Online Content Generation in Mobile Applications - Adaptation and Personalization for Location-based Game Systems

Abstract

Location-aware mobile applications like navigation services or fitness trackers are inherently incorporated into our everyday life. The success of Ingress and Pokemon GO popularized the genre of location-based games, even indicating positive health effects for avid players that voluntarily cover large distances. However, the inconsistent and area-dependent supply of playable content, combined with superficial cooperation mechanisms, pose a considerable challenge for novel games in this genre. Furthermore, a suitable content generation based on real-world location data is complicated by the heterogeneity of populated areas. Continuously, in these movement-based games an inflexible mobility integration penalizes players deviating from the game’s intended style of play. Online procedural content generation, which creates content during run-time, is well suited for such scenarios. In numerous application domains, online content generation is utilized based on individual scenario parameters and contextual user data, providing personalized outcomes. But the optimization of content availability based on real-world location data, combined with a dynamic online generation, remains an open research challenge. In this thesis, we design and implement automatic location-based content generation mechanisms as our first contribution, which address the aforementioned challenge. We develop an approach to identify suitable content locations and procedurally derive game areas. Consequently, we enable cooperation concepts by game area coupling, providing distinct game areas with similar urban characteristics, even in heterogeneous areas. Since movement is the main characteristic of such games, we propose a mobility personalization strategy as our second contribution. This incorporates a player routing approach in urban scenarios focusing on guided virtual reward optimization. On this basis, we develop a run-time mobility detection utilizing accumulated smartphone sensor data as our third contribution. To support this decision-making process, we employ a combination of accelerometer and location sensor readings to overcome the limitations of individual data sources for specific mobility types. Finally, based on our developed GeoVis prototype, we conduct an extensive evaluation of our contributions. Therefore, we assess the effectiveness of our content approach under varying ambient conditions. We show that our location-based content generation leads to a reliable and flexible content selection in inhabited regions. Additionally, we demonstrate the utilization of route planning and automatic mobility detection as suitable mobility personalization strategies. In summary, we show that our contributions allow novel location-based game systems to create content worldwide and adapt to application-specific area and mobility characteristics by examining user mobility data during run-time.