Pedestrian navigation using wrist-worn inertial sensors

Resum

Global Navigation Satellite Systems are currently the main source of location information. However, due to the need for Line-Of-Sight with satellites, it is not possible to use them continuously. This represents a limitation to the development of seamless Location-Based Services for pedestrians as they spend most of their time in indoor environments. In this context, the combination of inertial Dead Reckoning systems, which do not need additional infrastructure to be installed in the environment, and wrist-worn wearable devices, such as smartwatches and smartbands, which allow different types of sensors to be carried in a continuous and user-friendly manner, can play a key role in the development of seamless pedestrian localization systems. However, despite the potential they seem to offer, the upper limbs are a challenging body location and there are few proposals for wrist-worn inertial Pedestrian Dead Reckoning (PDR) systems. Additionally, the existing literature is not enough to provide a complete overview of the operation and performance of this type of localization systems. Therefore, the main goal of this thesis is to contribute to both the understanding and the improvement of the performance of the wrist-worn inertial PDR systems. To this end, firstly, an in-depth analysis of the characteristics of the signals coming from wrist-worn sensors was carried out, including their variability with respect to factors such as different users, the arm pose or motion, or the walking speed. Then, a complete and systematic review of the step length estimation methods based on inertial sensors aimed to identify the reasons why, despite the many methods proposed, this topic is still in research stage. In relation to the step length estimation methods, Ultra Wideband (UWB) technology was identified as a possible source of ground truth for calibrating PDR systems, thanks to its balance between accuracy and coverage. Four different step length estimation algorithms using UWB were tested. Later, the performance of a wrist-worn inertial PDR system was evaluated by participating in the IPIN Indoor Localization Competition 2016. Another evaluation stage was also done including multiple users and the simultaneous comparison against other state-of-the-art inertial pedestrian navigation systems. Finally, the improved Heuristic Drift Elimination (iHDE) method, originally designed for foot-mounted inertial navigation systems, was adapted for systems based on the integration of the length and orientation of each step. This simple heading drift reduction heuristic method helps to improve the estimation of the position in most buildings. The results obtained throughout the thesis indicate that, although clear weaknesses have been identified that still need to be addressed, the performance of inertial PDR systems carried on the wrist is not too far from that of other systems mounted on other parts of the body. Therefore, the main conclusion of this thesis is that work must continue to improve PDR systems on this part of the body, since the advantage offered by the wrist at a level of comfort for the user may be key for achieving that continuous localization systems become real products and reach the mass market.