Test 16: Affiliation to familiar exits

1. Definition

A room of 10 m by 15 m has two 1 m exits on the 15 m walls, at equal distance from the short end wall. In Scenario 1 one occupant who is unfamiliar with the exits (no pre-evacuation time, 1 m/s) starts 1 m from the short wall at a fixed position equidistant from both exits. In Scenario 2 one occupant affiliated with Exit 2 starts at a position that ISO’s text places in the central area at the start of the corridor. Each scenario is repeated until the exit usage changes by less than 1 % with an additional run. The test needs an exit-choice sub-model that represents familiarity. The tester documents whether affiliation has its own sub-model and whether exit choice is deterministic or predictive.

Source: ISO 20414:2020, Test 16, Table 18, pp. 27-28, Fig. 12. This is a paraphrase, not the text of the standard.

Reference: ISO (2020). ISO 20414:2020, Fire safety engineering - Verification and validation protocol for building fire evacuation models. First edition. International Organization for Standardization, Geneva. https://www.iso.org/standard/78348.html (Page numbers are the printed page numbers of ISO 20414:2020 (the copy read is a proof, so numbers can differ by a page in the published edition). ISO states that its tests are examples and that a tester may use different tests that verify the same component equivalently.)

2. Scenario

Scenario figure from the executed notebook.

3. What to expect

The usage of Exit 2 in Scenario 2 is higher than in Scenario 1.

Our criterion: Journey-weighted affiliation biases exit choice (80/20 -> exit 2 preferred)

The outcome is fixed by the input journey weights and the harness’s proportional split of agents over journeys, not by an affiliation model; same mechanism as NIST Verif.3.3.

4. Our result

Status: TEST-LOCAL

Test cases of this row

• test_affiliation_biases_exit_choice: bound: measured <= 0.10 (|exit1 - exit2| / total, 50/50 weights); comparison: affiliated_exit_counts[1] > affiliated_exit_counts[0] with 20/80 weights

Result figure from the executed notebook.

5. Our implementation

Deviations from the standard:

• Affiliation is represented by journey weights. Scenario 1 uses weights of 50 / 50. The affiliated case uses 20 / 80 in favour of Exit 2. The result follows from these input weights and from the harness’s proportional split of agents over journeys. No affiliation sub-model is involved, as in NIST Verif.3.3.
• Each scenario has two occupants instead of one. Both scenarios use the same start region.
• The pass rule is the suite’s own: with 50 / 50 weights the two exits differ in use by no more than 10 % of all arrivals, and with 20 / 80 weights Exit 2 is used more than Exit 1, over 20 fixed seeds. ISO’s repeat-until-stable rule is not used.

Model: CollisionFreeSpeedModel

• Test function: tests/vv/test_iso.py::test_affiliation_biases_exit_choice
• Scenario file: Iso-16-familiar-exits.zip

6. Our notebook

• Executed notebook (HTML)
• Notebook source (.ipynb) on GitHub