Improving the Legibility of Abstract Spaces

Legibility in City Environments

The basis for the current research is the book The Image of The City by Kevin Lynch [Lynch60]. In this work he defines the legibility of a city as "...the ease with which its parts may be recognised and can be organised into a coherent pattern." This is a reference to the formation in a persons mind of a cognitive map. This is a mental image of the environment which we use as a reference when performing wayfinding tasks. Lynch's work identified five features of city landscapes which are used as the building blocks of cognitive maps:

Landmarks

Static and recognisable objects which can be used to give a sense of location and bearing.

Districts

Sections of the environment which have a distinct character which provides coherence, allowing the whole to be viewed as a single entity.

Paths

Major avenues of travel through the environment such as major roads or footpaths.

Nodes

Important points of interest along paths, e.g. road junctions or town squares.

Edges

Structures or features providing borders to districts or linear obstacles.

We aim to make information spaces more legible by providing these features as part of the virtual environment.

The LEADS System - Design

We are currently developing a system which will provide the legibility features as a layer on top of the existing information visualisation and PITS systems available within the group.

The system is designed to be applied to a specific type of data space where the enhancements provided will be most effective. These spaces should satisfy three criteria: be persistent over relatively long periods of time; be relatively stable so that they evolve over their lifetime and are rarely disturbed by major upheavals in the database; be accessed repeatedly by a number of independent users. The criteria ensure that the features may be incorporated into cognitive maps over a period of time from experience of the space and be relatively stable so as to be useful references.

To place the legibility features LEADS uses districts as a starting point as this allows for a number of relatively simple techniques to be used to form the other features. Districts in a data space will be clusters of items which have some sort of internal similarity. LEADS applies a clustering algorithm to the data to identify these groups automatically. The algorithm chosen for the initial implementation is Zahn's Minimum Spanning Tree Algorithm [Zahn71], which works by forming a minimal spanning tree of the distances between the data items and then walking the tree to remove links which are significantly longer than others nearby. The sub-trees resulting from the use of the algorithms each form a district in the space.

Landmarks need to be placed in a position where they will be useful for navigation. We have used a premise that one such place will be where districts are dense. The first step in positioning a landmark is to find groups of three clusters which are mutually adjacent. A landmark will be placed in a position which is fairly central to these districts. This is found by finding the centroids of each district and placing the landmark at the centre of the triangle they form.

Edges in LEADS are structures which help to define the borders of districts. They are placed in the space between those districts which are significantly large. To accomplish this they are placed between the nearest neighbours in the cluster and aligned along the line that joins them. In most cases, especially where the clusters are essentially spherical, this results in an edge placement which effectively separates the space but does not cut into the individual districts.

Nodes and paths are co-dependent in the LEADS system. Eventually the aim is to have these features evolving out of the use of the data space. The initial prototype currently identifies nearest neighbours between districts as nodes and places a path between them. Within districts all those items identified as nodes are joined by a spanning tree.

The Leads system also aims to provide other tools to aid navigation in the virtual environments. Signposts are placed near the major nodes to aid in wayfinding and an optional "axes object" is available to float near the viewpoint and remain aligned with the major axes to provide a sense of orientation to the user which may be lacking in abstract environments without ground-planes or horizons. Future developments will include history and backtracking mechanisms and a more developed signposting system.

The LEADS System - Implementation

LEADS has so far been applied to the visualisations produced by three of the systems available in the group. The first, Q-PIT, is a system which works on databases where the items have a number of well defined fields [Benford94]. Three of these fields may be chosen so that the values they contain are mapped onto the three major axes of the space to give the position of the data items. The remaining fields may be used to define aspects of the representation of the items such as their shape, colour or speed of rotation.

The second system, FDP, is a 3D graph drawing tool. This takes a representation of an arbitrary network and produces a 3D visualisation by representing the nodes as balls and the links as springs. Initially the items will be placed randomly and the system will then go through a cycle of repositioning the nodes based on the tension in the springs until a relatively stable formation is found.

The third system, VR-VIBE, is a statistics based system for documentation visualisation which works by defining points of interest (POIs) containing keywords which are used as queries to a document database. The POIs are then positioned in the space and the matching documents are arranged betwen them with the position depending on the relative attraction to each POI.

References

[Benford94] Steve Benford, John Bowers, Lennart Fahlen, Chris Greenhalgh, John Mariani and Tom Rodden, Networked Virtual Reality and Co-operative Work, Presence, MIT Press, (in press).

[Lynch60] Kevin Lynch, The Image of the City, M.I.T. Press, 1960

[Zahn71] Zahn, C.T., Graph-theoretical Methods for Detecting and Describing Gestalt Clusters, IEEE Transactions on Computers, C 20, 68-86

Links

The following are links to related pages: