Abstract
To tackle the challenge of personnel evacuation following a mine fire, we propose an enhanced SSA (Salp Swarm Algorithm) simulation algorithm inspired by the intelligent behavior of marine organisms for personnel evacuation path planning. Our approach addresses the issues of local optimal solution trapping and slow convergence in the standard SSA algorithm by incorporating chaos optimization, differential evolution mutation, and non-uniform Gaussian mutation. We evaluate the performance of our improved SSA algorithm in path optimization through comparative testing with the standard SSA algorithm, particle swarm optimization (PSO) algorithm, and genetic algorithm (GA). Subsequently, we conduct grid-based path planning simulation experiments. Our findings demonstrate that our enhanced SSA algorithm excels in path optimization, achieving a 6% shorter optimal length compared to GA, an 11% improvement over PSO, and a 9% enhancement over the ant colony algorithm. Furthermore, in terms of path stability, our improved SSA algorithm outperforms GA by 75%, PSO by 89%, and the standard SSA algorithm by 83%. These results establish that our enhanced SSA algorithm surpasses other algorithms in obtaining the shortest and optimal paths.
Introduction
In the high-stakes environment of a mine fire, rapid and reliable evacuation is paramount. Traditional path-planning algorithms, such as Dijkstra’s or Floyd’s, often encounter "computational explosion" as environmental complexity increases, leading to delayed response times. While bio-inspired "intelligent biota" algorithms offer faster computation, they frequently suffer from trapping in local optima—finding a path that appears optimal locally but is suboptimal or dangerous on a global scale. This research introduces the Improved Salp Swarm Algorithm (ISSA) to overcome these limitations, ensuring fast convergence and high stability for emergency path planning.
Methods: The Improved Salp Swarm Algorithm (ISSA)
The standard Salp Swarm Algorithm (SSA) mimics the swarming behavior of salps—marine invertebrates that form chains to navigate and forage. To enhance its performance for mine fire scenarios, the authors introduced three critical modifications:
- Logistic Chaos Initialization: Replaces random starting positions with chaotic mapping to ensure uniform coverage of the search space, preventing the algorithm from overlooking the global optimum.
- Differential Evolution Mutation (DEM): Perturbs the leader salp's position to force exploration of new areas if the algorithm stalls, effectively escaping local optima.
- Non-uniform Gaussian Variation: Targets the "tail" of the salp chain with adaptive step sizes, maintaining population diversity and refining the search as it converges.
The algorithm's logic is structured to iteratively update positions and evaluate fitness until a global optimum is achieved, as shown in the flowchart below.
Experimental Validation and Results
The ISSA was rigorously tested using two methods: mathematical benchmark functions (Sphere, Rastrigin, Griewank, and Schaffer) and a simulated 50x50 grid map of a mine roadway.
Benchmark Performance
In mathematical tests, ISSA consistently reached the optimal value (zero error) or significantly outperformed competitors like Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The stability of ISSA—measured by the consistency of results across multiple runs—was particularly noteworthy, showing near-zero standard deviation in most tests.
Path Planning Efficiency
When applied to the simulated mine environment, ISSA demonstrated superior efficiency and stability:
- Path Length: ISSA produced paths 6% shorter than GA, 9% shorter than standard SSA, and 11% shorter than PSO.
- Stability: ISSA improved stability by 75% over GA, 83% over standard SSA, and 89% over PSO.
- Path Quality: The generated paths were smoother with fewer sharp turns, which is critical for real-world navigation in smoke-filled environments.
Discussion and Conclusion
The ISSA provides a significant advancement in emergency evacuation technology. By focusing on both the "leader" and the "tail" of the swarm, the algorithm maintains the diversity needed to find global optima while ensuring the stability required for life-safety applications.
While the current model uses a 2D grid with static obstacles, future research will focus on adapting the algorithm for complex 3D mine structures and integrating real-time sensor data to account for dynamic fire and smoke spread. Despite a slight increase in computational overhead due to mutation strategies, the substantial gains in path efficiency and reliability make ISSA a superior choice for mine fire emergency planning.
Summary updated