组合导航智能信息融合自适应滤波算法分析

Analysis of adaptive Kalman filter based on intelligent informationfusion techniques in integrated navigation system

针对当前自适应组合导航系统算法的研究趋势,总结了卡尔曼滤波技术的缺陷和利用智能融合技术提高滤波器性能的设计思想。对模糊控制自适应算法(FIR AKF)、神经网络自适应算法(NN AKF)和自适应神经网络模糊推理自适应算法(ANFIS AKF)进行了分析。着重研究FIR AKF采用滤波器新息序列和外系统状态的模糊控制器关键的模糊规则设计问题;分析NN AKF在组合导航系统模型调整、故障检测和隔离中的应用方法,并给出ANFIS AKF利用神经网络自动生成推理规则和建立自适应组合导航系统的基本方法。

The adaptive Kalman filtering (AKF) based on intelligent information fusion algorithm has currently became an effective approach to enhance the integrated navigation system's robustness and accuracy. Three main intelligent adaptive algorithms, i.e. fuzzy inference reasoning based AKF (FIR-AKF), neural network based AKF (NN-AKF) and adaptive neural network-fuzzy inference system based AKF (ANFIS-AKF), are chosen specially and studied. The design rules of fuzzy controller, the key problem in FIR-AKF, are in detail analyzed respectively based on two aspects, i.e., the innovation sequence of Kalman filter and the working states of external reference system. The NN-AKF approaches can be used to settle different limitations of traditional Kalman filtering in model modification, fault diagnoses and fault isolation, and their valid applications are provided. Using ANFIS-AKF to generate the fuzzy inference rules in an adaptive INS is the final concern of this paper, and the basic design procedures are given.