Inertial Odometry on Handheld Smartphones

Abstract

Building a complete inertial navigation system using the limited quality data provided by current smartphones has been regarded challenging, if not impossible. This paper shows that by careful crafting and accounting for the weak information in the sensor samples, smartphones are capable of pure inertial navigation. We present a probabilistic approach for orientation and use-case free inertial odometry, which is based on doubleintegrating rotated accelerations. The strength of the model is in learning additive and multiplicative IMU biases online. We are able to track the phone position, velocity, and pose in realtime and in a computationally lightweight fashion by solving the inference with an extended Kalman filter. The information fusion is completed with zero-velocity updates (if the phone remains stationary), altitude correction from barometric pressure readings (if available), and pseudo-updates constraining the momentary speed. We demonstrate our approach using an iPad and iPhone in several indoor dead-reckoning applications and in a measurement tool setup.

Deep Learning Based Speed Estimation for Constraining Strapdown Inertial Navigation on Smartphones

“…As inputs we use the three-axis gyroscope and accelerometer data from a smartphone. This data is passed to an inertial navigation system (labeled 'INS' and adopted following [3]) doing statistical inference on the current 3D position, velocity, and orientation. The blocks ' ş ', 'ö', and 'g' denote integration, rotation, and gravity, respectively.…”

“…In this paper we take a different approach and, instead of trying to solve the inertial navigation problem end-to-end using neural networks like [5], we aim at combining machine learning based speed estimation with a classical method [3], which is based on probabilistic sensor fusion using extended Kalman filtering (EKF). That is, we build upon recent work [3], which has shown that utilizing automatic zero-velocity updates (ZUPTs) and pseudo-measurements for limiting momentary speed can give accurate trajectory estimates in varying use cases. However, often in free handheld movement ZUPTs can not be established frequently enough to constrain motion sufficiently.…”

“…However, often in free handheld movement ZUPTs can not be established frequently enough to constrain motion sufficiently. Hence, in this work we aim at using machine learning to improve the accuracy of the momentary speed estimates which are used as pseudo-measurements in [3]. Thus, instead of assuming a globally constant momentary speed with high uncertainty like in [3], we predict more accurate data-driven momentary speed estimates that can be used as pseudo-measurements with smaller uncertainty (see Fig.…”

“…Recently, there has been a growing interest towards implementing inertial navigation systems (INS, [1,2]) on smart mobile devices [3,4,5]. Indeed, most current smart devices, like phones, tablets and watches are equipped with inertial measurement units (IMU) implemented as microelectromechanical systems (MEMS).…”

Gyroscope-Aided Motion Deblurring with Deep Networks

“…The rotation matrix R(t 2 ) is then formed as the direction cosine matrix corresponding to the quaternion q(t 2 ) (see, e.g., [26]). In theory, the translation could also be recovered using an accelerometer [31,7,10,21]. However, this requires knowledge of the initial velocity of the camera, or alternatively, known stationary points or reference points which can be used to aid zero-velocity updates or position updates in a Kalman filter [21].…”