A PARTICLE FILTER BASED SCHEME FOR INDOOR TRACKING ON AN ANDROID SMARTPHONE

Abstract

The development of stable outdoor positioning and tracking algorithms has ushered in a new era of technology. Today GPS based navigators are ubiquitous and GPS based positioning and navigation has been integrated into smartphones. This has made life easy for navigation over large distances in an outdoor scenario. A new horizon for research has developed which aims at providing that same ease of use in navigation and tracking in an indoor context. Over the last few years, advances in technology have allowed for the development and integration of MEMS sensors into smartphones. These sensors (an accelerometer, a gyroscope and a magnetometer) can be leveraged to develop an indoor tracking solution. However, the resource constrained nature of smartphones necessitates modifications to prior work achieving the same in a non-mobile context. Additionally, the fact that the tracking device lies in a user's palm and not on a protected, stable surface (as is the case in robotics) generates a new set of challenges. This thesis proposes and implements an online particle filter based algorithm for indoor positioning on an Android smartphone for pedestrians. The proposed algorithm is not only able to work well on the resource constrained device, it actually yields an accuracy better than prior work in this area. Such results have been made possible by a careful approach to the problem with decisions made on factual data determined before and during the implementation of the proposed work. The final system is able to achieve a mean error of 0.5m with a very low standard deviation value of 0.24m and provides a clean, intuitive tracked path. Results for comparison are produced using a simple dead reckoning approach and a Nearest Neighbour Wifi corrected tracking algorithm. Both the algorithms are tested in the same environment as the proposed work