This paper first introduces the advancement and importance of UAV laser communication technology,and then introduces Deep Neural Network (DNN) to develop wireless end-to-end communication systems,where DNN is applied to:encoding,decoding,modulation and equalization.Accurate channel state parameters (CSI) are necessary for computing DNN,but in laser communication systems,channel information is difficult to obtain in advance or varies greatly with time and environment.This paper will use conditional generation network (GAN) to represent channel.The effect,in which the coded signal of the transmitter and the received signal corresponding to the pilot data is used as condition information.The simulation results of additive white Gaussian noise (AWGN) and Lognormal channel show that the method can achieve high quality transmission without clear CSI.