A perceptual study into the behaviour of autonomous agents within a virtual urban environment

Abstract

Simulating vast crowds of autonomous agents within a procedurally generated virtual environment is a challenging endeavour from a technical perspective, however it becomes even more difficult when the subjective nature of perception is also taken into account. Agent behaviour is the product of artificial intelligence systems working in tandem, however the sophistication of these systems is not a guarantee of achieving believable behaviour. Within locations based upon reality such as an urban environment, the perceived realism of agent behaviour becomes even harder to achieve. This paper presents the development of a crowd simulation that is based upon a real-life urban environment, which is then subjected to perceptual experimentation to identify features of behaviour which can be linked to perceived realism. This research is predicted to feedback into the development processes of inhabited cities, especially those attempting to simulate perceptually realistic agents as it will highlight features of behaviour that are important to implement. The perceptual experimentation methodologies presented can also be adapted and potentially utilised to test other types of crowd simulation, whether it be for the purposes of computer games or even urban planning and health and safety.