Dynamic Fire-Evacuation Route Optimization in Complex Buildings Using Safe-Corridor-Guided RA-SSA

Dynamic fire evacuation in complex multistory buildings requires route planning across interconnected corridors, stairwells, transfer spaces, and exits while hazards, passage accessibility, and circulation capacity change over time. This study presents a risk-constrained method based on a safe-corridor-guided risk-aware sparrow search algorithm (RA-SSA). A time-varying evacuation network integrates building topology, horizontal and vertical circulation links, prescribed or independently generated fire-risk variables, passage states, and local congestion. In ten paired planar trials, RA-SSA achieved a mean model-estimated route travel time of 54.0 s and cumulative dynamic risk of 2.30, corresponding to a 16.1% lower time than GWO and 53% and 69% lower risk than GWO and DDPG, respectively. Across the three prespecified paired comparisons, Holm-adjusted p values were at most 1.11 × 10−4 and the absolute paired effect sizes |dz| ranged from 2.06 to 5.36. Profiling gave an RA-SSA algorithm-only P95 latency of 142.8 ms and a prototype update-cycle P95 of 168.5 ms, both well below the 5 s update interval. One-at-a-time sensitivity tests showed predictable safety-efficiency tradeoffs, while four multistory tests covered normal dual stairs, reduced stair capacity, dynamic stair failure, and an asymmetric fire-and-load condition. In an independently executed FDS 6.11.1 field, the feasible-update ratio remained 95%, route overlap with the prescribed-field result was 0.78, and medium-to-fine mesh changes remained at or below 3.02% for the reported tenability quantities and 1.61% for route-level metrics. These results support RA-SSA as an algorithmic decision layer that balances risk, route efficiency, feasibility, and stability; they do not constitute calibrated real-fire validation or prediction of actual occupant evacuation time.