DAVIS: Density-Adaptive Synthetic-Vision Based Steering for Virtual Crowds

DAVIS: Density-Adaptive Synthetic-Vision Based Steering for Virtual Crowds

Rowan Hughes         Jan Ondrej         John Dingliana
Trinity College Dublin



(Top) A scene with 500 virtual pedestrians on a collision course, using our approach virtually no bottlenecking occurs. (Bottom) A scene containing 8 walkers all aiming to get to the diametrically opposed position. Solutions are shown for a number of models. (from right to left) Our model, Ond¢§rej¡¯s model, Reynold¡¯s Model, RVO-Library, Helbing¡¯s Model
 
Abstract  

We present a novel algorithm to model density-dependent behaviours in crowd simulation. Previous work has shown that density is a key factor in governing how pedestrians adapt their behaviour. This paper specifically examines, through analysis of real pedestrian data, how density effects how agents control their rate of change of bearing angle with respect to one another. We extend upon existing synthe vision based approaches to local collision avoidance and generate pedestrian trajectories that more faithfully represent how real people avoid each other. Our approach is capable of producing realistic human behaviours, particularly in dense, complex scenarios where the amount of time for agents to make decisions is limited.


Publication    

Rowan Hughes, Jan Ondrej and John Dingliana
DAVIS: Density-Adaptive Synthetic-Vision Based Steering for Virtual Crowds, MIG 2015.

Download Paper (Pre-Publish) (2.66 MB)


Demo video