计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (15): 250-256.DOI: 10.3778/j.issn.1002-8331.1904-0166

• 工程与应用 • 上一篇    下一篇

基于LGMD的无人机避障方法研究

马兴灶,赵剑楠,朱齐媛,傅沁冰,胡诚,雷芳,岳士岗   

  1. 1.岭南师范学院 机电工程学院,广东 湛江 524048
    2.林肯大学 计算机科学学院,林肯 LN6 7TS
  • 出版日期:2019-08-01 发布日期:2019-07-26

Study on Obstacle Avoidance Method of UAV Based on LGMD

MA Xingzao, ZHAO Jiannan, ZHU Qiyuan, FU Qinbing, HU Cheng, LEI Fang, YUE Shigang   

  1. 1.School of Mechatronic Engineering, Lingnan Normal University, Zhanjiang, Guangdong 524048, China
    2.School of Computer Science, University of Lincoln, Lincoln LN6 7TS, UK
  • Online:2019-08-01 Published:2019-07-26

摘要: 为提高无人机避障的灵活性和可靠性,提出了一种基于LGMD(Lobula Giant Movement Detector)的无人机避障方法,通过将视场分割为上、下、左、右4个方位,形成4个方位竞争的LGMD(C-LGMD),并利用Matlab软件进行算法实现和视频仿真分析,最后将算法移植到无人机硬件系统,开展悬停测试和实时飞行实验研究。由视频仿真分析和悬停测试结果表明,该算法能有效分辨来自不同方位的障碍物,具有较好的避障性能和鲁棒性;在实时飞行测试中,无人机在室内环境中可实现三维空间有效避障,验证了该算法的可靠性。研究结果为进一步探索无人机高效、可靠避障提供参考依据。

关键词: 无人机, 竞争的小叶巨型运动检测器(C-LGMD), 避障, 生物启发, 神经网络

Abstract: In order to improve the flexibility and reliability of UAV obstacle avoidance, an obstacle avoidance method for UAV based on LGMD is proposed. By dividing the field of view into four directions:upper, lower, left and right, a Competitive LGMD(C-LGMD) is formed. The algorithm implementation and video simulation analysis are carried out by using MATLAB software. Finally, the algorithm is transplanted to the hardware system of UAV, and the hovering test and real-time flight test of UAV are developed. Video simulation analysis and hovering test results show that the algorithm can effectively distinguish obstacles from different directions, and has good obstacle avoidance performance and robustness. In real-time flight test, UAV can effectively avoid obstacles in three-dimensional space in the indoor environment, which verifies the reliability of the algorithm. The results provide a reference for further exploring the efficient and reliable obstacle avoidance of UAV.

Key words: Unmanned Aerial Vehicle(UAV), Competitive Lobula Giant Movement Detector(C-LGMD), obstacle avoidance, bio-inspired, neural network