摘要
GPS动态定位数据的处理广泛采用卡尔曼滤波技术,而应用卡尔曼滤波要求运动模型准确可靠,但由于载体真实运动的复杂多变,任何单一模型都难以全面描述,致使单一模型的滤波都容易出现模型误差。针对这一问题,将机动目标跟踪领域广泛应用的交互式多模型算法引入到车载导航中。通过分析车辆的运动特点,选取匀速直线模型和当前统计模型进行交互;同时考虑到车载终端计算能力有限,将状态变量在各方向解耦。仿真显示,在机动时改进的算法和单一模型的自适应算法基本相当,但在非机动时改进的算法明显占优。
Kalman filters were used extensively in data processing of GPS dynamic positioning, while the application of Kalman Filters required that the dynamic model was practical and reliable, but in actual the vehicle motion was complex and changeful, which made it was impossible to express the motion by single model, anyhow model errors often occured in the filters based on single model. In view of the problem, the interacting multiple model (IMM) algorithm, which was used extensively in maneuvering target tracking, was introducted to vehicle navigation. The constant velocity (CV) model and current statistics (CS) model were selected according to vehicle movement characteristics, and the state variables of every direction were decoupled in order to decrease calculation amount. Simulation results showed that, as the vehicle was maneuvering, the performances of the algorithm developed and the adaptive algorithm based on single model were almost same, but while the vehicle was non-maneuvering, the former was superior to the latter obviously,
出处
《测绘科学技术学报》
北大核心
2009年第3期170-173,共4页
Journal of Geomatics Science and Technology