Recent technologies involving indoor positioning rely on mobile data acquired via mobile sensors that accumulate errors, and they also exhibit various flaws because sensors are prone to fluctuations caused by the environment. Hence the performance of indoor positioning technology depends on how well it can be corrected and ease of removal of these errors to accurately provide the user’s location. Numerous approaches have attempted to solve this problem using secondary sensors or landmarks.
BatTracker uses sensors to estimate the user’s position based on inertial data and corrects errors by measuring distance using acoustic signals reflected off nearby objects [23
]. If there is a three-dimensional space consisting of a ceiling and two walls, it is possible to correct the position of the user based on the reflected sound waves in the space. However, BatTracker requires a three-dimensional medium, in which sound waves can be reflected. If the sound waves generated by the device are irregularly reflected off the medium, it is difficult for the sound waves to be detected. ApfiLoc [24
] is another successful Calibration-free smartphone based indoor localization system that uses least-squares support-vector machine (LS-SVM) algorithm [25
]. It fuses the augmented particle filter (APF) [27
] and measurements from the inertial sensors of smartphones to provide localization with an error less than 2 m without relying on Wi-Fi support. However, it exhibits limitation in terms of obtaining precise heading estimations. SmartLight measures the position of a user by sensing a digital signal generated by a blinking light-emitting diode (LED) from an auxiliary light source sensor [29
]. Not only can SmartLight use the light sensor of the device instead of the underlying auxiliary sensor, it can also reduce the cost of installing the infrastructure by changing the lighting in the room instead of installing the infrastructure to generate the light source separately. However, due to the nature of the technology for tracking location using the light source, it is impossible to measure the location if the device is not receiving light. This is a fatal drawback in the case of an emergency when there is a limited light source. visual simultaneous localization and mapping (vSLAM) is a technology used to create a visual landmark map using an object recognition algorithm based on a camera image from a mobile device [30
]. Unlike other technologies, vSLAM has an advantage in that it does not require the installation of additional infrastructure or auxiliary devices. However, it relies on visual landmarks and due to the nature of the video as a landmark, the landmarks database is very large, especially in a large environment such as a shopping mall or a stadium. This makes it difficult to use in a big space or building where indoor elements change rapidly in real-time. Geomagnetism and crowdsensing-powered indoor navigation (GROPING) is a system that involves the use of a map for localization and navigation [31
]. It is based on magnetic fingerprinting as well as Wi-Fi fingerprinting using the Dijkstra algorithm [32
] and Monte Carlo localization [33
]. The component that builds the map incorporates all semantic information as well as all discovered landmarks that make the map easy to use. Its main advantage is the non-dependence on any infrastructure support. WiSLAM is a technology that uses two SLAM technologies, PlaceSLAM and FootSLAM, to track the user’s location [34
]. WiSLAM uses landmarks effectively by utilizing each SLAM according to the type of landmark. However, since the two SLAM technologies are used simultaneously, the computation time increases, which makes it difficult to apply the calculation process to mobile devices. SemanticSLAM has the robustness for use in any building other than a specific building using the landmark retrieved from the user’s sensor and the landmark simultaneously generated from the indoor RSS value [37
]. In addition, SematicSLAM can reduce the time required to create a landmark map by generating one of the two landmarks in advance if the interior information is known. However, there is a problem that the SemanticSLAM does not perform the update of the landmark separately, so even if the landmark is no longer being used, it is not removed. The robust crowdsourcing-based indoor localization system (RCILS) is an indoor location tracking system that calibrates a user’s position using his movement data according to the indoor structure and RSS data of a corresponding location as landmarks [38
]. RCILS predicts indoor movement using sensor data and the user’s indoor map and combines these with observation result from Wi-Fi to generate landmarks. However, there exists a problem in that it is difficult to apply RCILS to all buildings because a map is required to acquire room information in advance. Phillips has proposed the visible light communication (VLC) [39
] based indoor positioning system which uses light from LED luminaires and does not require additional installations. The system locates a user by sending a unique code to a mobile device. The advantages of the solution are that it is empowered by iPhone operating system (iOS), Android software development kit (SDK) as well as cloud services, which allow third-party developers to incorporate the positioning capabilities into their customizable application. Yuanchao Shu et al. developed G-Loc [41
], a gradient-based fingerprint map system that uses a map (Gmap [42
]) and then leverages RSSI values to establish the user location using the extended particle filter algorithm. Although it can reduce the cost associated with fingerprint map calibration, its performance can be hampered by unexpected fluctuations of RSSI values. Chen, Zhenghua et al. propose a fusion approach of Wi-Fi, smartphone sensors and landmarks using the Kalman filter for indoor localization [43
]. They use the weighted path loss (WPL) algorithm to track the location of Wi-Fi and conducts pedestrian dead reckoning (PDR) using smartphone sensors. They use also the knife-only filter to fuse the two techniques. Without the direct Wi-Fi installation in the building, it is difficult to pinpoint the location and the number of Wi-Fi installed in the building. Zou, Han et al. propose an accurate indoor positioning algorithm using mobile phone inertial sensors, Wi-Fi and iBeacon [44
]. They use inertial measurement unit sensors for PDR and use Wi-Fi fingerprinting and iBeacon by sensor fusion based on a particle filter for corrections. Not only does this technology require both infrastructure to be installed indoors, but also the Wi-Fi’s desired use requires upgrading the router’s firmware.
The aforementioned existing indoor positioning technologies attempt to correct errors of sensor data using auxiliary sensors, combining several algorithms, or diversifying landmarks. However, the problem with using auxiliary sensors is that these sensors can only be used in limited environments. For example, in the case of BatTracker which uses sound waves, it cannot be used in large halls or places with many objects. When using composite algorithms, it is impossible to correct sensor errors quickly due to decreased calculation speed in Smartphones, which have lower processing power than normal central processing units (CPUs). In the case of techniques that diversify landmarks, they only consider the addition of landmarks and do not consider the deletion of unnecessary landmarks. Our proposed indoor positioning technique, IPSCL attempts to address these problems. IPSCL utilizes only the built-in basic sensors of smartphones and accurately finds the position of a user using the gyroscope sensor and the magnetometer sensor alternatively, based on characteristics of their values. IPSCL also enables rapid location correction using the Bluetooth beacon-based landmark approach. Finally, IPSCL allows users to add crowdsourced landmarks which in turn can be used. It also updates the number of detection of crowdsourced landmarks and spreads most recent and most frequently used crowdsourced landmarks to users.