ADMM for Mobile Edge Intelligence: A Survey

With the rapid development of mobile networks, centralized intelligent computing services struggle to meet the Quality of Service (QoS) requirements of mobile devices. Mobile edge intelligence (MEI) pushes artificial intelligence (AI) to the network edge to deliver intelligent services, including data analysis, model training, learning, decision control, optimization, storage, and communications. Traditional approaches within Mobile Edge Computing (MEC), face challenges such as communication overhead, task execution time, energy consumption, and dynamic network topology. Recent studies have demonstrated the significance of the alternating direction method of multipliers (ADMM). ADMM is a distributed optimization algorithm for efficiently solving constrained problems. This algorithm effectively decomposes a complex problem into simpler sub-problems, each of which can be processed independently and iteratively. The paper first thoroughly analyzes the challenges in MEI and demonstrates the advantages of ADMM over current methods. Then, we provide a comprehensive introduction to the ADMM algorithms, including detailed derivation and analysis of ADMM’s properties, classification of ADMM variants, and ADMM evolution in the MEI field. Finally, we show various applications and future research directions of ADMM in MEI, covering computation offloading, resource allocation, edge caching, privacy protection, unmanned aerial vehicle (UAV) trajectory planning, network slicing, and joint optimization.