基于二元正态分布匹配和非线性优化的激光SLAM研究

Laser SLAM Rearch Based on Binary Normal Distribution Matching and Nonlinear Optimization

AGV自导航的关键之一是拥有高精度地图,因此激光SLAM(同步定位与建图)地图构建系统非常重要。针对现有算法精度和效率的不足,该研究建立了一个适用于AGV上的激光SLAM系统,并进行了测试和实现。首先,系统由激光雷达获取激光帧,进行滤波处理,并由第一帧激光帧初始化自由概率地图;其次,将t-1时刻激光帧进行二元正态分布化,由t时刻激光帧在t-1时刻激光帧的相关区间进行投影和概率打分,找到离散最优解;其次,由二元正态分布概率密度函数和双三次插值的自由概率残差构建目标代价函数,在离散最优解位置进行非线性优化,得到最优解位姿;最后,将激光帧投影到地图中的最优解位姿,进行地图更新。结果表明,该算法匹配精度高,算法耗时少,满足AGV工业精度和实时性要求,对AGV在实际工业环境的应用具有重要意义。

One of the keys to the self-navigation of AGV is to have a high-precision map,so the laser SLAM(synchronous location and mapping)map construction system is very important.In this research,a laser SLAM system suitable for AGV was established,tested and implemented.First,the system obtains the laser frame from the laser radar,and filters the laser frame.The free probability map is initialized from the first laser frame;secondly,the laser frame at time t-1 is binary normalized,and the discrete optimal solution is found by projection and probability scoring of the laser frame at t in the correlation interval of the laser frame at t-1;after that,the objective cost function is constructed from the probability density function of the bivariate normal distribution and the free probability residual of the bicubic interpolation.The optimal solution position is iteratively optimized to obtain the optimal solution pose;finally,the laser frame at time T is projected to the optimal solution pose in the map,and the map is updated.