Source-linked AI summary
Generalized Centrifugal Force Model for Pedestrian Dynamics
Mohcine Chraibi, Armin Seyfried, Andreas Schadschneider
TL;DR
The paper addresses the need for quantitatively reliable pedestrian models across geometries without relying on geometry-specific parameter sets. It modifies a spatially continuous force-based model by introducing elliptical volume exclusion and evaluates it in pedestrian-flow settings, obtaining good agreement with controlled experimental data.
Problem
The paper seeks a robust, quantitatively verified, and simple pedestrian model that can describe movement across narrow and wide corridors with one parameter set.
Method
The paper modifies a spatially continuous force-based model by replacing circular pedestrian exclusion with elliptical volume exclusion and simplifying collision handling.
Results
The proposed model shows good agreement with experimental data, including the one-dimensional fundamental diagram and velocity-dependent elliptical semi-axes that reproduce its shape.
Takeaways & Limitations
Elliptical, velocity-dependent pedestrian representations improve the model’s quantitative description of movement in single-file and corridor settings.
Abstract
from arXiv · showhide
A spatially continuous force-based model for simulating pedestrian dynamics is introduced which includes an elliptical volume exclusion of pedestrians. We discuss the phenomena of oscillations and overlapping which occur for certain choices of the forces. The main intention of this work is the quantitative description of pedestrian movement in several geometries. Measurements of the fundamental diagram in narrow and wide corridors are performed. The results of the proposed model show good agreement with empirical data obtained in controlled experiments.
I. INTRODUCTION
The paper targets robust, quantitatively verified pedestrian models by proposing a simple force-based approach intended to work across narrow and wide corridors with one parameter set.
- Modeling context: Microscopic pedestrian models represent individual people and include rule-based and force-based subclasses.The paper focuses on spatially continuous force-based models.
- Model formulation: Each pedestrian’s motion is obtained by numerically integrating a high-dimensional system of second-order ordinary differential equations.Repulsive and driving forces determine the positions and velocities of all pedestrians.
- Modeling limitations: Existing force-based models reproduce collective phenomena qualitatively but can produce overlapping pedestrians, negative velocities, and high velocities in dense situations.Such behavior may require replacing the nominal equation of motion with additional procedures.
- Quantitative gap: Quantitative force-based studies often depend on geometry because their calibrated parameter sets change across scenarios.Prior investigations were commonly restricted to particular settings such as one-dimensional motion, bottlenecks, two-dimensional motion, or room outflow.
- Aim: The proposed work seeks a simple model that describes pedestrian movement quantitatively in both narrow and wide corridors using a unique parameter set.The stated scope is restricted to corridors.
- Contribution: The model is based solely on the equation of motion while retaining free parameters that can be calibrated to quantitative data.This combines a simple modeling structure with parameter fitting.
II. THE CENTRIFUGAL FORCE MODEL
The Centrifugal Force Model represents pedestrian motion through superposed driving and repulsive forces, with interactions shaped by distance, relative velocity, visibility, and viewing angle. Its collision-management supplement motivates a systematic modification of the repulsive force.
- Force-based formulation: The CFM models pedestrians as circular disks whose motion follows the superposition of driving and repulsive forces.Repulsive forces act between pedestrians and against obstacles, while the driving force specifies direction and desired speed.
- Driving force: The driving force represents a pedestrian’s intention to reach a destination at a desired speed, with a time constant τ.The supplied passages identify τ but do not provide the full displayed expression.
- Pedestrian interactions: The CFM repulsive force decreases with distance and incorporates relative velocity, while preventing slower pedestrians from responding to faster pedestrians ahead.The relative-velocity term is restricted to positive approaching motion.
- Perception: The reaction field is limited to pedestrians and obstacles within a 180° angle of vision, making repulsive-force strength depend on viewing angle.The angular coefficient is maximal for pedestrians ahead and minimal beyond 90°.
- Collision handling: The original CFM uses a Collision Detection Technique to manage conflicts and mitigate overlaps, but this adds complexity beyond the initial force-based equation.The paper therefore modifies the repulsive force to improve quantitative description while retaining the model’s central force-based idea.
III. OVERLAPPING VS. OSCILLATION
The model addresses a trade-off between pedestrian overlapping and trajectory oscillations by modifying the repulsive force and removing the collision detection technique. Simulations quantify both phenomena to calibrate the interaction strength η, reducing overlap while managing instability.
- Motivation: Pedestrian volume exclusion is treated as velocity-dependent because geometric overlap can displace a modeled pedestrian’s center of mass.The model allows limited overlap to represent elastic deformation, but identifies overlap as a serious issue requiring control.
- Model modification: The model simplifies the CFM by dispensing with the collision detection technique and modifying the repulsive force to compensate for its absence.The changes include transforming the force singularity and introducing intended speed in the numerator, with η adjusting force strength.
- Measures: The overlapping-proportion o(v) measures normalized overlap between pedestrian pairs, while the oscillation-proportion o(s) measures normalized oscillatory behavior during simulations.The overlap calculation uses pairwise overlapping areas, and both proportions are set to zero when their corresponding event count is zero.
- Trade-off: Increasing repulsive-force strength reduces overlapping but increases oscillations, creating a trade-off between exclusion and trajectory stability.Reducing force strength to avoid oscillations instead leads to pedestrian or obstacle overlap.
- Simulation results: In evacuation simulations, increasing η makes overlap negligible but raises o(s), so η must be calibrated to balance both phenomena.The authors propose the intersection of the o(s) and o(v) curves as a criterion for selecting an optimal η.
IV. HARD CIRCLES VS. DYNAMICAL CIRCLES: THE FUNDAMENTAL DIAGRAM FOR SINGLE FILE MOVEMENT
The section describes velocity-dependent pedestrian space requirements and the corresponding repulsive-force formulation for single-file movement.
- The effective pedestrian radius is modeled with a linear velocity dependence using rmin and τr.
- Pedestrian space requirement includes the torso, leg motion, lateral swaying, and a safety margin.
- The repulsive force uses the effective distance between pedestrians and the radius ri defined by the velocity-dependent formulation.
V. ELLIPTICAL VOLUME EXCLUSION OF PEDESTRIANS
The model replaces rotationally symmetric circles with velocity-dependent ellipses to represent pedestrians’ directional space requirements and lateral swaying in two-dimensional movement.
- Circular pedestrians occupy unnecessary lateral space in two-dimensional movement because circles are rotationally symmetric.
- Ellipses better approximate the projected space required by the body, leg motion, lateral swaying, and safety margin.
- The model defines each pedestrian as an ellipse centered at (xi,yi), with major semi-axis a along movement and minor semi-axis b orthogonal to it.
- The major semi-axis a is velocity-dependent through parameters amin and τa.
- Lateral swaying is represented by b, whose amplitude decreases from bmax during slow movement toward bmin at free velocity.
- Because a and b depend on velocity, the movement-direction axis is denoted a and the orthogonal axis b without assuming a fixed ordering.
VI. ELLIPTICAL VOLUME EXCLUSION AND FORCE IMPLEMENTATION
The section motivates mathematical treatment of the elliptical repulsive-force implementation using experimentally observed trajectory swaying as context.
- The implementation section provides mathematical insights into calculating the repulsive forces.
- Detected pedestrian trajectories can show strong swaying, while faster movement is associated with smoother and weaker swaying trajectories.
A. Repulsive Forces between Pedestrians
The model computes pedestrian and wall interactions using ellipse-to-ellipse and ellipse-to-line distances, while representing each wall with three moving interaction points.
- A. Repulsive Forces between Pedestrians: The pedestrian–pedestrian repulsive force requires the distance between ellipse borders along the line connecting their centers, distij.
- A. Repulsive Forces between Pedestrians: For two non-overlapping ellipses, the closest-approach distance l̃ can be non-zero and depends on their orientations.
- B. Repulsive Forces between Pedestrians and Walls: Each wall is modeled by three static point masses acting on pedestrians within an interaction range.
- B. Repulsive Forces between Pedestrians and Walls: The three wall points are recomputed as the pedestrian moves, with the middle point nearest the pedestrian’s center-to-wall segment.
- B. Repulsive Forces between Pedestrians and Walls: Three point masses were selected by trial and error because simulations found them sufficient for wall avoidance and computationally cost-effective.
- B. Repulsive Forces between Pedestrians and Walls: The ellipse-to-line distance uses the polar radius ri and the distance ki from the ellipse center point oi to the wall line, with a closest-approach distance k̃ also defined.
- B. Repulsive Forces between Pedestrians and Walls: For motion parallel to a wall, the normal velocity component is zero, so the lateral wall points have no effect in the force expression.
C. Numerical Stabilization of the Repulsive Force
The model limits repulsive interactions to nearby pedestrians and uses Hermite interpolation to stabilize the force near contact. The superposed forces do not guarantee the minimum effective-distance constraint.
- Pedestrian-pedestrian repulsion is truncated at rc = 2 m, restricting interactions to adjacent pedestrians.The original force has infinite range, which is treated as unrealistic for pedestrian interactions.
- A two-sided Hermite interpolation smooths the repulsive force for robust numerical integration.The interpolation defines left and right transition functions around the contact-distance region.
- The effective distance uses distij and the distance of closest approach ˜l, with s0 defining the minimum allowed magnitude.The construction is illustrated for constant relative velocity.
- As distij approaches the closest-approach distance from above, the interpolation prevents the force from diverging and caps it at fm.The force remains constant below the specified effective-distance threshold.
- Because forces are superposed, the inequality enforcing the minimum effective distance is not guaranteed for pedestrians i and j.
VII. SIMULATION RESULTS
Simulations evaluate the model in one-dimensional movement and a two-dimensional corridor using specified measurement setups and calibrated parameters. The resulting velocity-density relations agree well with experimental data, while the model remains limited in scope and requires tuning.
- Changing τa changes the slope of the one-dimensional velocity-density relation, while τa = 0 represents constant space requirement.The simulations keep amin = 0.18 m in this comparison.
- The one-dimensional fundamental diagram compares well with experimental data when velocity-dependent elliptical semi-axes are used.This velocity dependence also reproduces the shape of the fundamental diagram.
- The two-dimensional evaluation uses a 25 m × 1 m periodic corridor and calibrated lateral semi-axes bmin = 0.2 m and bmax = 0.25 m.A 2 m × 1 m measurement segment is placed in the corridor center.
- With the selected semi-axis dimensions, the model yields the correct velocity-density relation in single-file movement and wide corridors, although only a 1 m corridor width was investigated.
- Elliptical particles produce an upper-bound fundamental diagram relative to circular particles at low and medium densities, with no noticeable difference at high densities.
- The model describes operative behavior quantitatively but does not reproduce tactical or strategic phenomena such as cooperation, lane changing, or overtaking in bidirectional flow.The conclusions also state that free parameters must be tuned for a given scenario.
Appendix A: Distance between two ellipses
The appendix defines ellipse-to-ellipse and ellipse-to-line distances geometrically using borders along center-connecting lines. It also notes that these distance measures can remain nonzero when ellipses touch or overlap.
- distij is defined as the distance between ellipse borders along the line connecting their centers.
- The radius of an ellipse in a chosen polar direction is derived after expressing the ellipse in quadratic form and polar coordinates.The same procedure determines the corresponding radius for ellipse j.
- The distance between two ellipses can remain nonzero even when the ellipses touch or overlap.
- The distance of closest approach is the smallest border distance along the center-connecting line while the ellipses are not overlapping.For ellipses, this distance can be nonzero and depends on orientation.
- For an ellipse and a line, the closest-approach distance is computed using a parallel tangent line and the ellipse-line intersection equations.The final distance combines the tangent-line offset with the ellipse's polar radius.
Appendix C: Measurement method
The measurement method tracks pedestrians as they enter and leave a defined measurement area, using their transit times to compute mean velocity and occupancy to define density.
- A pedestrian's mean velocity is determined from the positions at which they enter and leave the measurement area.
- The velocity calculation uses the pedestrian's entrance and exit times.
- For one-dimensional movement, the transverse coordinate at entry is set to yin_i = 0.
- Density is defined from the number of pedestrians within the measurement area and its movement-direction length lm = 2 m.In one-dimensional space, the measurement area is reduced to a segment of length lm.