Intelligent Characters for Next-Gen AR in Museums

Abstract

In this master thesis we develop an intelligent virtual character as part of an augmented reality experience. The character serves the purpose of guiding the user through a scripted museum tour and automatically finds its way in a dynamic environment. We implement a system that uses an abstract map in the form of a waypoint graph for the broad scale navigation and infers walkable areas from real-time 3D scans of the environment, provided by the LiDAR scanner of an iPad Pro. To test and evaluate the system we performed a user study, where we gathered data to compare the effects of different environments and pre-existing 3D scans on the system’s performance. On one hand the study revealed that the system benefits from pre-existing scans. On the other hand we found the system to deliver a satisfying guiding performance in an initially unknown environment as well, succeeding to lead the users through the museum tours, all while being robust to occurrences of drift and the behaviour of the users in most scenarios.