🔬 physics

A well-motivated model of pedestrian dynamics

This paper proposes and validates a dynamic motivation model grounded in expectancy-value theory that captures how fluctuating internal drives reorganize pedestrian positioning and spacing near bottlenecks, thereby producing structured crowd heterogeneity that static models fail to replicate.

Original authors: Ezel Üsten, Anna Sieben, Mohcine Chraibi, Armin Seyfried

Published 2026-04-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ezel Üsten, Anna Sieben, Mohcine Chraibi, Armin Seyfried

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a crowd of people waiting to get into a concert. In most computer simulations of how crowds move, scientists treat everyone's "drive" to get in as a simple, unchangeable number. It's like saying, "Everyone here wants to get in at a speed of 5 miles per hour," and that's it. The computer doesn't know that if you're at the back of the line, you might feel discouraged and slow down, or if you see a gap open up, you might suddenly sprint.

This paper proposes a new way to think about that drive, which the authors call motivation. Instead of a fixed speed limit, they treat motivation like a dynamic mood ring that changes every second based on two things:

How close you are to the goal (The "Expectancy" of getting in).
How much you actually want to get in (The "Value" of the goal).

Here is the breakdown of their idea using everyday analogies:

1. The Old Way vs. The New Way

The Old Way (Static): Imagine a robot army. Every robot is programmed with the same "urgency" level. If the robot is told to be "urgent," it runs fast. If it's told to be "chill," it walks slowly. It doesn't matter if the robot is right at the door or stuck in the back; its speed setting never changes.
The New Way (Dynamic EVP Model): Imagine a real human waiting in line.
- The "Expectancy" (The Odds): If you are at the very front, you think, "I'm definitely getting in!" Your motivation spikes. If you are at the back and the line isn't moving, you think, "I'll never get a good spot," so your motivation drops.
- The "Value" (The Desire): Some people really want to be near the stage (high value). Others just want to get in eventually (low value).
- The Result: The model calculates a "motivation score" every second. This score doesn't just change how fast you walk; it changes how you behave. A highly motivated person might squeeze into smaller gaps, push a little harder, and change direction more aggressively to overtake others. A less motivated person might wait patiently and keep a larger personal bubble.

2. The "Payoff" Game

The authors introduce a concept called the Payoff Structure. Think of this like a game of musical chairs.

If the rule is "Only the first 10 people get a front-row seat," then being in position #11 is a "loss." Your motivation to rush drops because the "prize" is gone.
If the rule is "Everyone gets in eventually," then being in position #50 isn't a loss; you just wait your turn.
The model uses this logic to decide if a person should fight for a spot or just wait.

3. What Happened in the Simulation?

The researchers tested this new model in a computer simulation of people waiting for a closed door to open (like a concert venue). They compared their new "Mood Ring" model against the old "Fixed Speed" model.

The Old Model Result: Everyone looked the same. The crowd was a uniform blob. People at the back and people at the front had the same spacing and behavior.
The New Model Result: The crowd organized itself naturally.
- People who were close to the door and highly motivated squeezed into tight spaces, creating a dense cluster right at the entrance.
- People further back, or those with lower motivation, spread out more and kept larger gaps between them.
- The Key Finding: The simulation showed a "structured mess." It wasn't just a faster crowd; it was a different kind of crowd. The people at the front occupied less space than the people at the back, creating a natural gradient that matched real-world experiments.

4. Why This Matters

The paper argues that motivation isn't just about "running faster." It's about reorganizing the crowd.

When people are highly motivated, they don't just all sprint; they rearrange themselves competitively. They fight for the best spots, squeeze into tight gaps, and create a specific pattern of density.
The old models missed this because they only looked at speed. The new model captures the strategy people use: "I am close, so I will squeeze in," vs. "I am far away, so I will wait."

Summary

Think of the crowd not as a fluid of water, but as a group of individual players in a game.

Old Model: Everyone plays with the same rules and energy level.
New Model: Everyone plays based on their current position in the game and how much they want to win. This creates a realistic, shifting pattern where the front of the line is tight and aggressive, while the back is loose and patient.

The authors conclude that to understand crowds, we need to stop treating motivation as a simple "speed dial" and start treating it as a complex, changing strategy that reshapes how people stand, move, and interact with each other.

1. Problem Statement

In traditional pedestrian dynamics modeling, the internal drive propelling individuals toward a goal is typically represented by a static parameter: the desired walking speed ( $v_0$ ). This simplification assumes that motivation is either uniform across a crowd or fixed at the individual level throughout a simulation.

The authors argue this approach is insufficient because:

Motivation is dynamic: It fluctuates based on spatial proximity to the goal, relative position among peers, and the perceived likelihood of success.
Motivation is multi-dimensional: It influences not just speed, but also gap-closing behavior, interpersonal spacing, and directional changes.
Static models fail to capture heterogeneity: Real-world crowds exhibit structured heterogeneity (e.g., agents self-organizing into differentiated positions near bottlenecks) that static models cannot reproduce, often resulting in unrealistic uniform compression.

2. Methodology: The Extended Value-Payoff (EVP) Model

The paper proposes a dynamic motivation model grounded in Expectancy-Value Theory from psychology. The model calculates an agent's motivation ( $m_i$ ) as a multiplicative function of Value and Expectancy, where Expectancy is the sum of Spatial and Payoff components.

A. Mathematical Formulation

The motivation $m_i$ for agent $i$ is defined as:
$m_i = V_i \cdot E_i$
Where:
$E_i = SE_i + P_i$

Value ( $V_i$ ): Represents the intrinsic importance of the goal to the individual.
- Assigned at initialization as a fixed value drawn from a uniform distribution $[v_{min}, v_{max}]$ .
- Scaled to ensure bounded heterogeneity.
Spatial Expectancy ( $SE_i$ ): Represents the likelihood of success based on distance to the goal.
- Modeled as a function of distance $d$ to the exit.
- Increases as $d \to 0$ and plateaus at a baseline $\epsilon$ for distant agents.
Payoff ( $P_i$ ): Represents the collective context and relative standing.
- Based on the agent's normalized rank ( $q_i$ ) within the crowd (1 = closest to the door).
- Modeled using a logistic decay function where agents closer to the front receive higher payoff, reflecting a "winning" probability in a competitive resource scenario (the bottleneck).

B. Motivation-to-Motion Mapping

The calculated motivation $m_i$ does not merely set a speed; it modulates five operational parameters of the underlying Collision Speed Model (CSM) via logistic response functions:

Desired Speed ( $\tilde{v}_0$ ): Increases with motivation ( $\tilde{v}_0 = v_0 \cdot m_i$ ).
Time Gap ( $\tilde{T}$ ): Decreases with motivation ( $\tilde{T} = T / m_i$ ), causing agents to close gaps faster.
Interpersonal Buffer ( $\tilde{b}$ ): Decreases with motivation, allowing agents to stand closer to neighbors.
Neighbor Repulsion Strength ( $\tilde{A}$ ): Increases with motivation, making agents more sensitive to neighbors and more likely to change direction to overtake.
Turning Bound: Modulated to allow more aggressive directional changes.

3. Experimental Setup & Evaluation

Simulation Environment: The model was implemented in JuPedSim using a closed-door bottleneck scenario (agents waiting behind a closed door).
Population Sizes: $N=40$ and $N=80$ agents.
Comparison Baseline: A Uniform-Motivation Baseline where $m_i \equiv 1$ for all agents (static approach).
Empirical Reference: Trajectory data from the CROMA bottleneck experiments (Düsseldorf), which included low-motivation (reserved seats) and high-motivation (no reserved seats) conditions.
Statistical Analysis: Paired statistical comparisons (Wilcoxon signed-rank test) across 10 random seeds to ensure robustness.

4. Key Results

The study demonstrates that the EVP model produces structured heterogeneity absent in static models:

Rank-Area Correlation:
- Static Baseline: Shows a flat relationship between an agent's final rank and their occupied Voronoi area (homogeneous distribution).
- EVP Model: Shows a strong positive correlation ( $\rho \approx 0.66$ for $N=80$ ). Agents closer to the front (higher rank) occupy significantly less space (higher compression), while those further back occupy more space. This matches the "ordering structure" observed in real experimental data.
Spatial Motivation Heatmaps:
- Motivation is not uniform; it is spatially concentrated near the bottleneck and the advancing front, consistent with the goal-proximity hypothesis.
Trajectory Patterns:
- High Motivation (hM): Agents converge tightly toward the bottleneck, forming a dense cluster.
- Low Motivation (nM): Agents exhibit wider lateral spread and "waiting knots" (stopping and starting), mirroring experimental observations.
Ablation Studies:
- Removing the Payoff term eliminates the rank-dependent structure.
- Removing the Value term eliminates the difference in overall compression levels between low/high motivation scenarios.
- Removing Spatial Expectancy removes the distance-based gradient.

5. Key Contributions

Psychological Grounding: Successfully adapts Expectancy-Value Theory (typically used for long-term academic/career goals) to short-timescale pedestrian dynamics (seconds).
Multi-Parameter Modulation: Moves beyond the "speed-only" paradigm by demonstrating that motivation simultaneously alters speed, spacing, gap-closing, and interaction ranges.
Collective Context Integration: Introduces a Payoff Structure that explicitly models the competitive aspect of crowd movement (relative rank), bridging the gap between individual psychology and collective emergent behavior.
Empirical Validation: The model reproduces the specific "structured heterogeneity" (rank-dependent density) found in real-world bottleneck experiments, which static models fail to capture.

6. Significance and Implications

Redefining Urgency: The paper argues that crowd motivation should not be understood merely as a "global increase in urgency" (everyone running faster). Instead, it is a mechanism that reorganizes competitive positioning along spatial and social axes.
Safety and Design: Understanding that motivation creates structured density gradients (rather than uniform compression) is crucial for designing safer evacuation routes and managing crowd flow in venues like stadiums and transit hubs.
Future Directions: The framework is designed to be extensible to open-door throughput scenarios, larger populations, and more complex social interactions (e.g., group cohesion, cooperative strategies).

In conclusion, the EVP model provides a mathematically rigorous and psychologically sound framework that significantly improves the realism of pedestrian simulations by capturing the dynamic, heterogeneous, and context-dependent nature of human motivation in crowds.