Social lane-based cognitive model for simulating pedestrian flow in games

In computer game software, the implementation of simulated urban crowds is widespread. Representation of pedestrians in computer games has, to date, lagged behind what has been shown possible in academic studies and simulation software. Primary reasons for this are the strict CPU budgets that game AI has to function under, and algorithm implementation which is frequently complex for even a baseline approach (i.e. before performance optimisations). This paper presents a novel lane-based approach to pedestrian AI which combines the verisimilitude of AI based on behavioural studies, with the intuitive implementation and minimal computation cost required for games development. To achieve this, each agent filters its surroundings into virtual lanes, and then sidesteps or continues forward as dictated by social convention.