基于改进扩展卡尔曼滤波的姿态解算算法研究-计算机科学.pdfVIP

Vo1. 44 No. 9 第 44 卷第9 期 计算机科学 2017 年 9 月 COMPUTER SCIENCE Sep.2017 基于改进扩展卡尔曼滤波的姿态解算算法研究 冯少江1 徐泽宇2 石明全2 王晓东2 (重庆邮电大学计算机科学与技术学院 重庆 400065)1 (中国科学院重庆绿色智能技术研究院 重庆 400714)2 摘 要 为了解决标准扩展卡尔曼滤波器(EKF)在多旋翼元人机姿态解算中精度较低的问题，提出了一种改进扩展 卡尔曼滤泼算法(BPNN-EKF) ，使得解算精度得到较大提升。针对 EKF 存在预测模型参数要求具有先验已知性，在 工程实践中难以获得准确的参数，以及标准 EKF 对非线性系统采用线性化模型带来的误差等问题，利用神经网络的 非线性映射能力和自适应能力对标准 EKF 的估计值进行补偿，减小模型以及滤波参数误差对最优估计值的影响，从 而提高最优估计精度。仿真实验证明，BPNN-EKF 对多旋翼元人机姿态解算精度的提升具有显著作用。 关键词 扩展卡尔曼滤泼器，姿态解算算法，非线性系统，BP 神经网络 中固法分类号 TP391 文献标识码 A DOI 10. 11896/j. issn. 1002-137X. 2017. 09. 042 Research on Attitude Algorithm ßased on Improved Extended Calman Filter FENG Shao-jiang1 XU Ze-yu2 SHI Ming-quan2 WANG Xiao-dont (College of Computer Science and Technology ,Chongqing University of Posts and Telecommunications ,Chongqing 400065 ,China)1 (Research Institute of Chongqing Green Intelligent Techno1ogy ,Chinese Academy of Sciences ,Chongqing 400714 ,China)2 Abstract In order to solve the problem that standard extended Kalman filter (EKF) in the multi-rotor UAV attitude solver has lower accuracy ,an improved extended Kalman filter algorithm (BPNN-EKF) was proposed to improve the accuracy greatly. EKF prediction model parameters require the presence of priori known properties ,but in engineering practice it is difficult to obtain accurate parameters. And nonlinear systems using linear model will cause error problem for standard EKF. Aiming at above problems ,we used nonlinear mapping ability of neural network and adaptive ability to compensate the estimated value of the standard EKF ,reduce the impact of the model as well as f