Pedestrian evacuation efficiency is paramount to public safety and sustainable urban resilience. This study utilizes an agent-based model simulating evacuation dynamics in a built environment to assess the impact of route familiarity, interpersonal interactions, and storage layout on evacuation efficiency. The model incorporates an evolutionary game theory framework to capture strategic decision-making, featuring both symmetric and asymmetric interactions among evacuees with varying levels of exit information (complete, partial, or none). Results show that familiarity with exit location is the most decisive element for evacuation, significantly outweighing the influence of crowd interactions, imitation behaviors, group composition, or storage layout. Furthermore, the crowd composition exerts a significant moderating effect, so that asymmetric group structures yield superior evacuation performance compared to symmetric ones. The optimal storage layout for evacuation is contingent upon the availability of exit information. An orderly layout is superior when information is known, whereas a random layout proves more effective in the absence of information by preventing misleading paths. Thus, providing clear information, adaptable spatial designs and consciously constructing a heterogeneous population structure are more critical for evacuation. This work provides actionable insights for architects and safety planners, contributing directly to the development of safer, more sustainable built environments and supporting Sustainable Development Goal (SDG) 11, particularly Target 11.5.