Design and Implementation of an Inertial Navigation System for Pedestrians Based on a Low-Cost MEMS IMU

TL;DR

This paper presents a low-cost inertial navigation system for pedestrians that mitigates drift using an extended Kalman filter and innovative foot stance detection, achieving comparable accuracy to more expensive systems.

Contribution

The paper introduces a novel inertial navigation approach that uses only accelerometer and gyroscope data, incorporating sensor error modeling and a new foot stance detection heuristic.

Findings

Achieves sub-meter accuracy in pedestrian navigation

Performs better or comparable to existing systems with cheaper IMUs

Effectively mitigates drift without magnetometers

Abstract

Inertial navigation systems for pedestrians are infrastructure-less and can achieve sub-meter accuracy in the short/medium period. However, when low-cost inertial measurement units (IMU) are employed for their implementation, they suffer from a slowly growing drift between the true pedestrian position and the corresponding estimated position. In this paper we illustrate a novel solution to mitigate such a drift by: a) using only accelerometer and gyroscope measurements (no magnetometers required); b) including the sensor error model parameters in the state vector of an extended Kalman filter; c) adopting a novel soft heuristic for foot stance detection and for zero-velocity updates. Experimental results evidence that our inertial-only navigation system can achieve similar or better performance with respect to pedestrian dead-reckoning systems presented in related studies, although the adopted IMU is less accurate than more expensive counterparts.