It‘s difficult for fixed-wing Miniature Air Vehicles (MAV) in the unknown environment to avoid collision reactively when facing continuous obstacles such as the wall.To solve the problem, a MAV control method for obstacle avoidance is proposed, which combines multi-power reaching law of sliding mode control with super-twisting disturbance observer.First, the mathematical model of MAV collision avoidance system is established by using the spatial relationship and the MAV kinematics equation.Then, the guidance law of MAV collision avoidance is designed by using the characteristics of fast convergence of multi-power reaching law of sliding mode control.As to the uncertainty caused by sensor errors and system modeling errors, the super-twisting disturbance observer is introduced to compensate for the uncertainty.The convergence of the system is proved.Simulation results show that the proposed control method for collision avoidance enables the MAV to quickly avoid continuous obstacles, and has strong robustness to sensor disturbance.