Indoor positioning model based on people effect and ray tracing propagation

Abstract

WLAN-fingerprinting has been highlighted as the preferred technology in an Indoor Positioning System (IPS) due to its accurate positioning results and minimal infrastructure cost. However, the accuracy of IPS fingerprinting is highly influenced by the fluctuation in signal strength as a result of encountering obstacles. Many researchers have modelled static obstacles such as walls and ceilings, but hardly any have modelled the effect of people presence as an obstacle although the human body significantly impacts signal strength. Hence, the people presence effect must be considered to obtain highly accurate positioning results. Previous research proposed a model that only considered the direct path between the transmitter and the receiver. However, for indoor propagation, multipath effects such as reflection can also have a significant influence, but were not considered in past work. Therefore, this research proposes an accurate indoor positioning model that considers people presence using a ray tracing (AIRY) model in a dynamic environment which relies on existing infrastructure. Three solutions were proposed to construct AIRY: an automatic radio map using ray tracing (ARM-RT), a new human model in ray tracing (HUMORY), and a people effect constant for received signal strength indicator (RSSI) adaptation. At the offline stage, 30 RSSIs were recorded at each point using a smartphone to create a radio map database (523 points). The real-time RSSI was then compared to the radio map database at the online stage using MATLAB software to determine the user position (65 test points). The proposed model was tested at Level 3 of Razak Tower, UTM Kuala Lumpur (80 × 16 m). To test the influence of people presence, the number, position, and distance of the people around the mobile device (MD) were varied. The results showed that the closer the people were to the MD in both the Line of Sight (LOS) and Non-LOS position, the greater the decrease in RSSI, in which the increment number of people will increase the amount of reflection signals to be blocked. The signal strength reduction started from 0.5 dBm with two people and reached 0.9 dBm with seven people. In addition, the ray tracing model produced smaller errors on RSSI prediction than the multi-wall model when considering the effect of people presence. The k-nearest neighbour (KNN) algorithm was used to define the position. The initial accuracy was improved from 2.04 m to 0.57 m after people presence and multipath effects were considered. In conclusion, the proposed model successfully increased indoor positioning accuracy in a dynamic environment by overcoming the people presence effect.