Ambient Assisted Living (AAL) technologies aim at improving the safety, health conditions and wellness of the elderly. The human movement patterns can be monitored in their own homes or workplaces using indoor location solutions that can help to determine changes in the people’s health conditions.
Indoor location is of great importance for a range of pervasive applications, attracting many research efforts in the past decades. This technology allows a precise location where Global Positioning System (GPS) signal is not available, leading to potential applications in AAL (e.g. providing assistance to elderly people in their own homes), workforce management (e.g. finding nearby doctors), social networking, context-dependent information sharing, among others. The
current indoor location solutions based on smartphones built-in inertial sensors, are becoming popular due to the growing availability of these devices and advances in their embedded hardware. However, current systems either require specific and expensive equipment that needs to be installed and maintained, or require the collection of updated building data beforehand (fingerprinting-based solutions), which involves intensive costs on manpower and time, limiting the number of buildings where indoor location can be implemented. Moreover, the positioning accuracy of fingerprinting-based solutions is highly dependent on the density of the signal database and even impossible in areas without calibration data.
The aim of this project is to develop an algorithm that automatically construct environment fingerprint maps from data collected by smartphones, in a way that training data can be crowdsourced without any explicit effort on the part of users. The proposed solution will not require any infrastructure, relying instead solely on crowdsourcing data. The use of crowdsourcing techniques avoid the time-consuming site survey process, in which signatures of an interested area are annotated with their real recorded locations.
For this purpose, the solution will use environment measurements from opportunistic data, such as magnetic field, Wi-Fi scans, light, sound, among others. Combining this contextual information with inertial sensing, it is desired to employ a suite of techniques to infer the location in the building map. The activity sequence contained in the trajectories can be extract by activity detection and pedestrian dead-reckoning. Therefore, crowdsourcing trajectories can be located and fingerprints can be constructed based on localization results.
In this work, it is expected the development of crowdsourcing techniques to facilitate the current mapping in indoor location fingerprinting processes.
Author: Ricardo Santos
Type: MSc thesis
Partner: Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa