Vizagent
Planning with virtual alpine landscapes and autonomous agents

The visual landscape

Abstract

The visual quality of the landscape is an important but often neglected factor in landscape and environmental planning. Especially for the recreation potential, the quality of the visual landscape plays a key role.

Up to now, the judgement of human preferences about the landscape and landscape elements was based on the existing landscape, either by monitoring people in the real environment or by exposing test persons to visual surrogates such as photographs of the real landscape.

Recent developments in the virtual representation of landscapes suggest that it might be possible to reliably observe the behaviour and deduct landscape preferences of people in a more controlled 'laboratory environment' as it is provided through new developments in computer and projection technology. Such visualisations permit to show the visual landscape both at present and at potential future situations, giving a tool to systematically investigate how people react to changes in the landscape.

Such a visualisation of a changed landscape would be unpopulated, since no data about human behaviour existed for this changed landscape. Autonomous agents modelling, which is based on concepts from artificial life and computer simulations, enables one to populate the virtual world with rule-driven agents which can act as surrogates for real humans.

The research project seeks to visualise, model, and compare the judgments made by individuals in the real environment with judgments of people in the virtual environment and with judgments gained by autonomous agents modelling. This will be a central issue in calibrating the models.

The implementation will be via three threads. One will be the investigation of recreationist's behaviour in the real environment. This will be used as a benchmark. The second thread will be to generate a virtual representation of the real environment and then to investigate the behaviour of humans in this virtual environment. Third and last we will model synthetic agents in the virtual environment. This will enable us to populate the environment with many agents, meaning that we can investigate agent-agent interaction and crowd behaviour after landscape changes, something that would not be possible when using real humans in the virtual landscape as in the second thread.

The proposed research approach will help to explore, test, and evaluate landscape planning, design, and management alternatives, thereby assisting in the evaluation and management of potential changes.