Emergency evacuation during fire disasters remains a critical issue, especially in developing countries where evacuation planning and infrastructure may be insufficient. The leading cause of casualties during evacuation is not just the disaster but also the panic it induces. Under extreme stress, evacuees exhibit varying behaviors that are difficult to predict due to factors such as age, physical abilities, past experiences, and cultural differences. Indoor evacuations, particularly in building structures, necessitate careful consideration of multiple parameters, including both physical and psychological behaviors of individuals and groups. This study introduces an empirical multi-agent-based fire evacuation model that integrates Particle Swarm Optimization (PSO) to simulate human behavior under fire evacuation conditions. Each individual is modeled as an autonomous agent within a multi-agent environment, and PSO is utilized to optimize individual agent movement during the evacuation process, including exit selection under dynamic fire conditions. The proposed model accounts for human response to fire characteristics and considers both individual and social behaviors in determining evacuation strategies. This paper explores the impact of human and environmental factors on agent decision-making during evacuation, emphasizing the effect of fire-related variables on movement and strategy. The integration of PSO enhances the model’s ability to optimize evacuation paths and exit choices, ensuring a safer and more efficient evacuation process.