Machine learning techniques for indoor localization on edge devices

As part of the Internet of Things, location-based services play an important role to support context-aware applications, providing services that react and interact with the users, especially in indoor scenarios. However, indoor localization imposes several technical restrictions that cannot be solved following a common GPS-based approach. This chapter offers an overview of the technical challenges that need to be addressed in the development of indoor positioning systems and presents different techniques and mechanisms that have been proposed to deal with these issues. However, many of the proposed machine learning techniques are supported by cloud deployment, which is not feasible in many real applications (low latency, low bandwidth). As a solution, TinyML has emerged to support machine learning-based indoor localization techniques, for which two implementations are presented and compared.