The Integration of Agent-based Model and Social Force Model: Realistic Pedestrian Simulation

Abstract

The fitted pedestrian modeling is vital to construct realistic pedestrian movement dynamics based on the simulation objective. The inclusions of realistic human characteristics are essential in building a realistic pedestrian model to re-enact the actual movement of the crowd, especially during a panic situation. This study compared the discrete approach CA model with the continuous approach ABM model and the enhanced ABM-SFM model using 30 pedestrians in a sample layout to validate and improve the near realistic pedestrian simulation model. The ABM and ABM-SFM models were integrated with the Pythagorean Theorem (PT) to imitate the driving forces in the desired motion. The results showed that CA localization can imitate pedestrians' sudden, unorganized panic but cannot mimic human intelligence in decision-making and forces avoidance. ABM and ABM-SFM models were on par for the first 10 pedestrians evacuated. However, the ABM-SFM model reduced travel distance by 37.50% and evacuation time by 9.04% after 20 pedestrians were evacuated. Overall, the travel distance for escaped 30 pedestrians with the ABM-SFM model was 39.29% longer, and the evacuation time was 13.96% longer than ABM. The ABM model simulates the pedestrians’ movement towards the nearest exit point to form the “fast escape” approach. In contrast, the ABM-SFM model simulates the “safe escape” to provide balance usage of exit points.