Multi Agent Visualisation and Interactive Rendering

Addressing the fast simulation of accurate, dense crowds. This involves harnessing the GPU architecture, which has a theoretically higher computational throughput than equivalent generation CPU. The simulations are an agent-based model, which contains inherent parallelism, which is suitable for running on the GPU.

The work involves examining existing models of pedestrian collision avoidance, and creating novel extensions on the GPU, that can better deal with simulating more accurate people within a dense crowd. The work examines the effects of dense crowds and the collision avoidance that occurs, and attempts to address issues that arise in high densities by construction of new simulation models.

Research questions: What is the current state of simulating crowds? How true-to-life are the current models?

Approach: Development of pedestrian collision avoidance models using GPU architecture

Key aims: develop collision avoidance simulation models that can run for many people, and provide accurate behaviour in dense crowds. Validate developed models against real behaviour.