Detecting and Correcting for Human Obstacles in BLE Trilateration Using Artificial Intelligence

Abstract

One of the popular candidates in wireless technology for indoor positioning is Bluetooth Low Energy (BLE). However, this technology faces challenges related to Received Signal Strength Indicator (RSSI) fluctuations due to the behavior of the different advertising channels and the effect of human body shadowing among other effects. In order to mitigate these effects, the paper proposes and implements a dynamic Artificial Intelligence (AI) model that uses the three different BLE advertising channels to detect human body shadowing and compensate the RSSI values accordingly. An experiment in an indoor office environment is conducted. 70% of the observations are randomly selected and used for training and the remaining 30% are used to evaluate the algorithm. The results show that the AI model can properly detect and significantly compensate RSSI values for a dynamic blockage caused by a human body. This can significantly improve the RSSI-based ranges and the corresponding positioning accuracies.

Monitoring of Cow Location in a Barn by an Open-Source, Low-Cost, Low-Energy Bluetooth Tag System

“…The RSS of the RuuviTag studied in this research had a relatively high noise level and non-gaussian distribution (Figure 5a) similar to [12,27] and others. Particularly, it was caused by the aggregation of RSS received by different channels (37, 38 and 39, separated by [26]), which can be seen in the first image in Figure 5a. However, in the barn environment, the RSS distribution was closer to gaussian (Figure 5b).…”

“…Further improvement of the localization algorithm can be done in the following parts of the system. The RSS noise can be decreased by the separation of the BLE signal channels, as was done in [26] and [22]. The Viterbi algorithm can be improved by more detailed description of the cow behavior by the Markov chain, changing a method for calculation of the location probability in the mapping points and using additional information about the cow behavior measured by other sensors (such as cow body position estimated with the help of tag accelerometers) similar to [16].…”

“…The location accuracy of the system covering the entire barn was 3.27 m and 1.9 m for the smaller area (4 × 5 m 2 ), which makes the system much less accurate than systems based on UWB technology reaching 0.4–2 m of accuracy in entire barns [ 2 , 4 , 5 ], but with similar accuracy as the research system based on BLE, such as 4.2 m at [ 8 ] in the cow barn environment and 2.4 m (for 90% of tested cases) at [ 22 ], 3.8 m at [ 26 ] and 2.3 m at [ 16 ] in the office environment. The developed localization system can be used to recognize preferable cow locations, such as areas for lying ( Figure 13 a), feeding and crowded locations, such as queuing to the milking robot ( Figure 13 b) and a rough distribution of the cow activity ( Figure 13 c).…”

“…The developed signal propagation model included the geometry and material properties of the structures of the tag environment as well as the mapping of RSS depending on what the tag orientation was used in [16,24]. The RSS decreasing while passing through the dairy cow's body was modeled [25], and human-body signal shadowing was mitigated [26]. Algorithms for learning the environment features were used for the localization [27].…”

Sensors and Sensing Technologies for Indoor Positioning and Indoor Navigation

“…In addition, Bluetooth Low Energy (BLE) is commonly used as wireless technology to address indoor positioning. The authors in [ 16 ] face the challenges related to RSSI fluctuations and the effect of human body shadowing when using BLE. In order to mitigate these effects, the paper proposes a dynamic Artificial Intelligence (AI) model that uses the three different BLE advertising channels to detect human body shadowing and compensate the RSSI values accordingly.…”