Multi-sensor data integration for personal navigation

Abstract

GPS technology has made outdoor navigation ubiquitous and has led to consumers salivating at the idea of importing this capability indoors. Unfortunately, GPS is not reliable indoors. This presentation will discuss how multiple sensors of different modalities can be combined to achieve precise personal navigation over long periods of time. Our system is built with inertial sensing at its core as navigation signals from such sensors are inherently secure against tampering compared to WiFi or cellular radio fingerprinting. To overcome the accuracy challenges of inertial navigation, we first expand on a well known technique called zero velocity updates during the stance phase of walking with a pseudo-measurement that detects small, non-zero velocities from statistical analysis of the stance phase. While this reduces the time dependence of computed position error from cubic to linear, it is not effective in reducing heading gyro drift. So, secondly, we introduce the measurements of range between shoes using both ultrasonic and radio frequency sensors to enable observability of the heading gyro biases. Integrating the data from the velocity sensing, range sensing and inertial sensing leads to positioning accuracy of less than 3 m after 2 hours of a combination of sitting and walking.