This paper presents a social agent pedestrian model based on experiments with human subjects. Research studies of criminology and environmental psychology show that certain features of the urban environment generate fear in people, causing them to take alternate routes. The Crime Prevention Through Environmental Design (CPTED) strategy has been implemented to reduce fear of crime and crime itself. Our initial prototype of a pedestrian model was developed based on these findings of criminology research. In the course of validating our model, we constructed a virtual environment (VE) that resembles a well-known fear-generating area where several decision points were set up. 60 human subjects were invited to navigate the VE and their choices of routes and comments during the post interviews were analyzed using statistical techniques and content analysis.
IntroductionIt is natural to combine human character animation and artificial intelligence (AI), particularly, agent technology because an agent represents an autonomous entity like a human that behaves independently. Agent technology becomes more appealing when crowd simulation is needed because it is economic to use multi-agent systems. Pedestrian or crowd simulations have been studied for decades using agent technology.1,2 The AI community has been focusing on developing a comprehensive agent architecture so that the agent behaves realistically (in either 2D or 3D), whereas the computer animation community has been trying to produce realistic 3D animations using agent technology. However, there have not been many studies in connection with social science research. Currently, many social science researchers are beginning to seek ways to visualize or simulate their findings using available technologies. Our interdisciplinary research is a result of the demands from both applied and social science communities, particularly, computer science and *Correspondence to: A. Park, School of Interactive Arts and Technology, Simon Fraser University, 250-13450 102 Avenue, Surrey, BC V3T 0A3, Canada. E-mail: aparkd@sfu.ca criminology. Our goal is to develop a pedestrian model based on the findings of social science research and to validate it. Many research studies in criminology and environmental psychology have discovered that people are afraid of certain features of the urban environment such as narrow passageways, hidden space created by corners, tall bushes, and dumpsters. They also fear the confrontation with potential offenders who might be hiding themselves using these environmental characteristics. If they sense possible dangers or risks, they take alternative routes to reach their destinations. We employed these findings in our pedestrian agent model and simulation system. In order to validate our model, we constructed a VE resembling a well-known fear-generating urban area. We intentionally set up five decision points with distinct environmental features that we wanted to test. We then recruited 60 subjects to navigate our VE making decisions at the decision points....