基于梯度下降的自适应姿态融合算法  

Adaptive attitude fusion algorithm based on gradient descent

作　　者：陈卓 任久春[1] 朱谦[1] CHEN Zhuo;REN Jiu-chun;ZHU Qian(School of Information Science and Engineering,Fudan University,Shanghai 200433,China)

机构地区：[1]复旦大学信息科学与工程学院,上海200433

出　　处：《传感器与微系统》2019年第3期124-126,共3页Transducer and Microsystem Technologies

基　　金：上海市科学技术委员会科研计划项目(14231202102);上海市体育科技"综合计划"资助项目(16Z019)

摘　　要：针对第一代帆船姿态测量系统中采用的自适应卡尔曼滤波算法作用范围有限、动态性能较差等缺陷,基于微机电系统(MEMS)惯性传感器与梯度下降姿态融合算法提出了两种自适应方法,分别根据当前时刻之前N个采样点的平均运动加速度与加速度的变化量设计自适应控制因子,得到稳定的动态梯度下降步长。实验结果表明:两种算法性能均优于自适应卡尔曼滤波与单点加速度抑制法,其中,基于加速度增量的控制算法更加符合高速运动状态下加速度剧烈变化的实际规律,测量性能达到最优,符合海面帆船运动船体姿态测量的实际需求。Aiming at defect of the action range of the adaptive Kalman filtering algorithm used in the first generation of the sailing attitudes,measurement system is limited and dynamic performance is poor,two adaptive methods are presented based on micro-electro-mechanical system(MEMS)inertial sensors and gradient descent attitude fusion algorithm.Adptive control factors are designed according to the average motion acceleration and the variation of acceleration of N sampling points before the current time respectively to obtain stable dyamic gradient descent steps.Experimental results indicate that the performances of the two algorithms are prior to the adaptive Kalman filtering and single-point acceleration restrained algorithms,among them,the acceleration increment-based control algorithm is proved to be consistent with real rule of drastic acceleration changes in high-speed motion state,measurement performance achieve the optimum which meets the real requirement of sailing attitude measurement.

关 键 词：姿态解算 多传感器测量 梯度下降算法 自适应滤波融合 加速度控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]