Abstract:
This research was commissioned by Sandfield Associates to develop a framework for geo-contextual indoor augmented reality (AR). With the majority of their clientele operating in indoor environments, Sandfield is seeking to extend their commercial offerings into the field of geo-contextual AR, due to its potential to greatly improve business processes. Two use cases are presented at the end of this thesis to illustrate the value added by this technology in a commercial setting. Geo-contexual indoor augmented reality (AR) requires improved contextual specificity in order to become well-established in commercial settings. Many methods of providing contextual specificity require precise knowledge of the pose and location of a user in relation to their environment. In outdoor applications, GPS based methods have become ubiquitous for determining location. Unfortunately, GPS is unsuitable for use in indoor environments due to the attenuation encountered as a result of the building’s structure. For this reason, we turn attention toward finding alternative methods of indoor localisation. In this research, we outline a framework for developing geo-contextual mobile augmented reality applications. The required localisation is performed using an extended Kalman filter (EKF), which combines displacement measurements produced through pedestrian dead reckoning (PDR) with RSSI measurements retrieved from a set of Bluetooth low energy beacons in order to estimate the user’s location. In addition to the localisation algorithm, a method of dynamically initializing the environment is also proposed. This method uses particle filtering to estimate the true locations of the beacons based on measured RSSI and the output of the PDR subsystem. This initialization routine requires no prior knowledge of the environment, or the beacon locations. The performance of both algorithms is evaluated in a 54m2 rectangular environment with a beacon placed in each corner (a total of four). The dynamic initialisation algorithm successfully initialised four estimated beacon positions with an RMSE of 1.51m from their true locations. Furthermore, the localisation algorithm estimated the position of a static device with an RMSE of 0.82m, and could track a moving device with an RMSE of 1.21m.