Energy-Efficient Channel Knowledge Map Construction for 6G Networks Using Improved Transformer

With the rapid development of 6G communication technology, accurate channel modeling has become crucial to optimize wireless networks. Constructing channel knowledge map is one of the important methods for implementing channel modeling. The construction of channel knowledge maps mainly involves generating path loss information. In this paper, we propose an environment encoding method to compress the complex environment features into a compact matrix representation. Furthermore, an improved Transformer-based model is proposed to construct the CKM. Experimental results show the proposed model outperforms baselines, with a mean absolute error (MAE) down to 3.82 dB, superior training efficiency, and potential for energy-efficient network optimization in 6G systems.