With the advancement of deep learning technologies, significant progress has been made in deep learning-based robot pose control methods. However, these methods rely heavily on large-scale labeled datasets for model training, and the acquisition of labeled data is time-consuming and labor-intensive, significantly limiting the application of deep learning in pose control. Recent developments in generative models have made dataset augmentation using generative algorithms an effective approach to address this challenge. This paper proposes a method that leverages Generative Adversarial Networks (GANs) to learn the distribution of pose data and generate new robot pose samples. Analysis using the Maximum Mean Discrepancy (MMD) and 1-Nearest Neighbor (1-NN) metrics reveals that the distribution of generated data closely resembles that of real data, indicating that the GAN-generated samples are of high quality and can effectively mitigate the small-sample problem in training pose control models.