Landmark-Based Pedestrian Navigation Using Augmented Reality and Machine Learning

The prevalence of smartphones and tablets featuring various kinds of sensors and the improvements in computation capabilities of those devices, have led to an acceleration of using geospatial data in many domains. The large number of sensors deployed on the devices has made it possible to detect user’s location, heading and orientation as well as getting contextual information from various sources of online data. Combining the stream of data from positioning and orientation sensors with camera, has also made it more feasible to deploy practical Augmented Reality (AR) applications on mobile devices. This paper, explains a system and its related study that provides a view of the navigation experience which composed of the AR view as well as continuous personal feedbacks about the relative location of the user in relation to the closest landmarks. In the system, navigation and path finding are based on landmarks. Relative position of the user with regard to landmark is determined using GPS sensors as well as image processing algorithm for finding distance from a landmark. In addition, feedbacks for navigation instructions are customized for each user based on his/her movement profile and use of continuous tuning of a machine learning algorithm. Experiment of using the system showed a significant improvement in acquiring of spatial knowledge for the users in comparison with turn-by-turn systems.