摘要
针对现有小型无人飞行器在复杂未知环境中执行自主探索任务时,存在机载端GPU计算资源不足,在线建图速度慢、探索效率低的问题.本文在传统欧几里德符号距离场(ESDF)方法的基础上,融合光线跟踪原理,加速构建ESDF地图,以提高飞行器在复杂未知环境中的探索效率.首先使用整数运算提高光线跟踪遍历体素的速度,从而加速体素占用概率的更新;然后通过调整哈希数据结构,减少地图占用内存;最后使用广度优先搜索算法(BFS)实现地图更新与融合.为验证本文方法的有效性,分别在公开数据集、仿真环境和真实环境下,与当前前沿的建图方法进行对比,实验结果表明,本文所提出的方法在三种不同情况下的地图平均更新时间分别减少了72.37%、60.80%和56.79%,显著提高了建图速度,为小型无人飞行器在线探索奠定基础.
For the existing small unmanned aerial vehicles(UAVs)performing autonomous exploration tasks in complex and unknown environments,there are problems such as insufficient computing resources of the airborne GPU,slow online mapping and low exploration efficiency.Based on the traditional Euclidean signed distance field(ESDF)method,this paper integrates the ray tracing principle and accelerates the construction of ESDF maps to improve the exploration efficiency of aircraft in complex and unknown environments.Firstly,integer operation is used to improve the speed of ray tracing voxel traversal,thus speeding up the update of voxel occupancy probability;Then adjust the hash data structure to reduce the memory occupied by the map;Finally,the breadth first search algorithm(BFS)is used to update and fuse the map.In order to verify the effectiveness of the method proposed in this paper,it is compared with the current cutting-edge mapping methods in the open dataset,simulation environment and real environment respectively.The experimental results show that the average map update time of the method proposed in this paper in three different situations is reduced by 72.37%,60.80%and 56.79%respectively,which significantly improves the mapping speed and lays a foundation for online exploration of small unmanned aerial vehicles.
作者
唐嘉宁
刘志聪
李孟霜
彭志祥
谢翠娟
陈云浩
TANG Jia-ning;LIU Zhi-cong;LI Meng-shuang;PENG Zhi-xiang;XIE Cui-juan;CHEN Yun-hao(School of Electrical and Information Technology,Yunnan Minzu University,Kunming 650031,China;Institute of Unmanned Autonomous System,Yunnan Minzu University,Kunming 650031,China)
出处
《陕西科技大学学报》
北大核心
2023年第4期158-165,共8页
Journal of Shaanxi University of Science & Technology
基金
国家自然科学基金项目(61963038)。
关键词
地图构建
光线跟踪
符号距离场
梯度优化
mapping
ray tracing
signed distance fields
gradient optimization
