Benchmarking Pedestrian Dynamics Models for Common Scenarios: An Evaluation of Force-Based Models

TL;DR

This study addresses the gap in evaluating force-based pedestrian dynamics models at moderate-to-low densities by conducting controlled experiments across four common scenarios. It introduces a two-stage benchmarking framework (eligibility on 80% success, followed by cutoff-based scoring) and evaluates five models (UPL, SFMc, SFMe, CFMc, CFMe). Findings show significant shortcomings in predicting real-world pedestrian behavior, with UPL performing best at 57.14% but still exhibiting limitations, and MOSP obstacle-rich cases proving particularly challenging. The work provides a practical benchmarking baseline and highlights the need for model improvements to reliably predict everyday navigation in dense, obstacle-rich environments, such as those common in India.

Abstract

Extensive research in pedestrian dynamics has primarily focused on crowded conditions and associated phenomena, such as lane formation, evacuation, etc. Several force-based models have been developed to predict the behavior in these situations. In contrast, there is a notable gap in terms of investigations of the moderate-to-low density situations. These scenarios are extremely commonplace across the world, including the highly populated nations like India. Additionally, the details of force-based models are expected to show significant effects at these densities, whereas the crowded, nearly packed, conditions may be expected to be governed largely by contact forces. In this study, we address this gap and comprehensively evaluate the performance of different force-based models in some common scenarios. Towards this, we perform controlled experiments in four situations: avoiding a stationary obstacle, position-swapping by walking toward each other, overtaking to reach a common goal, and navigating through a maze of obstacles. The performance evaluation consists of two stages and six evaluating parameters - successful trajectories, overlapping proportion, oscillation strength, path smoothness, speed deviation, and travel time. Firstly, models must meet an eligibility criterion of at least 80\% successful trajectories and secondly, the models are scored based on the cutoff values established from the experimental data. We evaluated five force-based models where the best one scored 57.14\%. Thus, our findings reveal significant shortcomings in the ability of these models to yield accurate predictions of pedestrian dynamics in these common situations.

Figures (6)

• Figure 1: Controlled experiments illustrating four real-life scenarios. (a) Single Obstacle Single Pedestrian (SOSP): a volunteer passing through a non-living, human-sized stationary obstacle to reach their goal. (b) Head-On: two volunteers approaching from opposite directions, swapping their initial positions. (c) Parallel Ped: a faster volunteer overtaking a slower one while walking in the same direction aiming to a common goal. Snapshots at three different time intervals display the initial condition ($t = 0$), midway condition ($t = t_1$), and final condition ($t = t_2$) for head-on and parallel ped scenarios. d) Multiple Obstacles Single Pedestrian (MOSP): a volunteer navigating through a randomly placed maze of obstacles to reach their goal.
• Figure 2: Schematic of all the experimental setups within a boundary defined as $x = 0\;m$ to $x = 10\;m$ and $y = -1.75\;m$ to $y = 1.75\;m$. The measurement region spans from $x = 2\;m$ to $x = 8\;m$. For future reference, volunteer/simulated pedestrian A is represented by a filled circle, and an empty circle represents volunteer/simulated pedestrian B.
• Figure 3: (a) Trajectories for all models show overlap, showing no transverse displacement due to collinear forces. Pedestrians get stuck before the obstacle in all models except for SFMc, where pedestrian can pass through the obstacle due to low repulsion. (b) Normalized speed plotted against the x-component of the pedestrian's position. Speeds for all models approach 0 m/s, indicating pedestrians get stuck, except for SFMc. SFMe exhibits oscillation before getting stuck, indicated by negative velocity.
• Figure 4: SOSP: (a) Trajectories and (b) Normalized speed for experimental and simulated trajectories. The experimental data shows anticipatory direction changes with minimal speed change to avoid the obstacle. In contrast, simulated models, except SFMc, exhibit sudden direction changes near the obstacle, resulting in velocity dips. Overlaps with the obstacle are noted for SFMc, SFMe, and CFMc, with no oscillation or backward motion ($V < 0m/s$) observed in any model.
• Figure 5: Head-On: (a) Trajectories and (b) Normalized speed (Ped A) graphs are presented for both experimental and simulated data. The experimental data indicates an anticipatory change in direction to avoid collisions, with minimal speed fluctuations. In the trajectory plots, solid lines with filled circles represent pedestrian A, while dashed lines with empty circles represent pedestrian B. The velocity profiles show a dip when avoiding collisions in all models except SFMc.