论文部分内容阅读
针对基于图论的同时定位与制图中,非线性约束方程组维数随机器人运行距离和时间的增加而不断增大的问题,提出一种基于信息增益的图精简算法.该算法通过评估精简前后特征点信息矩阵相对变化,删除观测信息量小于给定阈值的机器人位姿及相应的观测,达到显著简化优化问题的目的.根据测量球形协方差矩阵假设,给出了信息增益的精确和近似计算方法.通过恢复性图剪枝方法,确保图精简过程中的连通性.蒙特卡洛仿真和开源实验数据计算结果表明,在不引入明显的优化误差前提下,该方法可实现位姿和特征点90%的精简,显著提高图优化效率.
In order to solve the problem of simultaneity mapping and cartography based on graph theory, the dimensionality of nonlinear constrained equations increases with the increase of the running distance and time of the robot, and a graph reduction algorithm based on information gain is proposed. The relative change of the information matrix of feature points, the deletion of the pose and the corresponding observation of the robot whose observational information is less than a given threshold can simplify the optimization problem significantly.According to the spherical covariance matrix hypothesis, the accurate and approximate calculation of information gain Method.Recordabilty graph pruning method is used to ensure connectivity during graph reduction.Calculation of Monte Carlo simulation and open-source experimental data show that this method can realize the pose and feature points without introducing obvious optimization error 90% of the streamlining, significantly improve the graph optimization efficiency.