Probabilistic neural network multi-model Kalman filter navigation algorithm

• Main Authors: Liang Longkai, Zhang Liying, He Wenchao, Lv Xuhao
• Format: Article
• Language: zho
• Published: National Computer System Engineering Research Institute of China 2018-06-01
• Series: Dianzi Jishu Yingyong
• Subjects: navigation; extend Kalman filter(EKF); probabilistic neural network(PNN); BDS; GPS
• Online Access: http://www.chinaaet.com/article/3000083941

Interacting multiple model extended Kalman filter(IMM-EKF) algorithm is a sub-optimal algorithm which can solve the positioning problem in which the motion model is uncertain. But this method still gets sub-optimal solution and wastes computational resources when the carrier does the motion of which the model is certain. Aiming at this kind of defects of IMM-EKF, the off-line training probabilistic neural network model is adopted to judge the classification of current motion model in real time. We choose to operate with the single corresponding model when the motion model is in the state of certainty, and choose the IMM-EKF algorithm when the motion model is in the uncertain state. Thus it not only ensures the positioning accuracy, but also reduces the unnecessary computation burden. Simulation experiments verify the validity and accuracy of the algorithm, while the contrast test verifies the advantages in accuracy of the new algorithm compared with IMM-EKF algorithm.