Sameh Nassar

Wavelet de-noising for IMU alignment

Inertial navigation system (INS) is presently used in several applications related to aerospace systems and land vehicle navigation. An INS determines the position, velocity, and attitude of a moving platform by processing the accelerations and angular velocity measurements of an inertial measurement unit (IMU). Accurate estimation of the initial attitude angles of an IMU is essential to ensure precise determination of the position and attitude of the moving platform. These initial attitude angles are usually estimated using alignment techniques. Due to the relatively low signal-to-noise ratio of the sensor measurement (especially for the gyroscopes), the initial attitude angles may not be computed accurately enough. In addition, the estimated initial attitude angles may have relatively large uncertainties that may affect the accuracy of other navigation parameters. This article suggests processing the gyro and accelerometer measurements with multiple levels of wavelet decomposition to remove the high frequency noise components. The proposed wavelet de-noising method was applied on a navigational grade inertial measurement unit (LTN90-100). The results showed that accurate alignment procedure and fast convergence of the estimation algorithm, in addition to reducing the estimation covariance of the three attitude angles, could be obtained.

Two-Filter Smoothing for Accurate INS/GPS Land-Vehicle Navigation in Urban Centers

Currently, the concept of multisensor system integration is implemented in land-vehicle navigation (LVN) applications. The most common LVN multisensor configuration incorporates an integrated Inertial Navigation System/Global Positioning System (INS/GPS) system based on the Kalman filter (KF). For LVN, the demand is directed toward low-cost inertial sensors such as microelectromechanical systems (MEMS). Due to the combined problem of frequent GPS signal loss during navigation in urban centers and the rapid time-growing inertial navigation errors when the INS is operated in stand-alone mode, some methodologies should be applied to improve the LVN accuracy in these cases. One of these approaches is to apply smoothing algorithms such as the Rauch-Tung-Striebel smoother (RTSS), which uses only the output of the forward KF. In this paper, the development of the two-filter smoother (TFS) algorithm and its implementation in LVN applications is introduced. Two different LVN INS/GPS data sets that include tactical-grade and MEMS inertial measuring units are utilized to validate the TFS algorithm and to compare its performance with the RTSS.

Wavelet Analysis For Improving INS and INS/DGPS Navigation Accuracy

The integration of the Global Positioning System (DGPS) with an Inertial Navigation System (INS) has been implemented for several years. In an integrated INS/DGPS system, the DGPS provides positions while the INS provides attitudes. In case of DGPS outages (signal blockages), the INS is used for positioning until the DGPS signals are available again. One of the major issues that limit the INS accuracy, as a stand-alone navigation system, is the level of sensor noise. The problem with inertial data is that the required signal is buried into a large window of high frequency noise. If such noise component could be removed, the overall inertial navigation accuracy is expected to improve considerably. The INS sensor outputs contain actual vehicle motion and sensor noise. Therefore, the resulting position errors are proportional to the existing sensor noise and vehicle vibrations. In this paper, wavelet techniques are applied for de-noising the inertial measurements to minimize the undesirable effects of sensor noise and other disturbances. To test the efficiency of inertial data de-noising, two road vehicle INS/DGPS data sets are utilized. Compared to the obtained position errors using the original inertial measurements, the results showed that the positioning performance using de-noised data improves by 34%-63%.

IMPROVING POSITIONING ACCURACY DURING KINEMATIC DGPS OUTAGE PERIODS USING SINS/DGPS INTEGRATION AND SINS DATA DE-NOISING

Abstract In the standard integration of a Differential Global Positioning System (DGPS) and a Strapdown Inertial Navigation System (SINS), the DGPS provides position information while the SINS provides attitude information. In addition, the DGPS measurements are used to estimate the inertial sensors systematic errors and the SINS is used to detect and correct GPS cycle slips. In case of GPS signal blockages, the SINS is used instead for positioning as a stand-alone system until the GPS signals are available again. To obtain accurate positions during DGPS outages, near real-time (or post-mission) techniques should be applied, where these techniques are known as bridging algorithms. In such algorithms, new and improved positions of the outage periods are estimated. In this paper, two different bridging methods are used namely: backward smoothing and parametric modeling. An SINS/DGPS data collected with a van has been used in the analysis. The results show that both bridging algorithms reduce the SINS positional errors for DGPS outages of 75 to 100 seconds with an average of 1.35 m to an RMSE of 19 cm in case of backward smoothing and 10 cm in case of parametric modeling. To separate between the actual motion dynamics and other disturbing vibrations, a de-noising of the SINS raw data is required. Therefore, a de-noising of the van SINS data has been applied using a wavelet decomposition technique to eliminate or minimize the effect of sensor noise and other high frequency disturbances (such as engine vibrations). An analysis of the SINS sensor kinematic raw data in the frequency domain shows clearly that the majority of the van motion dynamics are contained in the low frequency portion of the spectrum (below 3.0 Hz). Consequently, several levels of wavelet decomposition can be performed without losing any motion information. The application of both bridging methods after the SINS data de-noising reduces the positional RMSE to 11 cm and 7.7 cm using backward smoothing and parametric modeling, respectively.

An Efficient Method for Evaluating the Performance of MEMS IMUs

Advances in MEMS technology combined with the miniaturization of electronics, have made it possible to produce chip-based inertial sensor for use in measuring angular velocity and acceleration. These chips are small, lightweight, consume very little power and are extremely reliable. They have therefore found a wide spectrum of applications in the automotive and other industrial applications. Currently, new MEMS inertial sensors or IMUs developed by various manufacturers continue to emerge on the market. However, such sensors should be evaluated in terms of navigation performance. Common testing in the lab can provide parameters such as sensor noise density and bias instability but cannot predict the corresponding performance of a full navigation system. IMU/GPS field testing is the only way to evaluate the performance of MEMS IMUs especially when GPS signals are temporarily blocked. However, testing every MEMS sensor (or IMU) in the field is not practical since it is a time- consuming and costly task. Therefore, the main objective of this paper is the development of an efficient method for evaluating the navigation performance of any MEMS IMU using lab testing only. The developed method is based on using MEMS sensors static data signals to estimate the MEMS sensor errors. Hence, by grafting these errors into the signals of a high quality IMU (gyro drift of 0.005 deg/h), collected in a previously conducted typical field test, a quasi field dataset of the MEMS is obtained since the high quality IMU signals can be considered as the true inertial sensor. Such emulated MEMS IMU field data can then be processed with the corresponding GPS data collected in the same test to evaluate the MEMS IMU navigation performance. To test the efficiency of the proposed method, several land-vehicle kinematic datasets with GPS, a high-quality IMU and different MEMS IMUs were used. Static data of the same MEMS IMUs was collected and then the proposed method was applied. The performance of the MEMS IMU actual and emulated datasets is compared during several GPS signal blockage periods. The results show that both solutions have a similar behavior with an average difference of only 20% in terms of accumulated position drifts. This illustrates the usefulness of the proposed technique in addition to the cost and time savings.

Stochastic Error Modeling of Smartphone Inertial Sensors for Navigation in Varying Dynamic Conditions

This paper aims at investigating and analyzing the behavior of Micro-Electromechanical Systems (MEMS) inertial sensors stochastic errors in both static and varying dynamic conditions using two MEMSbased Inertial Measurement Units (IMUs) of two different smartphones. The corresponding stochastic error processes were estimated using two different methods, the Allan Variance (AV) and the Generalized Method of Wavelets Moments (GMWM). The developed model parameters related to laboratory dynamic environment are compared to those obtained under static conditions. A contamination test was applied to all data sets to distinguish between clean and corrupted ones using a Confidence Interval (CI) investigation approach. A detailed analysis is presented to define the link between the error model parameters and the augmented dynamics of the tested smartphone platform. The paper proposes a new dynamically dependent integrated navigation algorithm which is capable of switching between different stochastic error parameters values according to the platform dynamics to eliminate dynamics-dependent effects. Finally, the performance of different stochastic models based on AV and GMWM were analyzed using simulated Inertial Navigation System (INS)/Global Positioning System (GPS) data with induced GPS signal outages through the new proposed dynamically dependent algorithm. The results showed that the obtained position accuracy is improved when using dynamic-dependent stochastic error models, through the adaptive integrated algorithm, instead of the commonly used static one, through the non-adaptive integrated one. The results also show that the stochastic error models from GMWM-based model structure offer better performance than those estimated from the AV-based model.