Paper Summary. The paper “No Need to War-Drive: Unsupervised Indoor Localization” introduces UnLoc which is an indoor localization scheme that does not require application of war-driving. Despite the great variety of researches, made in the sphere, the issue of indoor localization remains outside the mainstream. Current paper also presents the summary of several ideas in mobile computing and introduces an unconventional approach to indoor localization. In particular, every spot has unusual magnetic fluctuations. These fluctuations are considered to be internal landmarks of a building and can be detected by mobile devices. At the same time, the process of localization is performed on the basis of computing the location of users and landmarks.
Objectives of the Paper. The goal of the paper is to prove that particular location within indoor environment has a set of signatures and sensor measurements. As a result, indoor environments can provide a basis for detecting localization.
Approach, Novelty, and Technical Depth. Novelty of the paper is represented by the approach itself. Based on the previous researches in the field, the approach represents an unconventional view on indoor localization. In particular, approach foresees that mobile devises remain dead-reckoned while using their sensor measurements. At the same time, mobile device measurements are applied for detecting particular signatures within an environment. Environmental signatures can improve the error of dead-reckoning and provide higher accuracy of localization.
In addition, the approach is helpful in analyzing the opportunities for simultaneous appliance of two principles. One of them represents sensor-based dead-reckoning. Another is defined as environment sensing. Both principles are used for the purpose of localization. In order to extract particular sensor signatures also referred to as landmarks, the paper applies a practical scheme. As a result, the paper introduces new scheme called UnLoc. The scheme is used for detection of unsupervised localizations.
Approach also provides an opportunity to overcome the GPS error. In particular, dead reckoning requires the usage of expensive inertial sensor. At the same time, dead reckoning is characterizd by existence of an error. A number of errors can increase even with application of mounted accelerometers in a mobile device. In order to make an error insufficient, researchers apply GPS recalibrations within outdoor environments. However, GPS systems cannot be applied within indoor environments. As a result, introduction of UnLoc scheme can provide localization within the indoor environment.
The novelty of the paper is also represented by performance comparison of localization schemes. In particular, UnLoc is compared to EZ scheme. Both indoor localization schemes are aimed at achieving calibration-free process of localization. However, EZ scheme relies mainly on GPS techniques that cannot be efficient within the indoor environment. Thus, UnLoc has better accuracy and, hence, is more sufficient than other schemes. In addition, it does not require application of war-driving or availability of particular infrastructure.
Major Strengths and Weaknesses. Strengths of the paper is represented by the structure of the research which presents an overview of UnLoc. The paper provides development of the core framework of the research as well as the algorithms and engineering details. In addition, UnLoc is compared with other schemes used to identify indoor localization. According to the research, UnLoc is more effective than other schemes, as it remains less sensitive to RF signal fluctuations. In addition, UnLoc does not require any kind of special infrastructure or war-driving. Another strength of the research is represented by application of the experiments in three different environments that use several landmarks. The paper also analyses other schemes similar to UnLoc. It is also necessary to mention, that indoor localization scheme can be applied for detecting of activity based landmarks. In particular, some objects, such as cafes, may have queues. While other schemes fail to detect such kinds of landmarks, UnLoc can analyze them. Activity based landmarks can be characterized in accordance with their signatures. Thus, it is possible to detect objects that can change location. Finally, dead reckoning of the paper was made in accordance with two sub-tasks. The first task included evaluation of the users’ displacement from the accelerometer. The second task was to track the movement direction. Completing of both tasks alllowed evaluating possible error of the scheme. As a result, accuracy of UnLoc reaches 98 percent.
However, research has several weaknesses. In particular the scheme is not applied for detection of sensors that are characterized by high energy footprints. There are also drawbacks in implementation of the results of the research. Retracing and correction of the landmarks cannot be successfully applied in real time tracking of the users. It is also difficult to make large scale testing of UnLoc scheme.
Learn from the Paper. The approach represented in the research demonstrates that mobile devices that use measurements can detect environmental signatures within an area or a building. At the same time, there is no need in applying the war-driving.
Future Work. There are several issues that can improve future development of the approach. For instance, detection of the landmarks differs greatly depending on the smart phone platform used. The research was not held on different platforms. As a result, UnLoc has to be tested on various types of smart phones. In order to achieve this goal, the future research on cllecting and indexing smart phone data has to be conducted. Indeed, data collected can help to develop unique landmark signature for a particular phone model. Another area of prospective improvement on the issue includes analysis of the handling arbitrary orientation of the smart phones and discussion of its influence on localization. Calculation of the results was made upon the usage of smart phones in real-world settings, such as handling phones or holding them in pockets. Further research on the issue can explore application of localization scheme regarding activity based landmarks.
Other sphere of the issue that requires further research is represented by the necessity of the energy footprint analysis. The paper did not pay attention to energy consumption of the localization scheme. Finally, despite the fact that UnLoc scheme has been tested within three different environments, there is a necessity of testing the scheme over a large scale. In particular, the number of users and the amount of data has to be increased in order to test efficiency of UnLoc. As a result, further research on the issue can significantly improve localization performance.