A Refined Energy Optimization Model for Edge Computing with Machine Learning.

Edge computing (EC) with artificial intelligence (AI) /machine learning (ML) is a promising paradigm in current 5GAdvanced and future 6G wireless technologies. Energy optimization is a primary issue in most EC/ML systems. In this paper, a refined model is developed to seek optimal energy allocation. The present work introduces several features which were often omitted in existing studies, including fine-grained discrete optimization, non-singular CPU cycle allocation, longtailed data traffic, and non-singular pathloss terms. The energy optimization model is solved by the ML methodology and compared with a conventional solver. Simulations showed that the efficiency may be improved more than 90%.