1. Information about the paper

Citation:

Wang, He, Souvik Sen, Ahmed Elgohary, Moustafa Farid, Moustafa Youssef, and Romit Roy Choudhury. “No need to war-drive: unsupervised indoor localization.” In Proceedings of the 10th international conference on Mobile systems, applications, and services, pp. 197-210. ACM, 2012.

Abstract:

We propose UnLoc, an unsupervised indoor localization scheme that bypasses the need for war-driving. Our key observation is that certain locations in an indoor environment present identifiable signatures on one or more sensing dimensions. An elevator, for instance, imposes a distinct pattern on a smartphone’s accelerometer; a corridor-corner may overhear a unique set of WiFi access points; a specific spot may experience an unusual magnetic fluctuation. We hypothesize that these kind of signatures naturally exist in the environment, and can be envisioned as internal landmarks of a building. Mobile devices that “sense” these landmarks can recalibrate their locations, while dead-reckoning schemes can track them between landmarks. Results from 3 different indoor settings, including a shopping mall, demonstrate median location errors of 1.69m. War-driving is not necessary, neither are floorplans – the system simultaneously computes the locations of users and landmarks, in a manner that they converge reasonably quickly. We believe this is an unconventional approach to indoor localization, holding promise for real-world deployment.

paper-review-6

2. My review of the paper

Summary:

The motivation behind this paper is that we need higher accuracy for indoor localization. This paper offers a novel idea to combine dead reckoning, urban activity sensing, and WiFi based partitioning. The paper uses combined raw data gathered from accelerometer, magnetometer, compass, gyroscope, and WiFi access points. The main idea of this paper is that each part of a building has its own distinct signature that can be used for indoor localization. Moreover, each part of a building might forces a user to do “certain” behavior while interacting with that part of the building. The paper identifies this distinct pattern as landmarks, as long as the pattern happens in a small area (the paper uses 4 m2) which is inferred from WiFi localization. Then the paper offers a framework to learn this pattern recursively over time. In addition, the paper also uses GPS signal to mark an entrance to a building to make sure the accuracy of starting learning point. The experiment was performed in three different buildings on top of Android smartphone.

Contributions:

This paper offers new method to achieve 1,69 m accuracy with zero calibration, thus as the title says, no need to war-drive. This method is also better because it does not need additional equipment to perform its operation. Previous indoor localization works like Horus, perform better accuracy (~1 m), but with the need for war-driving as a trade-off. While other non-war-driving approaches like EZ and SLAM performed with lower accuracy, 2-7 m and ~5m.

Moreover, it has been proven to achieve a relatively desired accuracy. It also performs better over time, when more users supply more data. However, early users will experience less accurate localization result.

Comments

Pros:

  1. This paper introduces novel idea for indoor localization.
  2. This paper is cited by 130 following works (according to Google Scholar) in just two years, which is a good sign of how useful is the paper to inspire following works. I have not seen a paper cited this much in two years.
  3. It is also featured in various technology related websites such as: Gizmag, TG Daily, Scientific American, News Observer, The Verge, PhysOrg, Connected World, Fast Company, I Programmer, Duke Today.

 Cons:

  1. Some limitations are already stated in the paper:
  2. Result may varies between different smartphone hardware and operating system, the paper has not tested this plausible real life implementation.
  3. Phone orientation effect. The experiment asks users to do a certain orientation, what if users face other direction in real practice?
  4. This paper avoids using sensors that will need higher energy consumption such as light and sound, which can offer higher localization accuracy.
  5. The paper did not test a larger scale application.
  6. The paper did not say about how long do the user should wait to get the localization result, which should be important.
  7. The paper also did not mention how much the resources, like the battery used by this approach.
  8. The paper uses two phones on each user, one in the pocket, and one in hand. What if we only use one phone in real practice?
  9. It uses floorplan as a guide. Will it be really realiable when relaxing this assumed floorplan and uses one GPS signal as a guide?

 Other comments:

  1. It could be developed further to achieve real time result.