A model for mobile spatial services

Abstract

Mobile context-aware computing is a paradigm in which mobile devices have access to information, known as context, about the situation in which they are being used, and dynamically adapt application behaviour to support user tasks and mobility. A primary concern of mobile context-aware computing is awareness of the physical environment surrounding a user. This concern is addressed through spatial-awareness, where mobile applications employ knowledge about the physical location of real world objects to compute, for example, estimated journey times, routes between activities, perform proximity-based information retrieval, and render map-based interfaces. Geographic Information Systems (GIS) have traditionally performed these operations on static spatial data, vector-based geometry and attributes describing real world objects. However, spatial data and supporting services are confined to central servers, accessible to mobile devices in the form of a static graphic representation via wireless networks. The range of services that spatially-aware applications on hand-held mobile devices can provide are limited by technical factors such as the inherent unreliability of wireless networks and the limited nature of mobile devices in terms of battery power, memory constraints and screen size. Because these limitations are common to all mobile applications, a generic model for spatial services is needed that is designed not to overburden the limited resources of mobile devices and is not dependent on continuous network connectivity. This thesis proposes a model for spatial middleware featuring algorithms designed to minimise the processing time and power consumed on hand-held mobile devices, while providing uninterrupted access to common spatially-aware application services, such as rendering of geospatial information, real time generalisation of dynamic scenes, route generation, and visibility determination. The algorithms included in the model are designed to reduce the complexity of spatial data, thereby reducing the processing

time and power consumed while manipulating it on mobile devices. These algorithms include a multiple representation database designed to approximate continuous scale adaptation with stepped levels of detail. Access to this data structure is facilitated by a hierarchical spatial index that uses minimum bounding boxes to approximate more complex spatial objects. The levels of detail are computed using algorithms that eliminate objects based on rendered size at particular scales and simplify geometry using shape and location preserving algorithms. Limited processing resources are further preserved by clipping geometry that extends beyond the viewport of the device to avoid computing projections and rendering coordinates that will not be seen. A graph-based topological representation of spatial data is searched using pluggable graph traversal algorithms with configurable cost functions to provide navigation capabilities. The increased accessibility of spatial information offered by the model allows for the development of innovative services such as visibility determination, which expands mobile context-aware computing's environmental awareness beyond the physical location of real world objects to include the visibility of those objects. Spatial objects within the applications' viewport are filtered based on a field of view determined by compass direction and the human visual system. The resulting objects are then stored in a depth bu er on which a variation of occlusion culling is performed on their bounding volumes to reduce the set of geometry in the buffer to a possibly visible set (PVS). Spatial services are incorporated into a generic framework, supporting the development of spatially-aware applications. Performance evaluations demonstrate that spatial services such as map rendering, generalisation, route generation and visibility determination can be provided locally on mobile devices. In addition, empirical experiments demonstrate that the model for spatial middleware presented in this thesis is is more energy efficient than existing server-based approaches. The reusability and extensibility of the framework to support the development of a range of mobile, spatially-aware applications is evaluated through the development of a case study application.