Due to the advance of the internet of things and business opportunities, indoor navigation systems have been deployed in many large buildings, such as big train stations, shopping malls, hospitals, and government buildings. After installing a navigation mobile app, users can select a point of interest on a menu list. Then, the app will determine a route to the destination, which is usually the shortest path. Nowadays, the most commonly used user interface (UI) of navigation applications is a 2D map with a route. Users are provided with navigation instructions, such as turn left, turn right, and go straight, when they are close to an intersection. However, due to the limitations of a 2D navigation map, it could add an additional cognitive load for users to construct the relationship between the 2D navigation map and the real environment. Extra mental pressure may also be induced and make users confused [1
]. Therefore, eliminating possible user confusion is important for navigator UI design.
In order to create a good user experience, several research efforts have been devoted to developing an indoor navigation system by utilizing augmented reality (AR) technology. A. Mulloni et al. [2
] and L. C. Huey et al. [4
] deployed markers as location anchors in the environment. A user can know their location by matching the markers with the associated location information stored at a remote server or on a user’s phone. However, the angle of the camera must be in proper alignment with markers before the matching process can start. In addition, the markers could get dirty easily and become unrecognizable, therefore increasing maintenance costs. S. Kasprzak et al. [5
], J. Kim et al. [6
] first performed an image search for pre-tagged objects, such as billboards and trademarks, in the environment, and then determined the user’s location based on the obtained objects. However, the more complicated the environment is, the more difficult it will be to identify pre-tagged objects. The image matching processing becomes even challenging when the layout and decoration of different parts of the space are similar. Feature matching is another method to determine user’s location [1
]. However, constructing point clouds of a real indoor environment is time consuming and costly, especially for a large building.
In this paper, we designed ARBIN, an augmented reality-based navigation system, by extending our previous work, WPIN [7
]. WPIN utilized Bluetooth low energy (BLE) beacons, named Lbeacons version 1 (BiDaE Technology, Taipei, Taiwan), deployed at each intersection and point of interest (POIs), to get the coordinates of the current position. 2D images, such as turn left, turn right, and go straight, were provided to users as direction indicators along the route to the destination. Unlike WPIN, ARBIN uses AR technology that combines virtual objects and the real world. Navigation instructions, as well as AR 3D models, are posted on the screen on the surrounding environment through the smartphone camera. Therefore, there is no need for users to make a connection between the navigation instructions and the real-world environment. In our implementation, Google ARCore (Google Inc., Mountain View, California, United States) [8
] is adopted to create AR 3D models, obtain gyroscope sensor readings, and determine where to put the models. Accuracy is the key factor for the success of an AR-based indoor navigator. The difficulties in achieving accuracy of indoor positioning, and that of AR 3D model placement are described as follows.
In WPIN [7
], Lbeacons were adopted at waypoints to periodically broadcast their own coordinates to smartphones nearby. A waypoint can be an intersection, a point of interest (POI), or the middle of a corridor. After receiving a broadcast message sent from a Lbeacon, the positioning module, running on the user’s smartphone, starts to estimate the distance between itself and the Lbeacon according to a RSSI (received signal strength indicator) distance model. The stronger the received signal is, the closer the user is to the Lbeacon. When the user and the Lbeacon are close enough, for example less than 5 m, a new direction indicator will pop up to guide the user to the next waypoint. The above process continues until the user arrives the destination. However, because of the machine cutting error, the size of the antenna board of each Lbeacon may not be identical, which could affect its capability for transmitting and receiving signals. Furthermore, the characteristics of the RF (Radiofrequency) circuit of each Lbeacon may also be different due to the nature of an analogy circuit. Therefore, the RSSI distance model of each Lbeacon is not exactly identical according to our experience. In our previous work, to achieve the required positioning accuracy, we constructed a RSSI model for each Lbeacon, which was time consuming and unscalable. To overcome this unavoidable and challenging hardware problem, a novel RSSI modeling method was developed to overcome the problem of the heterogeneous issues of Lbeacons, which is given in Section 3.2
The AR 3D models, such as a left arrow or a right arrow, should be placed properly in a real-world environment to avoid possible user confusion. An inaccurate placement of the 3D model may make users confused, and avoid using it. For example, displaying a 3D model in the wrong orientation, an incorrect elevation angle, or an incorrect depression angle. Many parameters should be carefully considered before having the correct placement of a 3D model, such as the face orientation of a user, the location, and orientation of the smartphone. Constructing a relationship between these parameters and the coordinates of a 3D model is challenging. The detailed method is presented in Section 3.4
to Section 3.5
In order to investigate the applicability of ARBIN, we first evaluated the responsiveness of the positioning module of ARBIN. We then set up a field trial in a hospital. For the former, we conducted a series of experiments in the engineering building No. 5 of the National Yunlin University of Science and Technology, crossing three floors with a total area of around 250 m2. The experiment results showed that the adopted RSSI (received signal strength indicator) model could accurately determine the distance between a Lbeacon and a smartphone. Thus, the AR models could be displayed correctly on the smartphone screen. Furthermore, a field trial was conducted at the outpatient area of the National Taiwan University Hospital YunLin Branch, which is nearly 1800 m2, with 35 destinations and point of interests, such as a cardiovascular clinic, x-ray examination room, pharmacy, and so on. Four different types of smartphone were adopted for evaluation. Our results show that ARBIN can achieve 3 to 5 m accuracy and give users correct instructions on their ways to the destinations. ARBIN proved to be a practical solution for indoor navigation, especially for large buildings.
In this paper, we presented ARBIN, an augmented reality-based indoor navigation system, to guide users to their destinations in an indoor environment. When users enter the range of a waypoint, ARBIN posts a 3D directional indicator into the real-world surrounding environment. With the support of augmented reality, it is easier for users to determine their locations when walking inside a building. To address the heterogeneous problems of Lbeacons, four types of RSSI model are proposed. Experiences in correctly placing a 3D model in a real-world were also explained. Further, we conducted both in-house experiments and a field trial to verify the responsiveness and practicality of ARBIN. The in-house experiments showed that in 92.5% of the test cases, ARBIN could provide users with a proper directional indicator when they came close to a Lbeacon. For the field trial, four volunteers were invited. Of the of the user feedbacks, 97% (35/36) were moderate. Our results show that ARBIN can achieve a 3 to 5 m accuracy, and provide users with correct instructions on their way to the destinations. ARBIN proved to be a practical solution for indoor navigation, especially for large buildings. To further enhance user experience, in the future we plan to extend the capability of ARBIN by adding landmark objects into real-world environments, and showing advertisement messages provided by a surrounding information system.