Multi-Sensor PVT Solution for Android Devices

One of the greatest assets of smartphone devices is the availability of a diverse set of sensors embedded in the phones or tablets. These include but are not limited to Global Navigation Satellite Systems (GNSS) receivers, modems for network location positioning, inertial navigation sensors (INS), magnetometers, and barometers. Although the capabilities of the sensors inside the devices are constrained by size, weight, cost and power (SWaP-C), utilization of all the sensors can provide a complementary position solution that leverages the positives and mitigates the limitations of each individual component. A standard method of fusing these sensors is through a loosely-coupled (LC) GNSS/INS solution, where GNSS positions and velocities are used to update the states consisting of position, velocity, attitude, and INS biases. When no GNSS measurements are available, INS is used to time update the states. One critical problem with this approach is that the observability of heading is dependent on the magnitude of the velocity. When the user is stationary or traveling at low velocities, the accuracy of the heading is very limited. Therefore, in this study, we use an additional sensor, magnetometer, to assist in heading determination for low velocity cases, because the accuracy of the magnetometer heading is less dependent on the device velocity magnitude. We describe the methods that can be used to calibrate the magnetometer to obtain valid heading information, use an Attitude and Heading Reference System (AHRS) filter to obtain accurate heading estimates, and use an interactive multiple model (IMM) filter to fuse the magnetometer heading measurements with the LC GNSS/INS solution. The IMM solution will be compared to the GNSS/INS and GNSS/AHRS solutions to show how it can leverage the benefits of both solutions to achieve robust and accurate final state estimate. The proposed solutions will be assessed using the Google Smartphone Decimeter Challenge (GSDC) 2023 training data. Overall, this paper will assist in improving the understanding of each Android sensors’ capabilities and demonstrate a successful Android multi-sensor solution using publicly available data.