考虑融合多传感器误差项优化方法的定位技术研究

为了实现移动机器人的精确自主定位,根据不同传感器的测量原理定义了视觉误差和惯性测量单元误差项,采用基于图优化的思想构建一个最小二乘问题的位姿估计器数学模型,并把多种传感器的误差项添加到估计器中,使用优化工具求解出最优的位姿,实现多传感器的融合定位。通过在仿真实验平台上运行公共数据集,实验结果表明单传感器的定位方案因为尺度模糊和累计漂移的问题在绝对位姿误差平均值达到7.942 m,而融合多传感器的定位方案的绝对位姿误差平均值为0.234 m,说明融合多传感器的定位方案比单传感器定位方案在定位上更加准确和鲁棒。

Research on Positioning Technology Based on Optimization Method Considering Fusion of Multi-sensor Errors

In order to achieve precise autonomous positioning of mobile robots,visual errors and inertial measurement unit error terms are defined according to the measurement principles of different sensors.Graph-based optimization is used to construct a mathematical model of the pose estimator of the least squares problem,the error terms of multiple sensors are added to the estimator,and the optimal pose is solved using optimization tools to achieve the fusion of multiple sensors.By running the common data set on the simulation platform,the experimental results show that the average value of absolute position error analysis of single sensor positioning scheme is 7.942 m due to the problem of scale ambiguity and cumulative drift,while the average value of absolute position error analysis based on multi-sensor positioning scheme is 0.234 m.It shows that the multi-sensor positioning scheme is more accurate and robust than the single sensor positioning scheme.