Source: Journal of WSCG vol.14, 2006
Keyword: Crowd Navigation, Behavioral Navigation, Pedestrian Simulation
The focus of this paper is on the collective behavior of pedestrians.
Problem
In general path-planning algorithm are not a good option for dense crowd navigation because continuous replanning would be needed due to the high number of dynamically moving objects.
The flow grid mechanism is introduced to simplify navigation and enable crowd self organization.
The avatars avoid area of high density or high opposing flow, thus lane formation takes place.
Measuring the Flows
The flow grid measures the densities of pedestrians and their velocities at various directions.
Each avatar registers his position and velocity on the flow grid as soon as he moves to a new position.
The velocities are separated into X and Z axis components. Positive and negative axis are stored separately, Thus four velocity values are stored at each point, (+Vx, -Vx, +Vz, -Vz).
Each avatar is registered on the grid by distributed his density and velocity to the 4 neighboring points as shown in figure below.
Using the Flow Grid to Navigate
The flow grid is used to extract information for navigation purposes.
Densities and velocities are interpolated between grid positions when information is needed at any in-between point.
By consulting the flow grid at regular intervals the avatar choose to head for the area with smaller density and smaller opposing flow.
To help the avatar decide which area is best, a special weight formula has been constructed. This formula is explained in the box below:
The weights are calculated each time the avatar needs to find a new intermediate destination.
The spot with the lowest weight is chosen as a temporary local target.
No comments:
Post a Comment