一种改进的高斯混合概率假设密度SLAM算法

An improved Gaussian mixture PHD SLAM algorithm

针对高斯混合概率假设密度SLAM(GMPHD-SLAM)算法存在的估计精度低和计算代价高的问题,提出一种无迹高斯混合概率假设密度SLAM算法(unscented-GMPHD-SLAM).其主要特点在于:将无迹卡尔曼滤波器应用于机器人位姿粒子权重计算及概率假设密度更新过程中,可提高算法整体估计性能;将更新的高斯项按照传感器视域分类,有效降低了算法计算量.通过仿真实验,将所提出算法与传统PHD-SLAM算法进行比较,结果表明该算法在提高估计精度和降低计算负担方面是十分有效的.

For two problems in Gaussian mixture probability hypothesis density SLAM（GMPHD-SLAM） algorithm of low estimation accuracy and high computational cost,the GMPHD-SLAM algorithm based on unscented transform,called unscented-GMPHD-SLAM,is proposed.The main contribution lies that：the unscented Kalman filter is used in the calculation of particle＇s weight and PHD update process,which improves the performance of the algorithm;the updated Gaussian components are classified based on the sensor＇s field of view（FoV）,which reduces the computational cost.The proposed algorithm is compared with the traditional PHD-SLAM algorithm.The results show that the proposed algorithm is effective in accuracy improvement and reduction of computational cost.