Performance comparison of wearable-based pedestrian navigation systems in large areas

Abstract

Wearable devices are a key driver for the development of pedestrian navigation systems. In this work, we consider inertial navigation systems (INSs). There is a diversity of such INSs. Normally, the comparison of INSs is restricted to indoor environments, or to outdoor small areas. However, it is of interest to study the behaviour of INSs in large areas. To that end, we present a ground truth system with cm accuracy to evaluate navigation systems. The ground truth system is distributed in an area of 14380m2 approximately. The ground truth system is used to evaluate three INSs based on three different body locations: the thigh, which is denoted as pocket, the wrist and the foot. Additionally, the data from a glasses-mounted inertial measurement unit (IMU) are also collected. The data, as well as the ground truth, have been made available for download. The results of evaluating 995 ground truth points indicate that the foot INS outperforms the pocket INS in, at most, 2cm/s. The pocket INS has, in contrast, a better standard deviation of the position error, and a robust step detection. The wrist INS is the most sensitive system to outliers. Therefore, its average position error is the highest. All in all, there is still room for improvement in the performance of all evaluated INSs.