SmartSLAM - An Efficient Smartphone Indoor Positioning System Exploiting Machine Learning and Opportunistic Sensing

Abstract

The most promising solution to the ubiquitous positioning problem is the smartphone, and many smartphone-based indoor tracking methods exist today. To ensure consumer acceptance of the technologies, it is critical that these systems do not have a significant effect on the battery life of the device. Methods exploiting signal fingerprinting have been shown to provide good performance with low processing overhead but require prior surveying. Methods exploiting opportunistic sensing and machine learning techniques such as Simultaneous Localization and Mapping (SLAM) need no prior data but at the cost of high computational load. This paper describes a smartphone-based indoor positioning system that exploits a new intelligent filtering approach to reduce this computational load. SmartSLAM moves between different sensor fusion algorithms depending on the current level of certainty in the system, reducing the computational load of the tracking engine, maintaining good positioning performance, improving battery life and freeing CPU cycles for foreground processes.