Obstacle Avoidance Based on Gradient Descent Distance for Self-driving Vehicles

作者：Yao, Yongqiang[1,2]; Ma, Nan[1,2]; Li, Jiahong[2]; Wu, Zhixuan[1,2]; Zhang, Guoping[1,2]

第一作者：Yao, Yongqiang

机构：[1] Beijing Key Laboratory of Information Service Engineering, Beijing Union University, Beijing, 100101, China; [2] College of Robotics, Beijing Union University, Beijing, 100101, China

第一机构：北京联合大学北京市信息服务工程重点实验室

会议论文集：Proceedings of 2021 International Conference on Autonomous Unmanned Systems, ICAUS 2021

会议日期：September 24, 2021 - September 26, 2021

会议地点：Changsha, China

语种：英文

外文关键词：Autonomous vehicles - Gradient methods - Object detection - Simulation platform - Virtual reality

摘要：Computer vision plays an important role in vehicle or pedestrian detection and distance estimation tasks for self-driving car in the surrounding environments. However, due to the mismatch of both the object detection and the intrinsic parameters of the camera model, the fluctuation of the distance from a camera to a marker is actually a very well studied problem in straightforward and succinct techniques, e.g., the triangle similarity technique. In order to solve this problem, a gradient descent distance estimation method is proposed to realize the solution of self-driving vehicles emergency obstacle avoidance. In the algorithm, the gradient descent approach is used to fit the pixel difference of the object with the actual distance for the difference between the object and the camera in the image. The fitting results are used to calculate the distance between the object and the camera in real-time to solve the problem of unstable distance calculation. The performance of this method is better than triangle perspective transformation in distance test on static and dynamic objects. The proposed method is applied in the virtual simulation of 2021 World Intelligent Driving Challenge, which used simulation platform of AD Chauffeur to avoid obstacles in emergencies. This method not only realizes the interaction between vehicle and environment effectively, but also improves the stability of distance change between vehicles and front vehicles significantly, and reduces the influence of perception error on interaction. ? 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.