Research on the application of ensemble neural network in integrated navigation system

In this article, a prediction method based on a radial basis function neural network and an improved unscented Kalman filter, is proposed to improve the accuracy of position and velocity of the Global Positioning System/Strapdown Inertial Navigation System (GPS/SINS) integrated navigation system, especially in the presence of GPS signal outages. The improved unscented Kalman filter based on the adaptive theory is adopted to enhance the positioning accuracy of the GPS/SINS integrated navigation system when a GPS signal is available. A radial basis function neural network and a non-stationary time series analysis are used to predict and compensate for the positioning error of the GPS/SINS integration in the presence of GPS signal outages to improve the reliability of the navigation system. The effectiveness of the proposed method is verified by the simulation experiment and vehicle test. The simulation and test results show that the proposed method can improve the position accuracy and velocity accuracy in different GPS outage time periods.

For land-based navigation, Euler angles are often used in INS/GPS integrated navigation systems. However, the trigonometric operations required in the updates and forming of the rotation matrices for transforming the INS measurements from the body frame to the navigation frame turns the system model to be highly nonlinear. Besides, using low-cost MEMS-based IMUs, the gyroscope bias errors must be correctly estimated and compensated, which makes the nonlinearity problem a critical one. In this contribution, three Kalman filtering methods (i.e., Extended Kalman filter with simplified system model, Extended Kalman filter with linearized system model and Unscented Kalman filter with nonlinear system model) are utilized in INS/GPS tightly-coupled integration. Simulations and field experiments are conducted. Numerical results are compared in terms of both estimation accuracy and processing time.

Aiming at the difficulties of high-precision position and orientation system in aerophotogrammetry system, drift error in traditional Strap-down Inertial Navigation system, noise characteristic mutarions in integrated navigation system, an algorithm based on IMM-SINS/GNSS Kalman filter is proposed. The algorithm takes the advantages of SINS's autonomy and anti-interference characteristics, GNSS's high absolute precision and error-compensating, IMM's multi-model existence and noise adaption characteristic. The error is estimated by using the improved Kalman filter supported by SINS/GNSS integrated navigation system. Simulation results show that the location and attitude error is effectively limited, verify the feasibility of this scheme and the efficiency of the algorithm, showing that this algorithm is valuable in navigation and aerophotogrammetry field.

Filtering algorithm is One of the key technologies that affect the performance and reliability of INS/GPS integrated navigation systems. The paper adopts hybrid filtering algorithm based on neural network. Using the powerful learning and identification capabilities of the neural network, the hybrid filtering algorithm solve the problem of the Kalman filtering precision degradation caused by the system noise and the measurement noise and the absence of the GPS signal. The simulation results show that the hybrid filtering algorithm is effective.

The extended Kalman filter (EKF) has been used widely in Global Position System (GPS) and Strapdown Inertial Navigation Systems (SINS) integrated navigation systems,which simply linearizes all nonlinear models.However, careful treatment of the nonlinearity of the system models is particularly critical when the integrated systems use low cost micro-electro-mechanical (MEMS) inertial sensors. Hence application of the EKF in MEMS inertial sensor-based integrated navigation systems will likely lead to inaccurate results. To overcome the shortcomings of the EKF, the unscented Kalman filter (UKF) has been proposed.The UKF does not require the linearization of the system models. Alternatively it uses a set of deterministically selected "sigma-points",which completely capture the true mean and covariance of the original random vector. Then these sigma-points are propagated through the nonlinear models.This captures the mean and covariance to second order accuracy for arbitrary nonlinear functions. In this paper the EKF and UKF are applied in GPS/SINS integrated systems respectively. Simulation result demonstrates that the performance of the UKF is better than the EKF in MEMS inertial sensor-based SINS/GPS integrated systems.

Aiming at the integrated navigation system with model errors, the model errors is assumed as a process noise for Gaussian white noise to process in Kalman filtering, thus causing larger state estimation error of nonlinear filtering system and even divergent. This paper presents a nonlinear model predictive particle filtering method considers the model error of real-time estimation, and then corrects the nonlinear and non-Gaussian system model with model error. Compared and simulated the new method with the predictive filtering and the particle filtering in Strap-down Inertial Navigation System/Synthetic Aperture Radar integrated navigation system. Experiment and comparison results demonstrate that the proposed method can improve the filter performance and integrated navigation calculation accuracy significantly.

In this study, the integrated navigation system, consisting of radio and inertial navigation system (INS) altimeters, is presented. INS and the radio altimeter have different benefits and drawbacks. The integration is achieved by using an indirect Kalman filter. Hereby, the error models of the navigators are used by the Kalman filter to estimate vertical channel parameters of the navigation system. In the open loop system, INS is the main source of information, and radio altimeter provides discrete aiding data to support the estimations. At the next step of the study, in case of abrupt faults, the performance of the integrated system is examined. The optimal Kalman filter reacts with abnormal estimates to this situation as expected. To recover such a possible malfunctioning, the robust Kalman filter algorithm is suggested.

Embedded system has outstanding characteristics in real-time, portability, application specificity and resource possession of the system. The application of embedded technology in integrated navigation system can make the system smaller and more specific, therefore, this paper designs and implements BDS /GPS/SINS integrated navigation system based on embedded system. The BDS/GPS/SINS integrated navigation system can combine the advantages of BDS, GPS and SINS with higher precision and better stability. The system design covers the hardware and software, including the overall design of the hardware, the selection of the sensor module, the algorithmic analysis of the software and the system transplant. The precision and reliability of BDS/GPS/SINS integrated navigation system were tested and verified by dynamic experiment. The results show that the error range of BDS/GPS/SINS integrated navigation system in longitude, latitude and height was better than GPS/SINS navigation system.

Most autonomous vehicle navigation systems rely on Global Navigation Satellite System (GNSS) as a primary positioning sensor. However, a standalone GNSS receiver may not be able to meet the required positioning performance in aspects of position accuracy, robustness against signal blockages or signal reflections, and position output rates. Therefore, an integrated navigation system with an Inertial Measurement Unit (IMU) and GNSS receiver is often used to provide a continuous, reliable, high-bandwidth navigation solution. In general, Kalman filter is often used to integrate IMU measurements and GNSS output with predefined characteristic models under stochastic assumptions. If there is a discrepancy between the predefined model and actual sensor behavior or when proper correction on IMU measurements is not possible due to GNSS outages, the positioning performance of the filter will be degraded. For the application having a stringent positioning performance such as autonomous vehicle navigation, a more sophisticated sensor model or methods should be employed. A Recurrent Neural Network (RNN) – Long Short-Term Memory (LSTM) network has been found to have strengths in making a prediction with time-series data and has been applied for various navigation problems from the prior research. Therefore, in this paper, to reflect the non-linearity and time-correlated characteristics of the actual IMU sensor measurement, an RNN-LSTM network is employed for the GNSS-IMU integration. The paper investigates RNN-LSTM based GNSS-IMU integration methods through simulation and experiment and the results showed that the proposed RNN-LSTM network can provide relatively stable positioning performance even in a situation where a GNSS signal can not be received for 200 seconds.

Celestial navigation system (CNS) is a completely autonomous navigation technology with strong anti-interference ability and high navigation accuracy, which has a great application prospect. Aim at defect of traditional integrated navigation system using attitude angle measurement celestial navigation system and strapdown inertial navigation system (SINS), which can only correct of strapdown inertial navigation system’s gyro drift. This paper proposes an improved integrated navigation system that add double-star positioning celestial navigation to traditional SINS/CNS integrated navigation system, meanwhile, introduces the fault detection technology based on statistical distance to solve the problem that inaccuracy of measurement information leads to the integrated navigation system’s performance degradation. By analyzing the simulation results, the improved navigation system can inhibit divergence of system and obtain high accurate and stable navigation information.

It is a reasonable question. Why would we want to go through the time and effort to integrate two different navigation systems? The simple answer is that we integrate multiple systems when no one system by itself will satisfy all the necessary requirements. More generally, we integrate multiple systems in order to construct a super-system that in some way outperforms any of the individual systems.We are going to be spending a lot of time developing the framework behind the Kalman filter and ultimately driving toward the goal of developing an architecture that will enable us to integrate multiple navigation systems into a single integrated unit. We will start off by looking at navigation system integration and consider some key issues along with goals and requirements. We will investigate subsystem characteristics and will look at some challenges to integration.Navigation engineers must be cognizant of the fact that their product is a component within a larger system. This larger system is typically a vehicle (specifically, a land, air, sea, or space vehicle). As I have told many students throughout my years of teaching: an inertial navigation system will not transport me from New York to Chicago. It will tell me where I am. It will tell me how fast I am going. It will tell me the current attitude of my vehicle and how fast it is turning. However, it will not answer the most basic navigation question that all drivers, pilots, and captains must answer in order to get where they want to go: Which way to steer?The position, velocity, and attitude information determined by the navigation system, in virtually every case, is handed off to another system. Frequently it is either a flight management system or a flight control system or some kind of display for the pilot (or driver or captain...). It is also possible that it may be pointing a sensor or a weapon or providing the position and velocity needed to process synthetic aperture radar images. In any case, the navigation system needs to provide its information in a form and with a quality that satisfies the needs of the downstream user or system.

This paper proposes an in-coordinate interval adaptive Kalman filter for INS/GPS/SMNS (Inertial Navigation System/Global Position System/Scene Matching Navigation System) integrated navigation system in which SMNS has the characteristics of randomness and in-coordinate interval when it outputs the reliable positioning information. The proposed filter sets the period of the INS parameters analysis as reference, the periods of parameters output of GPS and SMNS are adjusted to be integral multiple of the INS analysis period by interpolation or extrapolation method; The system is updated only based on INS information when the observation information don't reach, it is updated based on all navigation information when the observation information reach; In addition, this method can overcome the drawbacks of the traditional Kalman filtering method when the system measurement noise is unknown. The proposed algorithm and the method proposed by reference [8] are applied to INS/GPS/SMNS integrated navigation system. Simulation results show that the in-coordinate interval adaptive Kalman filter outperforms the traditional Kalman filter.

For the long-range voyage and light Autonomous Underwater Vehicle (AUV), the high-precision navigation and positioning is the key point in the completion of the task. Therefore, this paper designs an underwater integrated navigation system, including the strapdown inertial navigation system (SINS), Doppler velocity loci (DVL), magnetometers and depth sensor. Based on the principle of each navigation sensor and the corresponding error model, the mathematic model of each sub-filter of the integrated navigation system is firstly established. Secondly, combining the combination filter and federated filtering theory, the integrated navigation system based on federated filtering is designed. Finally, to eliminate the influence of unknown or time-varying statistical characteristics on the Kalman filter (KF), a new federated filtering algorithm based on Sage-Husa adaptive KF is proposed to estimate each local filter. And the corresponding simulation results will show the performance of the research.

The centralized Kalman filter is always applied in the velocity and attitude matching of Transfer Alignment (TA). But the centralized Kalman has many disadvantages, such as large amount of calculation, poor real-time performance, and low reliability. In the paper, the federal Kalman filter (FKF) based on neural networks is used in the velocity and attitude matching of TA, the Kalman filter is adjusted by the neural networks in the two subfilters, the federal filter is used to fuse the information of the two subfilters, and the global suboptimal state estimation is obtained. The result of simulation shows that the federal Kalman filter based on neural networks is better in estimating the initial attitude misalignment angle of inertial navigation system (INS) when the system dynamic model and noise statistics characteristics of inertial navigation system are unclear, and the estimation error is smaller and the accuracy is higher.

Aiming at the characteristics of large volume, low degree of integration in traditional missile-borne distributed tightly-coupled navigation system, a compact integrated navigation system is designed. Low cost IMU(inertial measurement unit) and GNSS receiver are used as navigation sensors, and the algorithm is programmed on a navigation computer designed based on DSP C6747+FPGA shallow-parallel architecture, the size of the whole system is only 13cm∗14cm∗12cm without external antenna. In terms of algorithms, an adaptive Kalman filter algorithm is adopted for information fusion. Meanwhile the status of GNSS is judged by a new proposed algorithm before Kalman filter to avoid the affect of abnormal values of pseudo-range and pseudo-range rate. In addition, satellite-selection algorithm is designed and used step by step to avoid the influence of long time cost of satellite-selection algorithm. Finally, the results of vehicle experiments show that the designed system has very high navigation precision, the dynamic level positioning error is less than 2m(1σ) and velocity error is less than 0.2m/s(1σ).