Robot trajectory tracking aims to quantitatively characterize its operational capabilities and provide data support for robot reliability. However, due to the lack of prior knowledge and sufficient data, the single-dimension monitoring feature cannot satisfy the data gap for robot trajectory evaluation. In order to provide a quantitative description of multi-dimensional trajectory data, a trajectory state characterization model based on multi-attribute data augmentation is proposed in this paper. Firstly, a distributed trajectory state characterization architecture is constructed based on fuzzy membership probability and knowledge mapping. To solve the problem of evaluation under insufficient samples, a Multi-Channel Generative Adversarial Network (Multi-CGAN) is constructed to achieve multi-attribute trajectory sample data augmentation. To validate the proposed model, the evaluation performance of the proposed method in different neural network-based models is analyzed, which shows that the proposed data augmentation method demonstrates improved accuracy.