In the simulation of emergency evacuation of aircraft, the cellular automaton model is mostly used to simulate evacuated people. This model is faster in calculation but regards the evacuees as the same particles, ignoring individual differences, so there is a difference from the actual evacuation situation . In this paper, the agent model is used to establish personal behavior rules, combined with the A* pathfinding algorithm, a valuation function in the A* algorithm suitable for aircraft emergency evacuation is proposed, and a character model is established. Finally, the simulation of the emergency evacuation process of the cabin of a certain type of Boeing 737-700 aircraft was carried out, and compared with the previous experimental results and the authoritative simulation software airEXODUS. The simulation results were close to the real experimental results. Therefore, an agent-based algorithm suitable for passenger cabin emergency escape simulation is proposed. Compared with the cellular automaton model used by traditional algorithms, this algorithm is closer to the real evacuation situation.