Abstract: Objective Based on the actual demands of emergency evacuation,this paper establishes a multi-objec-tive optimization model which takes the flow of each path as a control variable.Maximum flow,mini-mum cost,and maximum reliability are considered as objectives to integrate the timeliness,economyand security of emergency evacuation.An improved Pareto multiple objective genetic algorithm is pro-posed,to encode the control variables directly.It introduces a fitness function based on the degree ofPareto domination and self-adaption punishment,and designs a selection operator based on tourna-ment and niche technology.The algorithm provides a practical tool to solve the problem with complexconstraints and multiple objectives.Finally,a real world road network is used for simulation and ana-lyses,which validates the effectiveness and applicability of the proposed methodology.