Robust Heterogeneous Sensor Fusion for Orientation & Motion Estimation
Accurate orientation estimation plays a critical role in different fields, including human activity and motion analysis, space and marine vehicle navigation, human-robot interaction, gaming, virtual reality, etc. As compared to sophisticated optical systems for orientation and motion tracking, the widely available inertial measurement units (IMU) are considered the most viable option due to their source-less nature, low-cost, high data rate, less power consumption, and miniaturized size. However, despite their decisive advantages, these low-cost sensors impose several substantial signal-processing challenges as well. The first and foremost challenge is the interpretation of sensor measurements as these measurements rarely coincide with actual motion states of interest. It is mainly due to the reason that the sensors are attached to a moving coordinate system, often requiring three-dimensional strap-down integration. Second, the data from these sensors is corrupted by time-variant sensor biases and measurement noise, yielding large integration drifts and errors during the estimation phase. Taken together, these challenges require combining measurements with complementary information in carefully designed fusion architectures and mathematical models.
In this work, we developed advanced sensor fusion and state estimation methods, which combine the complementary information from these sensors and minimize the effect of sensor disturbances, biases, and noise on the orientation and motion states of interest. This work is divided into two parts. The first part of the dissertation is focused on the compensation of motion-induced uncertainties while estimating orientation and external acceleration. The second part focused on compensating the impact of external magnetic fields on magnetometers.
In this respect, four different algorithms are proposed. The first algorithm tries to compensate for the impact of external acceleration using gyroscope while simultaneously estimating gyroscope bias. The second algorithm considers the problem of complementing accelerometer with magnetometer without the use of gyroscope and utilizes objects' motion constraints for better state estimates. This algorithm is designed for ground vehicles where there are some natural constraints on the motion of the body. These constraints enable us to efficiently compensate for sensor disturbances such as gyroscope bias and external accelerations. In our formulation, we demonstrate that these constraints can be embedded inside fusion architectures, which enable very efficient estimation of orientation information even during prolonged dynamics. In the third algorithm, we fused the gyroscope, accelerometer, and magnetometer in complementary manners such that they can withstand prolonged dynamic state and sensor biases. In this algorithm, we ease the motion constraints assumed in our previous algorithms and make it more robust and independent of vehicle type. Our fourth filter is designed to estimate the heading (yaw) when the magnetometer is exposed to prolonged and severe magnetic disturbances. The proposed fusion schemes have been tested using simulations and real-world experiments, and their performance in terms of accuracy, robustness, and computational time has been analyzed in comparison to several state-of-the-art techniques.
List of Publications:
1) M. A. Javed, M. Tahir, and K. Ali, ”Attitude in Motion: Constraints Aided Accurate Vehicle Orientation Tracking in Harsh Environment,” in IEEE Transactions on Industrial Informatics, 2022. Doi: 10.1109/TII.2022.3181798
2) M. A. Javed and M. Tahir, ”A Gyroless Attitude Estimation for Ground Vehicle Under Severe Dynamic Conditions,” in IEEE Transactions on Intelligent Transportation Systems, 2022. Doi: 10.1109/TITS.2021.3125712.
3) M. A. Javed, M. Tahir, and K. Ali, ”Cascaded Kalman Filtering-Based Attitude and Gyro Bias Estimation With Efficient Compensation of External Accelerations” in IEEE Access, 2020. Doi: 10.1109/ACCESS.2020.2980016
4) M. A. Javed and M. Tahir, ”Orientation and Acceleration Estimation of a Ground Vehicle on a Rough Surface in GPS Denied Environment” accepted at The 96th IEEE Vehicular Technology Conference, 2022.
5) M. A. Javed and M. Tahir, ”Finding the right direction in the magnetically perturbed environment,” manuscript under preparation for submission to IEEE Transactions on Intelligent Transportation Systems.
Final Defense Committee (FDC):
Dr. Muhammad Tahir (Supervisor)
Dr. Ijaz Haider Naqvi (Member)
Dr. Hassan Jaleel (Member)
Dr. Mian Muhammad Awais (Member)
Dr. Khurram Ali (External Examiner)
Meeting Link (Zoom): https://lums-edu-pk.zoom.us/j/97545939596
Meeting ID: 975 4593 9596