IEEE Access (Jan 2023)
Empirical Approaches for Improving Inertial Pedestrian Navigation: Approximating Gravitational Acceleration and Digitizing Heading Change Angles
Abstract
Spatial mobility provides a rich context for human life, and the demand for indoor navigation and localization has increased. Pedestrian dead reckoning (PDR) is a viable solution thanks to its cost advantages and robustness to environmental changes. In this paper, we propose two generic ideas for improving inertial pedestrian navigation. One is about the estimation of gravitational acceleration and the other is about the estimation of heading change angles. Gravitational acceleration is dynamically extracted on the fly rather than assuming sensors to be fixed in a certain orientation. It can be used to estimate step lengths and heading change angles as a baseline. Also, heading change angles are digitized by combining the estimated gravitational acceleration with a simple threshold-based turn detection algorithm. Turns tend to occur across multiple steps and we separate turns from steps in walking. To demonstrate the effectiveness of the ideas a simple scheme, steps-and-a-turn (SnT), is designed for inertial pedestrian navigation. In experiments using a complete daily route, we show that the estimation of gravitational acceleration is consistent and robust, and that the digitization of heading change angles is highly effective in typical building environments: the positioning error is about 1.2% of the total length of the experimental path. Various state-of-the-art schemes served on top of pure inertial pedestrian navigation are expected to benefit by utilizing the proposed ideas as basic building blocks.
Keywords
