摘要
针对移动机器人在未知复杂环境中动态目标跟踪存在的数值不稳定、计算量大和精度较差等问题,提出基于平方根容积卡尔曼滤波的移动机器人动态目标跟踪算法(SR-CKF-SLAM-OT)。该算法的系统状态由地图环境特征、机器人和目标作为一个整体构成。建立目标和机器人的动态模型进行预测、数据关联和更新,在更新过程中直接传递目标状态均值和协方差矩阵的平方根因子,降低了计算的复杂度。此外,通过数据关联环节能够有效的降低伪观测值对系统状态估计的影响。仿真结果表明:相比基于EKF的动态目标跟踪算法,所提出的动态目标跟踪算法目标和机器人均方根误差分别降低了36.3%和38.2%,SR-CKF-SLAM-OT算法有效地满足了移动机器人动态目标跟踪的需求。
When exploring the unknown complex environment, the single mobile robot has many problems, such as unstable numerical value, large computation capability and poor accuracy, therefore, a dynamic target tracking algorithm based on square root cubature Kalman filter （SR-CKF-SLAM-OT） is proposed for mobile robots in this paper. In terms of the algorithm, the environmental characteristics of map, the mobile robot and the target as a whole constitute the system state. Moreover, the system state was predicted, associated and updated according to the dynamic model of the target and the robot, meanwhile, the mean value of target state and the square root factor of the covariance matrix were transmitted directly during the updates, which reduced the computation complexity. In addition, the influence of the pseudo observation value on the system state estimation can be effectively reduced by the data association. The simulation results show that compared with the dynamic target tracking algorithm based on EKF, the dynamic target tracking algorithm based on square root proposed in this paper make the root mean square errors from target and robot decrease 36.3% and 38.2% respectively, which means that the SR-CKF- SLAM-OT algorithm effectively satisfies the requirement of dynamic target tracking for mobile robots.
出处
《电子测量与仪器学报》
CSCD
北大核心
2016年第8期1198-1205,共8页
Journal of Electronic Measurement and Instrumentation
基金
2016年安徽高校自然科学研究项目(KJ2016A794)资助
