MEMS Sensor Assisted Terrestrial Vehicular Navigation on Portable Devices

S. Bharadwaj, S. Murali, J. Balakrishnan, A. Deshpande, Y. Shekar, G. Dutta

Abstract:	We present a GNSS sensor blending approach employing MEMS sensors specifically tailored to terrestrial vehicular navigation using portable units, such as smart phones and Portable Navigation Devices (PNDs). GNSS systems perform poorly in poor satellite visibility conditions, and sensor augmentation can help in improving the overall positioning accuracy and reducing outages under such scenarios [1, 2, 3]. We first address the task of providing sensor assistance to a GNSS solution using an Inertial Measurement Unit (IMU), consisting of a 3-axis accelerometer and a 3-axis gyroscope, which is docked to the vehicle in a fixed but arbitrary, unknown orientation. In literature, the unknown orientation between the IMU and the vehicle (relative attitude) is estimated in two steps. First, the horizontal alignment (tilt) is estimated when the vehicle is at rest [1, 2, 3], and then the angle between the sensor frame and the vehicular frame after de-tilting (yaw) is estimated using GNSS aiding whenever the vehicle is moving with constant heading [3]. In this paper, we propose techniques for estimating the relative attitude (both tilt and yaw) even when the vehicle is in motion, with or without GNSS aiding. Further, we augment this with an IMU attitude disturbance monitoring and re-estimation mechanism, which improves the solution’s robustness to real-world use-case scenarios, where the vehicle may not come to rest for long time durations, and the PND containing the IMU may potentially be undocked/docked by the user while the vehicle is in motion. Typical PNDs have low-cost (hence, low-grade) MEMS sensors, which need calibration before they can be used to provide assistance to the GNSS system. Also, with low-grade sensors, it becomes impractical to calibrate and track sensor non-idealities in all three dimensions, and some simplifying assumptions specific to terrestrial navigation are typically employed. Essentially, for terrestrial navigation, it is assumed that it is sufficient to track speed variations by monitoring acceleration along the drive-axis of the vehicle, and heading changes by monitoring angular velocity about a vertical axis [1, 3]. Most solutions employ a tightly coupled formulation of blending, where the raw sensor data are directly fed as measurements into a unified Extended Kalman Filter (EKF) in the GNSS Position Engine (PE), which in addition to tracking the position and velocity state variables, also tracks sensor non-idealities such as biases and drifts [3]. This complicates the tuning of the EKF because of a larger set of state variables, and renders the solution inflexible in terms of ease of accommodation of complementary information, such as that from the vehicle’s tachometer. We propose an alternate, cross-coupled, formulation of blending, wherein sensor calibration is performed external to the PE, in a dedicated Sensor Engine (SE) EKF, which uses an unfiltered version of the positioning information from the PE for calibration; and the calibrated sensor assistance is in turn provided to the PE as a control input. We explain how this arrangement allows for a more generic and simplified abstraction of the interface for assistance information going into the PE, without burdening the PE with the additional task of tracking sensor errors. The primary problem in using a “single-axis” tracking scheme for low-grade sensors is the issue of gravity leakage: road inclination changes resulting in a time-varying bias component on the drive-axis of the vehicle, making the accelerometer calibration unreliable in blockages and urban canyons. In literature, accelerometer bias drifts are typically modeled as a first-order Gauss-Markov process [1]; however, such a model, while suitable for tracking inherent drifts in MEMS sensor biases, does not work well for situations where gravity leakage dominates. Instead of attempting to track this unwieldy bias component, we apply a two-pronged strategy: For dead reckoning in blockage scenarios, we partially address the problem of positioning error growth due to gravity leakage, by proposing an error containment mechanism that exploits correlations between accelerometer and gyroscope measurements that exist during turns. For urban canyon blending, we propose a strategy to selectively weaken the accelerometer blending while leaving the gyroscope blending undiluted, in order to strongly steer the evolution of the direction (heading) of the GNSS position engine’s velocity state while not as strongly influencing its magnitude (speed). We present results from field trials performed on an embedded system based evaluation platform. Our results on dynamic attitude estimation show that after an attitude disturbance (undock/dock) event, the time taken for sensor blending re-engagement is typically under tens of seconds even when the vehicle continues to be in motion. In terms of improved quality of fixes, we show that even in harsh urban canyons such as downtown Chicago, large position outliers of up to 200m, which are possible with GNSS-only solutions under low satellite visibility and/or poor geometry conditions, are contained to less than 50m with sensor blending. We also show how the error containment mechanisms help restrict large over/under shoots due to gravity leakage, of the order of 140m, during dead-reckoning in multi-level parking lots, to well under 15m. We conclude by summarizing how the proposed algorithms addresses several practical issues related to using low-cost MEMS sensors for terrestrial navigation. References: [1] Davidson, P.; Hautamäki, J.; Collin, J., “Using low-cost MEMS 3D accelerometer and one gyro to assist GPS based car navigation system,” in Proc. Conference on Integrated Navigation Systems, May 2008, Saint Petersburg, Russia. [2] Davidson, P.; Vazquez, M.A.; Piche, R., "Uninterrupted portable car navigation system using GPS, map and inertial sensors data," in Proc. ISCE, pp. 836--840, May 2009. [3] Chowdhary, M., Colley, J., and Chansarkar, M., “Improving GPS location availability and reliability by using a suboptimal, low-cost MEMS sensor set,” in Proc. ION GNSS, pp. 610–614, Sep. 2007.
Published in:	Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013) September 16 - 20, 2013 Nashville Convention Center, Nashville, Tennessee Nashville, TN
Pages:	1084 - 1091
Cite this article:	Bharadwaj, S., Murali, S., Balakrishnan, J., Deshpande, A., Shekar, Y., Dutta, G., "MEMS Sensor Assisted Terrestrial Vehicular Navigation on Portable Devices," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 1084-1091.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In