Wei Yan Yehuda Kalay University of California, Berkeley SIMULATING HUMAN BEHAVIOR IN BUILT ENVIRONMENTS Wei Yan Yehuda Kalay University of California, Berkeley
Human behavior in built environments Sproul Plaza UC Berkeley
Behavior simulation Simulated Sproul Plaza, Berkeley
Predicting behavior Behavior patterns in New York City plazas, William Whyte, 1980
Predicting behavior Whyte’s question: Our question: why some places work well, other do not? Our question: how can we predict which places will work well, which ones will not? Behavior patterns in New York City plazas, William Whyte, 1980
Methodology Develop virtual users with behavioral traits. Add usability traits to the environment, as locational input to the VUsers. Assure that the simulated behavior corresponds to actual behavior.
Video Tracking Statistics Methodology Behavior Simulation Video Tracking Statistics Verification Environment Modeling Usability Geometry User Modeling Behavior Geometry Perception
Environment modeling Measurements in Sproul Plaza
Environment modeling 2D DXF
Environment modeling Usability Model – Discrete Space Model Step cell Cell Properties: Sittable? In the sun or in the shade? Occupied by a user? ……… Ground cell Fountain side cell Fountain water cell Bench cell Usability Model – Discrete Space Model
Environment modeling 3D VRML
User modeling Human form. Human movement. Human traits: shortest path
User modeling Human form. Human movement. Human traits: modified shortest path
User modeling – social space Personal-space bubble (Deasy, 1985) Personal space Social space (closer) Social space (farther) Public distance Proxemics (Hall, 1966)
User modeling - movement Artificial Life. Boids (Reynolds, 1999)
User modeling – goals & preferences Poisson distribution for arrival rates Sproul Plaza UC Berkeley Summer 2003
User modeling – goals & preferences William Whyte, 1980
User modeling – video tracking system
User modeling – video tracking system Target region highlighted: people at the fountain Background subtraction Foreground image Intensity thresholding
User modeling – statistics Numbers of people entering the plaza on different days Numbers of people sitting in different places on different days
User modeling – paths
Results - simulation
Results – design alternatives
Conclusions
Conclusions Computers in design processes: representation synthesis evaluation Evaluation of physical factors: structures energy etc. Evaluation of human factors: behavior learning behavior modeling/simulation