针对现有基于深度学习的自适应调制编码算法在信道环境改变时出现的模型泛化能力下降的问题，提出了一种基于元学习的自适应调制编码策略。该方法利用元学习算法快速适应新任务的优势，使得模型仅需通过新环境下的少量样本微调就能获得良好的性能。仿真结果和讨论都表明，本文所提算法比基线算法在性能上表现更为优越。

A meta-learning-driven adaptive modulation coding strategy was proposed for dealing with the model generalization capability degradation problem encountered by the existing deep learning algorithms when the channel environment was changed. The proposed approach employed the Model-Agnostic Meta-Learning (MAML) algorithm to predict the modulation and coding schemes based on channel characteristics. Initially, two neural network models were proposed and trained. Subsequently, a small number of samples from new channel scenarios were used to fine-tune the trained model parameters, thus enabling rapid adaptation to new environments. Both simulation results and discussions demonstrate that the proposed algorithm outperforms the baseline algorithms in term of throughput performance and model generalization capability.