扑翼机器人机载视觉避障系统设计
Design of an On-Board Vision-Based Obstacle Avoidance System for Flapping-Wing Robots
付强 1张祚卿 1刘胜南 1吴晓阳 1贺威1
作者信息
- 1. 北京科技大学智能科学与技术学院,北京 100083,中国;北京科技大学人工智能研究院,北京 100083,中国;北京科技大学智能仿生无人系统教育部重点实验室,北京 100083,中国
- 折叠
摘要
扑翼机器人是一种高仿生度的飞行器,它可以执行军事侦察和民用监测任务.在执行这些任务时,避障是一种确保扑翼机器人安全的必要功能.本文设计了一种基于单目视觉的扑翼机器人自主避障系统,其中所有图像的处理计算均使用机载处理器实现.在这个系统中,机载处理器的质量被减至48 g,以便于扑翼机器人可以携带它稳定飞行.该系统的工作流程可以分为以下几步:首先,图像采集模块获取周围环境的视频图像;然后,机载处理器通过处理扑翼机器人第一视角的光流信息来计算方向舵角度和转弯方向;最后,飞控板接收计算结果并控制扑翼机器人避开障碍物.地面站对扑翼机器人飞行过程进行实时监测,实验结果验证了本文所设计的避障系统的有效性.
Abstract
The flapping-wing robot(FWR)is a kind of high-biomimetic aerial vehicle that can perform military reconnaissance and civil monitoring missions.When performing these missions,obstacle avoidance is a necessary function to ensure safety of FWRs.In this paper,an autonomous monocular-vision-based obstacle avoidance system is designed for FWRs,where all the image processing computations are implemented by using an on-board computer.In this system,the weight of the on-board computer is reduced to 48 g so that the FWR could fly properly.The workflow of the system can be divided into the following steps.First,the image acquisition module captures videos of the surrounding environment.Second,the on-board computer calculates the rudder angle and turning direction by processing the optical flow information from the first-person view.Finally,the flight control board receives the calculated results and controls the FWR to avoid obstacles.A ground station performs real-time monitoring of the FWR flight process,and experimental results demonstrate the effectiveness of the obstacle avoidance system designed in this paper.
关键词
扑翼机器人/视觉/避障/机载处理器/光流Key words
flapping-wing robots/vision/obstacle avoidance/on-board computer/optical flow引用本文复制引用
基金项目
国家自然科学基金(62225304)
国家自然科学基金(61933001)
国家自然科学基金(62173031)
国家自然科学基金(U20A20225)
Interdisciplinary Research Project for Young Teachers of Fundamental Research Funds for the Central Universities(FRF-IDRY-22-029)
Beijing Top Discipline for Artificial Intelligent Science and Engineering,University of Science and Technology Beijing()
出版年
2024
国家科技期刊平台
NSTL
万方数据