Directed rooted forest based direction setting method: A step toward automated dynamic exit signs

Dynamic exit signs are generally considered to outperform static ones in guiding evacuees toward exits in the changeable indoor environment during an emergency. However, a crucial challenge lies in equipping them with a rapid and effective adjustment method. This study proposes a directed rooted forest (DRF)-based planning method that can boost the efficiency of iterative optimization with heuristics from simulation. In this method, the DRF structure is introduced to store direction information to ensure the feasibility of the solution. Two algorithms, Branch Grafting and Leaf Grafting, are developed to enhance the global and local search capabilities. Consequently, the optimizer composed of them converges fast by leveraging network flow data obtained from the simulation in the optimization. Through testing, the method demonstrates a significant reduction in the number of simulation executions and maintains a fast convergence rate when the problem size multiplies. In addition to the engineering application potential offered by the accelerated simulation module, this method holds promise in contributing to the development of a machine learning model for real-time guidance direction setting by rapidly generating optimized evacuation guidance plans in a variety of scenarios for training purposes.