Source-linked AI summary
Modelling dynamic route choice of pedestrians to assess the criticality of building evacuation
A. U. Kemloh Wagoum, A. Seyfried, S. Holl
TL;DR
Evacuation route choice is difficult to model because it depends on subjective factors such as facility familiarity and affects evacuation and jam times. The paper models event-driven routing with four shortest- and quickest-path strategies, then compares their evacuation and jam outcomes, which vary with congestion.
Problem
Route choice depends on subjective influences such as facility familiarity, yet it affects overall evacuation time and time spent in jams.
Source: §1 · p. 1Method
Results
Quickest-path routing was associated with evacuation averaging 1 minute faster and jam time averaging 30 seconds, while in the highly congested full-facility test mean jam sizes were 46 m2 for static and 40 m2 for dynamic strategies.
Source: §4 · p. 10Takeaways & Limitations
The approach assesses buildings across possible route choices, and the authors report that quickest-path routing is not sensitive to initial distribution or special topologies such as symmetric exits.
Source: §5 · p. 14
Abstract
from arXiv · showhide
This paper presents an event-driven way finding algorithm for pedestrians in an evacuation scenario, which operates on a graph-based structure. The motivation of each pedestrian is to leave the facility. The events used to redirect pedestrians include the identification of a jam situation and/or identification of a better route than the current. This study considers two types of pedestrians: familiar and unfamiliar with the facility. Four strategies are modelled to cover those groups. The modelled strategies are the shortest path (local and global); They are combined with a quickest path approach, which is based on an observation principle. In the quickest path approach, pedestrians take their decisions based on the observed environment and are routed dynamically in the network using an appropriate cost benefit analysis function. The dynamic modelling of route choice with different strategies and types of pedestrians considers the manifold of in uences which appears in the real system and raises questions about the criticality of an evacuation process. To address this question criteria are elaborated. The criteria we focus on in this contribution are the evacuation time, the individual times spent in jam, the jam size evolution and the overall jam size itself. The in uences of the different strategies on those evaluation criteria are investigated. The sensibility of the system to disturbances (e.g. broken escape route) is also analysed. Keywords: pedestrian dynamics, routing, quickest path, evacuation, jam, critical state
1. Introduction
The introduction frames pedestrian route choice as a difficult modeling problem because subjective influences, including facility familiarity, shape evacuation and jam times. It introduces four shortest/quickest-path strategies and additional criteria for assessing evacuation criticality.
1. Introduction
- Route choice is difficult to model because subjective influences affect pedestrians’ choices and therefore overall evacuation time and time spent in jams.The study restricts its scope to pedestrians who share the motivation to leave the facility.Source: §1 · p. 1
- The introduction positions pedestrian simulation as relevant to escape-route design while noting that existing approaches are only a first step toward modeling evacuation influences.It situates the work amid growing attention to pedestrian dynamics and challenges for facility and event planning.Source: §1 · p. 1
- Graph-based routing represents routes as a network of destination points, with shortest paths commonly computed using Dijkstra or Floyd–Warshall algorithms.The introduction contrasts this with quickest-path approaches that seek to avoid congestion using dynamic fields, observed density, exit loads, or travel-time optimization.Source: §1 · p. 2
- The paper introduces event-driven graph-based way finding with four strategies: local and global shortest paths, each combined with quickest-path routing.Pedestrians observe their environment and use the observations to make routing decisions.Source: §1 · p. 2
- The study assesses criticality through individual time spent in jams, jam-size evolution, and total jam size, defined as the area under the jam-size curve.It investigates these criteria for familiar and unfamiliar pedestrians using the four routing strategies.Source: §1 · p. 2
2. Modelling
The model combines graph-based routing with pedestrians’ changing route choices, distinguishing familiar and unfamiliar pedestrians. Its quickest-path strategy uses observations of queues and visible alternatives to assess rerouting.
2. Modelling
- The model focuses on strategic route choice, assigning familiar and unfamiliar pedestrians a next destination while their routing strategies may change during evacuation.Walking is represented at the operational level by the Generalized Centrifugal Force Model.Source: §2 · pp. 2–3
- The facility is represented as decision areas connected by nodes, with a visibility-based network and Euclidean distances used as edge weights.Decision areas abstract rooms and corridors; destinations are exits and corridor ends.Source: §2 · p. 3
2.1. Shortest Path
- Familiar pedestrians may choose the global shortest path to the outside, while unfamiliar pedestrians choose the nearest node at each decision point.The global paths are determined on the graph using Floyd–Warshall with path reconstruction.Source: §2.1 · p. 3
2.2. Quickest Path
- Quickest-path rerouting begins when a pedestrian enters a new room or identifies a jam, after an initial local or global shortest-path assignment.Jam-triggered searches use a pedestrian-specific patience time, initially 10 seconds and increased by 1 second after each unsuccessful attempt in the same room.Source: §2.2 · pp. 3, 6
- The approach compares gains for the current destination and alternatives, and reroutes only when the cost-benefit result exceeds a pedestrian-type threshold.The thresholds are 0.20 for familiar pedestrians and 0.15 for unfamiliar pedestrians; the approach is not tied to facility geometry or initial pedestrian distribution.Source: §2.2 · pp. 6–7
- For visible alternatives, pedestrians select reference pedestrians from queues using visibility and distance, then observe them to estimate expected travel time.A reference is the nearest visible pedestrian in the queue; the model also defines visibility using the current location and other pedestrians.Source: §2.2 · pp. 3, 4, 6
3. Evacuation assessment
Evacuation assessment considers evacuation time, individual time spent in jams, and jam size and its evolution. The quickest-path approach reduces overall evacuation time and narrows the evacuation-time and jamming-time distributions.
3.1. Evacuation time
- The quickest path reduces overall evacuation time and distribution width, while also producing more realistic simulation shapes.The evacuation-time distribution has its smallest variance under GSP and largest under LSP.Source: §3.1 · pp. 7–8
3.2. Jamming time
- Pedestrians moving below 0.2 m/s for at least 10 seconds count as jammed, and the quickest path yields a narrower distribution of time spent in jams.Each pedestrian's total time in jam is recorded to form the distribution.Source: §3.2 · p. 8, Figure 5 · p. 8
3.3. Jam size
- Jam size is measured as the summed effective areas of jammed pedestrians and requires separate analysis from evacuation time and individual jam duration.Short-lived or small jams are described as relatively uncritical; large, long-lived jams warrant consideration against evacuee and environmental conditions.Source: §3.3 · p. 8
- Quickest-path jam-size evolution peaks at approximately 150 m² at all exits before declining, with a less constant slope as pedestrians change destinations.The figure depicts 2,500 pedestrians in a 50 m × 50 m room with four exits on each wall side.Source: §3.3 · p. 9, Figure 6 · p. 9
4. Simulation and Analysis
The simulations compare route-choice strategies in a 2144 m² facility across a 250-pedestrian block, a 1,000-pedestrian full-facility evacuation, and a broken-exit scenario. Quickest-path routing improves outcomes in the first and disturbed scenarios, while heavy congestion limits differences in the full-facility case.
Simulation setup
- The test structure is a simplified section of Düsseldorf’s ESPRIT arena, with 250- and 1,000-person initial distributions across a 2144 m² simulation area.The simulations also examine evacuation dynamics at 60 seconds for the two initial distributions.Source: §4 · pp. 9–10, Figure 7 · p. 10, Figure 8 · p. 11
Single-block evacuation
- For 250 pedestrians, quickest-path routing cuts average evacuation time by 1 minute and average time in jam to 30 seconds, versus nearly three times as long with static strategies.The facility’s static strategies produce five long-lasting jams; quickest-path strategies produce more but shorter-lived jams, with mean jam sizes of 11 m² versus 9 m².Source: §4 · p. 10, Figure 9 · p. 11, Figure 10 · p. 12
Full-facility evacuation
- In the 1,000-pedestrian full-facility test, heavy congestion leaves little difference in mean evacuation and jamming times; mean jam size is 46 m² for static and 40 m² for dynamic strategies.The authors report that most pedestrians have to follow the flow, with no distinct quickest path in this scenario.Source: §4 · p. 10, Figure 12 · p. 13
Disturbed system
- With exits E2 and E8 broken, quickest-path routing produces faster evacuation and less time in jam, and the model reproduces the resulting route asymmetry.All mean values are lower than in the corresponding case without disturbance.Source: §4 · p. 10, Figure 13 · p. 14
5. Summary
The paper combines shortest- and quickest-path strategies in a graph-based way-finding approach to assess evacuation criticality across varied route choices. It reports that observation-based quickest routing is robust to initial distribution and symmetric-exit topology, and evaluates multiple evacuation and jam criteria across different scenarios.
Summary
- Four graph-based strategies combine local or global shortest paths with quickest-path routing to assess buildings across possible route choices.The quickest-path approach uses observation and is reported as insensitive to initial pedestrian distribution and symmetric-exit topologies.Source: §5 · p. 14
- Observation-based quickest routing is reported to be general across initial distributions and special topologies, including symmetric exits, and to produce more realistic evacuation dynamics.Source: §5 · p. 14
- The assessment examines evacuation-time distributions, individual time spent in jams, and average jam size across symmetric, asymmetric, and broken-route scenarios.The distributions, their meanings, and their impacts are quantitatively analysed.Source: §5 · p. 14