A Smart Phone based Multi-Floor Indoor Positioning System for Occupancy Detection
Md Shadab Mashuk, James Pinchin, Peer-Olaf Siebers, Terry Moore, University of Nottingham, UK
To understand how to better manage resources in urban environment it is imperative that we develop predictive models related to the behaviour of building occupants. The energy demand and performance of a building depends on the behaviour of occupants engaged in various activities. In office buildings it can be electrical equipment, heating, ventilation and lighting. In residential buildings there are household appliances consuming water as well as electricity. To understand these energy demands it is important to understand how the building spaces are being used by individuals i.e. the occupancy pattern of individuals. At present there is a lot of research being done simulating building environment with artificial agents and predicting energy usage and other building performance related factors that helps to promote understanding of more sustainable buildings. All these simulation tools try to incorporate models of interaction and most importantly models of occupancy. There has been previous work in detecting occupancy of building users using PIR sensors or ambient sensors detecting changes in carbon dioxide composition or sound. The major shortcoming of the previous methods is the sensors limitation in detecting presence and absence of individual occupants in the room only. It is not possible to know the transition of movement of a person e.g. from office room to meeting room. The occupancy pattern of individual building users shows either presence in a room or absence from the building. Indoor positioning technology can help overcome this problem by detecting movement profile of an individual’s journey throughout the building.
An understanding of occupancy in the building environment requires a robust indoor positioning solution to be deployed. Although the positioning system might not require a very high degree of positioning accuracy, the application still requires correct identification of transitions between zones and rooms, and entry and exit from the building. These details together help to get a clear picture of the journey the occupant makes which has possible applications in built environment modelling. The use of Wi-Fi in radio map fingerprinting and Bluetooth Low Energy as a proximity sensor has seen extensive research in indoor positioning over the last couple of years. It should be noted that most of the previous research was based on trials and tests in controlled environments giving generally good results with average static positioning accuracy between 3 to 5 m but deteriorating significantly when in motion. This can be further improved with advanced pedestrian dead reckoning solution using sensors like UWB and foot mounted inertial sensors. This sensor though requires involvement of devices which has to worn and is not very convenient for participants to be in their natural state of occupancy over a long period of time like typical office hours. Thus it becomes a very different scenario when it comes to deployment of indoor positioning system in large areas involving continuous data collection over a long period of time in an environment like academic or office buildings.
The use of BLE in indoor positioning research has made a lot advances recently due to the fact that it is extremely cost effective, easy to deploy and has the potential to be used in a variety of context. Due to its short but adaptable range it has been very appealing for research in industries related to smart cars, supermarkets or hospitals requiring proximity based localization. At the same time due to the widespread availability of Wi-Fi it is also very convenient to be used for indoor localization. Both of these technologies have been used as an aiding positioning sensor with more accurate sensors like UWB and Inertial sensors in multisensory integrated indoor positioning solution. The availability of both Wi-Fi and BLE sensors in smart phones makes it more convenient for the user to easily carry potential positioning capability on their own pocket and although the hardware capability of inertial sensors like accelerometer and gyroscope may fall short of widely used state of the art foot mounted Inertial MEMS sensors it nevertheless allow for research opportunities in novel application of indoor positioning using smart phones for personal positioning.
The research presents a novel application of indoor positioning technology proposing a prototype positioning system for occupancy detection in building environment. The solution is based on a particle filter that combines Bluetooth Low Energy and Wi-Fi fingerprinting together, which can be quite time consuming and physically exhausting as well. The positioning algorithm utilizes environmental features like map matching and computes relative heading change information from a Orientation filter, based on accelerometer and gyro data from the smart phone. The entire proposed positioning solution is dependent on a simple smart phone to be carried by participants in their pockets all the time and carry on with their daily work in an office environment. Due to the absence of foot mounted inertial sensors accurate step detection is not possible since the smart phone will be in the pocket; instead filter particles are propagated forward by a simple but novel motion detection algorithm using accelerometer data from the smart phone. The motion detection algorithm looks for step like motion using a simple pattern matching algorithm in every constant short interval of time. Transition between floors is detected using Bluetooth Low Energy that can be used as a trigger on and off the stairways by reducing the range of BLE ibeacons and deploying them strategically along the stairways.
This paper starts with a brief review of current occupancy detection methodology and indoor positioning technology. The following sections discusses experimental results of the relative heading computation based on the gyro data and the results from motion detection involving trials to account for different walk patterns (fast and slow). We then discuss the Wi-Fi and Bluetooth fingerprinting and static positioning accuracy when both Wi-Fi and BLE fingerprinting are combined compared to their accuracy when used on their own. Finally different individual sensor data are integrated in a simple multisensory integrated indoor positioning system and the results of actual trials carried out throughout the building are discussed. The fact that smart phones are easy to carry helps participants carry on with their usual daily work without any distraction but at the same time provide a reliable pedestrian positioning solution for detecting occupancy information.