Variable-scale representation of road networks on small mobile devices

Abstract

A method is proposed for the adaptive multi-scale representation of road networks for location-based service applications. The method is able to automatically set a feasible scale according to geographic scope, the complexity of the road network, and the distance to the viewer. Moreover, the method achieves multi-scale representations of road networks on a display screen. The key steps of the method and the initial experimental studies undertaken to evaluate its feasibility are described.

Introduction

Visualization of spatial data (e.g., images, vector data) is of vital importance for spatial recognition, augmented reality, and many other different fields of application. Among these are location-based services (LBS) and car navigation systems, the typical characteristics of which are that the representation medium of the spatial data relies on mobile devices with small screens. As a consequence and because of the nature of mobile devices, display, particularly on small mobile devices, cannot simply rely on the techniques designed for traditional web or desktop applications.

The key factors affecting the display on small mobile devices can be summarized as follows:

• Data presentation and exploration on mobile devices are strongly affected by the small size and resolution of the displays.

• The computing capability is weak and the limited storage space implies that large datasets cannot be loaded on the device.

• Frequent zoom in/out and pan operations are tedious and cognitively complicated due to global context loss.

To communicate spatial data in both overview and detail, the system has to allow the user to flexibly zoom in and out (Brenner and Sester, 2005). Map data, particularly, road networks, are critical to routing, positioning, and guided navigation in car navigation systems. Extensive research has been done regarding the visualization of spatial data on small mobile devices (e.g., Harrie et al., 2002; Alan et al., 1996). Brenner and Sester (2005) implemented a method in which generalization operators were used for displaying building polygons on small mobile devices. Reichenbacher (2003) proposed a framework for mobile visualizations. Agrawala and Stolte (2001) developed techniques for the generalization of cartographic data that improved the usability of maps for road navigation on mobile devices. Their techniques are based on cognitive psychology research and are meaningful for personal navigation, as all turning points along the route are shown and less attention is paid to the length and direction of each road. Dong et al. (2007) also proposed a method to generate semantic road maps for mobile navigation. In their approach, semantic road maps are generated by distorting road lengths and angles and by simplifying road shapes. However, the maps may be too large to be displayed on the small screens of mobile devices. Several studies have explored the generation of variable-scale maps based on the principles of the Fisheye view (Sarkar and Brown, 1992), which shows a detailed representation of a circular area surrounding a point of interest (POI), e.g., a mouse point, whilst using a small scale and applying generalization and distortion operations to fit the remaining map area in the available space. However, with this technique road maps cannot be displayed on the screens of car navigation systems.

Fig. 1 shows a planned route consisting of a long and straight road. Suppose that a set of generalization operators are applied to generalize the road network. The generalized road map may still be too large to be displayed on a small screen, and the user might lose the context based on such a map. Thus, neither a generalized road map at a fixed scale nor a Fisheye view road map necessarily meets the requirements of displaying road maps on the small screens of mobile devices. In the following, a method is proposed to display variable-scale road maps on the small screens of mobile devices. First, the road network is divided into different regions of interest (ROIs). Subsequently, the road network of each ROI is generated at different levels of detail (LoDs). In the proposed method, a road network of a defined geographic extent is displayed at different LoDs. Driver support is guided along a planned route. The road network of a region close to the current location is displayed in greater detail while that at a greater distance is shown with less detail.

In the remainder of this paper, Section 2 elaborates on the proposed method while Section 3 discusses the experiments undertaken to illustrate the feasibilities of the proposed method. Conclusions are drawn in Section 4.