The performance of the maximum correntropy criterion filter is affected by the kernel size, while the present kernel size adaptive methods are prone to excessive convergence. To achieve the adaptive adjustment for the kernel size without excessive convergence problems, an adaptive multikernel size-based maximum correntropy cubature Kalman filter (CKF) is proposed. The adaptive factor for each measurement element is constructed, and the kernel size is adjusted within a reasonable range based on the adaptive factor. Then the correntropy matrix with the adaptive multikernel size is established to achieve measurement-specific outliers processing in state estimation. The target tracking simulation and the cooperative positioning experiment are conducted to verify the proposed method. The results demonstrate that the proposed adaptive multikernel size-based maximum correntropy CKF (AMCCKF) can effectively optimize the kernel size for different noises and is more convenient for selecting tuning parameters, thus effectively achieving robust state estimation against outliers while ensuring filtering accuracy.

In this research, a novel autonomous underwater vehicle (AUV) cooperative positioning algorithm is proposed to solve the implementation problem of multi-sensor-fusion applications. Different from the traditional methods [i.e., the extended Kalman filter (EKF), unscented Kalman filter (UKF), and iteration extended Kalman filter (IEKF)], which have large linearity error under the condition of nonlinear observation equation when multiple AUV are cooperative positioning, the proposed algorithm utilized the Baysis filter to solve the AUV cooperative problem. Factor graph and sum-product (FGS)-based cooperative positioning algorithm is established to mathematically implement the Bayse filter by converting the global function estimation problem into a local function sum-product estimation problem. Furthermore, to improve the performance of the proposed algorithm, a robust data processing method is presented by introducing a transform matrix to the estimated position information. To demonstrate and verify the proposed methods, the simulation and real tests in different scenarios are performed in this research. Compared with the traditional EKF, UKF, and IEKF cooperative positioning algorithm, the positioning error of the proposed improved FGS (IFGS) cooperative positioning algorithm is obviously smaller than that of the other three algorithms. Moreover, the IFGS algorithm can reduce the complexity of the algorithm, available improving the computational speed of the whole system. This proposed algorithm has important theoretical and practical value for the both industry and academic areas.

Along with the development of computer technology and informatization, the unmanned vehicle has become an important equipment in military, civil and some other fields. The navigation system is the basis and core of realizing the autonomous control and completing the task for unmanned vehicles, and the Strapdown Inertial Navigation System (SINS) is the preferred due to its autonomy and independence. The initial alignment technique is the premise and the foundation of the SINS, whose performance is susceptible to system nonlinearity and uncertainty. To improving system performance for SINS, an improved initial alignment algorithm is proposed in this manuscript. In the procedure of this presented initial alignment algorithm, the original signal of inertial sensors is denoised by utilizing the improved signal denoising method based on the Empirical Mode Decomposition (EMD) and the Extreme Learning Machine (ELM) firstly to suppress the high-frequency noise on coarse alignment. Afterwards, the accuracy and reliability of initial alignment is further enhanced by utilizing an improved Robust Huber Cubarure Kalman Filer (RHCKF) method to minimize the influence of system nonlinearity and uncertainty on the fine alignment. In addition, real tests are used to verify the availability and superiority of this proposed initial alignment algorithm.

With the advantages of anti-interference and no accumulated error, bionic polarization heading measurement has important military significance and research value for autonomous navigation. However, the heading robustness is seriously affected by the accuracy of skylight polarization model and the existing methods all use the Rayleigh scattering model without considering the atmospheric depolarization. Therefore, this paper proposed an innovative method of polarization heading measurement based on Berry model to considering the influence of depolarization neutral points. And this model is improved by controlling the neutral points with correction coefficients to realize high-robustness heading measurement. On this basis, the improved Berry model is further used to calibrate the sensor parameters outdoors, which separates the skylight polarization model error from the sensor error without expensive instruments and complicated processes, and improves the heading measurement accuracy. In the experiments with changing solar altitude, the average of heading error STD after field calibration is 20.86% lower than that of Berry model and 92.85% lower than that of Rayleigh model, which shows great advancement in actual measurements.

Integrity quantifies the confidence level in the position solution and is essential for positioning safety-critical applications. To monitor the integrity with a protection level (PL) of the multiple fault biases for a sequential filtering framework in challenging environments, a novel robust sequential IM approach is proposed. First, the impact of estimation consistency on PL is analyzed theoretically, and a front-end Student’s t distribution-based filter variant is adopted to provide consistent posterior estimates for constructing dynamic IM regression models and suppressing outliers. Then, under the multiple fault biases assumption, a maximum eigenvalue-based PL is calculated in a sequential filtering framework. Finally, two GNSS/INS in-vehicle experiments are conducted to validate the proposed method. The results indicate that the proposed method has a higher PL reliability (100% and 98.13%) than other methods, and did not suffer any hazardous misleading cases during the experiment. Therefore, the proposed method can assess the confidence of the position estimation of GNSS/INS and effectively monitor position integrity in challenging environments.