Inertial Odometry on Handheld Smartphones (1703.00154v2)

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 double-integrating 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 real-time 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.

• The paper presents a probabilistic approach for real-time inertial odometry that utilizes only standard smartphone sensors for accurate positioning and motion tracking.
• The methodology employs an extended Kalman filter to learn IMU biases online, enabling robust performance in challenging scenarios like elevators and escalators.
• This research demonstrates the potential of consumer smartphones for advanced navigation, enabling applications in augmented reality, indoor localization, and mobile gaming.

Overview of "Inertial Odometry on Handheld Smartphones"

This paper presents a comprehensive approach to inertial navigation using consumer-grade smartphones. It tackles the challenges faced by traditional inertial navigation systems (INS) that are not typically designed for the low-quality data produced by smartphone sensors, specifically focusing on the integration of accelerometer and gyroscope data. The methodology is built on a probabilistic approach capable of achieving real-time inertial odometry without relying on external sensors or additional hardware, which distinguishes it from many existing systems.

Key Contributions

The authors propose a model capable of learning inertial measurement unit (IMU) biases online through an extended Kalman filter (EKF). This model is robust, computationally lightweight, and compatible with smartphones, allowing it to accurately track positioning and motion in a variety of scenarios, including dead-reckoning and step-free applications such as elevators and escalators. This versatility addresses significant limitations of traditional step-and-heading INS approaches that are often restricted by known or constant device orientation and pedestrian motion.

Technical Approach

The system leverages a set of state variables, including position, velocity, orientation, accelerometer biases, gyroscope biases, and scale errors. The core component of the method is the dynamic model described by the mechanization equations, which employ double integration of rotated accelerations to estimate the system's state over time. The model fuses inertial sensor data using the EKF, generating sequential estimates of the smartphone's position and velocity while iteratively calibrating biases. To mitigate accumulative errors associated with the integration of sensor data, the system incorporates zero-velocity updates, altitude corrections, and pseudo-updates.

Results and Demonstration

The proposed system demonstrated its effectiveness using an iPhone 6 in various indoor localization tasks—a scenario where GNSS cannot function accurately. Empirical results showed the system's ability to handle dynamic and abrupt motions typical of handheld devices while maintaining accurate positioning. This is evidenced by the successful execution of tasks spanning different environments like open-air spaces and controlled indoor settings without the requirement of any pre-calibrations.

Implications and Future Directions

The insights obtained from this paper hold practical importance in fields such as augmented reality, indoor navigation, and mobile gaming, where precise localization is critical. From a theoretical perspective, this work underscores the potential of utilizing existing consumer devices for advanced navigation solutions without hardware modifications. Looking forward, continued exploration into machine learning techniques and further refining the probabilistic inference models are likely to enhance robustness and applicability across diverse operational contexts. The adaptation to various perturbations and environments will be central to broadening the application scope of inertial navigation solutions using smartphones.

Overall, the paper provides a foundation for future work to harness inertial data from smartphones, paving the way for innovations in mobile-based navigation technologies.