CROWD EVACUATION PLANNING USING CARTESIAN GENETIC PROGRAMMING AND AGENT-BASED CROWD MODELING

This paper proposes a newevolutionary algorithm-based methodology for optimal crowdevacuation planning. In the proposed methodology, a heuristic-based evacuation scheme is firstly introduced. The key idea is to divide the region into a set of sub-regions and use a heuristic rule to dynamically recommend an exit to agents in each sub-region. Then, an evolutionary framework based on the Cartesian Genetic Programming algorithm and an agent-based crowd simulation model is developed to search for the optimal heuristic rule. By considering dynamic environment features to construct the heuristic rule and using multiple scenarios for training, the proposed methodology aims to find generic and efficient heuristic rules that perform well on different scenarios. The proposed methodology is applied to guide people's evacuation behaviors in six different scenarios. The simulation results demonstrate that the heuristic rule offered by the proposed method is effective to reduce the crowd evacuation time on different scenarios.