A Practical Multiplayer Multi-armed bandit Algorithm for Smart City Communication System

摘要

In this paper, the authors have addressed a Multiplayer Multi-Armed Bandit (MMAB) problem, where M players work together towards maximizing their cumulative reward following some Bandit policy. We have identified that the Cognitive Radio’s (CR) decision making inside CR equipment can be formalized as Multi-Armed Bandit problems. Cognitive radio enables us to develop a smart city heterogeneous communication system. First, we discuss the Multiplayer-MAB issue. Hence proceeding with the problem, authors propose optimal decentralized algorithms that achieve similar asymptotic regret as equivalent to that obtained by an optimal centralized algorithm. We then aim for a practical algorithm that is more suitable for real world problems and which have applications for smart city communication technology.