Multi destination Indoor Navigation Critical Analysis

This paper is all about a system of indoor navigation on the basis of Wi-Fi fingerprint localization and multi-destination path planning.

In the start, it has been described that the demand of navigation systems has been increasing due to the broad use of mobile communication and mobile devices technology. In navigation systems, the crucial tasks are route planning and positioning. In terms of outdoor environments, GPS or Global Positioning System has been actually used widely as it can offer reliable positions. Its service, however, is limited greatly in environments which are indoor. In recent years, IPS or Indoor Positioning System has taken a challenging form. On several technologies like Wi-Fi, Zig bee, Bluetooth, and RFID, the present indoor positioning systems are based. Among the technologies which are mentioned, Wi-Fi IPS is considered one of the greatest studies because of the presence of several access points of Wi-Fi, specifically in environments of indoor and the ease of a receiving module of Wi-Fi which is developed in all devices of mobile (Kannan, Meneguzzi, Dias, & Sycara, 2013).

Piggybacking information is used in the system in addition to the present system with almost no additional modification to the environment of indoor. The Wi-Fi fingerprint that includes the signal features or characteristics is usually used, specifically on the basis of fingerprint on RSS or Received Signal Strength. Although the strength of signal can be impacted by shadowing and multi-path influenced by diffraction, scattering, and reflection, for each position, the fingerprint is different it can be no doubt a good source for positions. Performed phases are two: localization and training. In the phase of training, a map of fingerprint is designed by collecting the vector of RSS achieved from the APs which are nearby at each and every point of reference in the selected sector. In the phase of positioning, it is more or less a task for identifying the position which is targeted. The issue can be considered a classification of multi-class. For identifying the needed positioning, many methods can be used like the KNN or k-nearest neighbor, SVM or Support mixture model, Naïve Bayes Classifier, statistical Gaussian mixture model, and WKNN or weighted KNN (Fallah, Apostolopoulos, Bekris, & Folmer, 2013).

The theme of KNN is to select fingerprints with k shortest distance in the database in comparison with the measured fingerprint. Moving on, the position result is the k-nearest location’s average coordinate. But if the APs which are k-nearest are scattered, large errors will be caused by it in the result of positioning. By using the calculated weights from the RSSI distance inverse can minimize the error. But limited reference points with WKNN cannot always obtain the accuracy which is required for the reason similar as the easy and simple KNN. For estimating the positioning, SVM was presented by finding the most suitable hyperplane that seems to separate fingerprints of one from the other class. The method of bilinear median interpolation was actually applied for reducing the calibration on the designing of fingerprint map. On the basis of fingerprint classification, Naïve Bayes has the nature of probabilistic which has actually been proven for being effective to handle RSS’s time variation. The new data point is associated to the class with probability that is highest.

Moving further, it has been explained that in the paper, the presented indoor system of navigation is described by the authors which actually permits navigation of multi-destination with the path using A* algorithms and 2-opt that is automatically updated with the fingerprint positioning of Wi-Fi. Localization and path planning have been integrated for providing find navigation solution with precise prediction. The user can detour any time and is permitted to specify various destinations. A position coordinate is provided by the Wi-Fi IPS which determines the position of user. It can be said that path finding is presented for organizing almost all destinations for forming the path route which is the shortest. Whenever the position of user is out of the route which is predefined, it will be activated again for navigation refinement and route updating.

In the second section of the paper, proposed algorithms have been described. 3 main and important components are included in the system architecture: Route Updating, Multi-destination Path Planning, and IPS or Indoor Positioning Service. At the beginning, an outline is offered by the user including destinations in a good number to the server. Moving on, Multidestination path planning is performed by the server that finds the route consisting of destination as per the requirements of the user and offers the outcome to the portable device of the user. In the process of tour, the Wi-Fi signal is continuously sensed by the portable device while retrieving the position of the user from the IPS. Whenever it is detected by the system that the user is detouring from the predetermined route, the report is automatically sent to the server by the portable device for a brand new route.

Moving on in the third section of the paper, the indoor positioning service using the localization of Wi-Fi fingerprint is explained. The service of indoor positioning is more or less a module for estimating the current position of user on the basis of RSS fingerprint of Wi-Fi where the connection between the given location’s radio signature and its corresponding is also analyzed. In every fingerprint, each AP’s RSS information is contained and fingerprints from the given location has a combination that is unique in comparison with map fingerprints. Again, the issue can be considered as a classification of multi-class. The presented methods of classification can be classified into two classes or groups: Probabilistic and deterministic algorithms. In the algorithm which is deterministic, the measured point’s RSS fingerprint in the database with the minimal signal of statistical distance between the detected vector of location and itself is usually regarded as the location estimation that is best. The methods of nearest neighbor are recognized as examples of the algorithms which are deterministic on the basis of Manhattan distances’ statistical distance computation. In the algorithm that is probabilistic, probability inferences are utilized for determining the possibility of the given location’s occurrence on the basis of conditional probability distribution in the database of reference.

Further, it has been explained that the classifier of Naïve Bayes is selected as the classification of probabilistic based fingerprint. More or less it is the simplest form of network classifier of Bayes that is a learning algorithm which is supervised on the basis of stochastic model with the independent features assumption. Regardless of its simple nature, it is less responsive or sensitive to RSS time variation, no estimation of iterative parameter with easy building which is quite useful for the dataset which is large and frequently exceeds in terms of performance compared to sophisticated classifiers.

In the end, it has been concluded that the system of indoor navigation has been presented on the basis of fingerprint positioning of Wi-Fi RSS and a mix of A* and 2-opt algorithms for path finding of multi-destination. Localization and path planning are integrated by it for flexibly navigating users to their several destinations even when a user takes another route. It has been determined that the demand of users on the indoor navigation service can be met by the presented service while improving the experiences of indoor, therefore decreasing time of search for the position that is required.

References of Multi destination Indoor Navigation Critical Analysis

Fallah, N., Apostolopoulos, I., Bekris, K., & Folmer, E. (2013). Indoor human navigation systems: A survey. Interacting with Computers, 25(1), 21-33.

Kannan, B., Meneguzzi, F., Dias, M. B., & Sycara, K. (2013). Predictive indoor navigation using commercial smart-phones. Proceedings of the 28th Annual ACM Symposium on Applied Computing, 519-525.