In order to solve the problem of autonomous reconnaissance mission planning of multi UAV when communication is interrupted, a distributed mission planning method based on congestion game is proposed. According to the problem scenario, the reconnaissance payoff function of each UAV and the social welfare function of each sub-region are given. Based on the congestion game, a congestion game model for autonomous reconnaissance of multi-UAV is established. Next, the existence of pure Nash equilibrium is verified. Give the boundary between the cost of anarchy and the cost of stable state of multi-UAV autonomous reconnaissance congestion game. Furthermore, a strategy updating rule with performance guarantee and easy implementation is designed and simulated by Monte Carlo method. Simulation results verify the applicability of the model and the validity of policy updating rules.