Full e-thesis link is provided below. If the link is broken, the thesis can also be found in the University of Nottingham e-Theses Repository.
Abstract:
Assessing building performance related to energy consumption in the post-design-occupancy stage requires knowledge of occupancy patterns together with many other behavioural factors. Typically to evaluate building performance in the post-occupancy-evaluation (POE) period, it would require users to participate in occupancy detection trials covering the building and generate occupancy-related data. This would allow architects and built environment professionals to understand the current occupancy pattern and evaluate by comparing it with the predicted performance done at the design stage. Due to the limitations of passive sensors in detecting an individual’s occupancy throughout the building, indoor positioning can provide a viable alternative but requires the deployment of a robust positioning solution. It also demands a user-friendly device capable of providing positioning-related data that users can carry around without any obstruction or hindrance to their daily work. Most modern smartphones come with an array of sensors quite capable of providing the desired solution. These requirements create the opportunity for research in new application areas of indoor positioning to solve real-world problems such as understanding building occupancy.
It is well-known that Global Navigation Satellite System (GNSS) based positioning performs poorly in the indoor environment, but alternative options making use of signals of opportunity from existing wireless devices and sensors can provide reliable and to some extent, very high accuracy indoor positioning solutions. Some common examples of indoor positioning systems widely researched are based on Wi-Fi and Bluetooth Low Energy (BLE), inertial navigation solutions using foot-mounted or smartphone Inertial Measurement Unit (IMU), Ultrawideband (UWB) or Radio Frequency Identification (RFID). Unfortunately, the use of such technology is dependent on the context of usage and most of them have not yet been commercialised. After careful exploration of the current state of research related to indoor positioning technology, techniques, positioning capability, infrastructure costs and ease of deployment, an indoor positioning system is proposed which makes use of the smartphone and its built-in integrated sensors: Wi-Fi, Bluetooth, accelerometer, and gyroscope. The positioning system uses the traditional Wi-Fi and Bluetooth fingerprinting in combination together with pedestrian dead reckoning to develop a cheap but effective multi-floor indoor positioning solution. The combination of all these performance critical elements involved in the positioning system requires continuous collection of Wi-Fi and BLE signal observations from the fingerprint radio map, checking step events and heading and stairwell detection. These are all fused in a particle filter engine for positioning state estimation epoch-by-epoch in a multi-sensor pedestrian dead reckoning solution.
To test the capability of the proposed positioning system each of the dead reckoning elements; step detection and heading estimation have been thoroughly tested in an objective manner and the positioning system validated by carrying out experimental trials against ground truth obtained from a Robotic Total Station. The relevant positioning algorithms and techniques were carefully chosen to ensure ease of implementation with minimum complexity without compromising the core performance specification necessary for occupancy detection.
Real-life occupancy is very complex, and an extensive trial in a real-world setup is necessary. The potential of the proposed prototype positioning system was investigated by involving non-expert participants in real-world trials to show the limitations and effectiveness of using such systems for detecting occupancy. The trials were designed in the form of spatially distributed tasks to imitate real-life occupancy. Participants were required to move around without any guidance to cover different parts of an office building with multiple floors. The appropriate ethics approval was obtained for the study beforehand and participants were recruited with their explicit consent. In total, fifty trials were carried out by around twenty-two participants. The trials comprised of a variety of routes and floors of a sample office building within the limited time scale of the PhD research. All the trials were validated by comparing the estimated occupancy against real occupancy recorded by the researcher who shadowed the participants during the trials. The trial results were analysed to demonstrate the level of accuracy achievable in terms of occupancy against a specific set of performance metrics. Sensor deployment challenges faced during the research are discussed as well.
Finally, to demonstrate a possible contribution, an agent-based building occupant model is developed to understand the local dynamics of electricity consumption using different lighting policies with inputs from the developed indoor positioning system. The model uses the Nottingham Geospatial Building’s lighting and floor plan configuration as an example. The energy usage obtained from simulation for the default lighting policy is validated using real electricity data from the local energy supplier, the University Estates Department for any required calibration of the model. One example is then chosen to illustrate how the positioning data collected from observed individuals’ trajectories from the original trials contribute to the understanding of electricity usage in multi-floor offices when a different lighting policy is applied on the base model and any potential impact analysed.
This research has intended to demonstrate how indoor positioning data can be applied to the improvement of building occupant modelling and what the shortcomings and limitations of this approach are. Shortcomings of the positioning system, when it comes to detecting occupancy, have been identified from the multi-user trials. Simplifications and assumptions made when implementing the algorithms for individual elements of the positioning system, especially motion detection and heading estimation, have been discussed together with their potential implications on the expected accuracy of the positioning system. Key areas of improvement, choice of technology, techniques and application methodology were identified for future work related to the development of the positioning solution.
Faculty of Engineering
Supervisors: Terry Moore, Peer-Olaf Siebers, James Pinchin
Keywords Building Occupancy, Building Performance, Electricity Consumption, Modelling, Simulation, Agent Based Modelling, Indoor Positioning, Smartphone, Inertial Sensors, Sensor Fusion, Bluetooth, WiFi, Accelerometer, Gyroscope, Fingerprinting, Radio Map, Particle Filter
