Distributed cross-layer resource allocation using Correlated Equilibrium based stochastic learning

摘要: In recent years, cross-layer Resource Management (RM) has been widely considered as an efficient approach for improving the network performance. However, due to factors such limited knowledge about wireless environment and spontaneous characteristics of device behaviors, designing distributed RM scheme using paradigm becomes extremely difficult. this paper, a distributed, multi-users transmitting scalable video in ad-hoc CDMA is presented. For design, end-to-end performance criterion at APP layer applied resource allocations PHY, LINK layers are unified into single decision process. Due dynamic property channels uncoordinated user interactions, modeled framework stochastic game. game, each gradually learns its policy Correlated Equilibrium (CE) based Q-Learning algorithm. order calculate CE way, local devices apply regret-matching mechanism state-action value updating. Simulation experiments show that with proposed management scheme, system can be improved by 15% compared layered 10% purely on local-level learning.