Design and experimental validation of path planning for fire rescue robots in long traffic tunnels

China has an increasing number of long traffic tunnels exceeding 10 km, which face severe fire risks owing to enclosed spaces and limited evacuation paths. Fires in such tunnels are characterized by dynamic fire spread and real-time fluctuations in environmental parameters such as temperature and visibility. Traditional path planning algorithms, however, fail to dynamically incorporate environmental factors into global planning or adapt to fire evolution during local planning, leading to suboptimal paths for rescue robots. This study aims to propose an improved hybrid algorithm, called improved ant colony optimization-improved artificial potential field (IACO-IAPF), to achieve efficient and safe path planning for fire rescue robots under dynamic tunnel fire scenarios and to verify its performance through simulation experiments to provide technical support for practical tunnel fire rescue operations.

A two-stage IACO-IAPF algorithm following a framework of global guidance and local correction was designed. First, a 500 m × 30 m scaled tunnel model was constructed using PyroSim, discretized into 1.0 m × 1.0 m × 1.0 m grids, and instrumented with 500 temperature sensors and 500 smoke sensors to collect real-time environmental data, forming a dynamic grid map based on NFPA 72-2025 risk grading standards. For global planning (IACO), an equivalent length heuristic function integrating temperature and visibility factors was proposed to replace geometric distance and an adaptive pheromone evaporation factor update strategy based on the environmental change index was designed to enhance convergence and scenario adaptability. For local planning (IAPF), a backtracking direction-finding strategy with escape force was adopted to resolve the local minimum problem; a dynamic repulsion field with decreasing repulsion at the target point addressed the target inaccessibility issue; and a Gaussian-distributed dynamic virtual fire source repulsion field was established to accommodate fire spread. Four dynamic fire scenarios with different fire spread ranges (20 m × 30 m/30 m × 50 m) and obstacle distributions were set up, and IACO-IAPF was compared with traditional algorithms (ACO, GA, PSO, and ACO-APF) in terms of path length, running time, path safety, and path effectiveness.

Experimental results demonstrated the superior performance of IACO-IAPF over the traditional algorithms. In global planning, IACO generated the shortest path (65.18 m) with the highest safety score (0.94) and only a marginally longer running time (20.71 s) compared with ACO, whereas GA failed to converge. In full path planning across four scenarios, IACO-IAPF reduced path length by 8.2%–15.7% and running time by 11.3%–29.3% compared with ACO-APF. Notably, IACO-IAPF maintained path safety consistently above 0.90 in all scenarios, markedly higher than that of ACO-APF in complex scenarios. The paths planned by IACO-IAPF were smoother with fewer turning points, responded to fire evolution in real time, and achieved targeted avoidance of high-risk areas, with optimal path effectiveness in all tests.

The proposed IACO-IAPF algorithm effectively addresses the limitations of traditional path planning algorithms under dynamic tunnel fire conditions. The improved heuristic function and adaptive pheromone evaporation factor in IACO enhance global planning ability, enabling paths that are more responsive to dynamic fire environments while avoiding local optima. The optimized strategies in IAPF successfully resolve the local minimum and target inaccessibility problems inherent in the traditional artificial potential field method, enabling precise real-time avoidance of fire sources and obstacles. The PyroSim simulation results confirm that IACO-IAPF delivers strong performance in path optimization, operational efficiency, and safety assurance, with strong practicality and scenario adaptability for long traffic tunnel fire rescue operations.