Spectrum Management Of Cognitive Base Station Based On Cooperative Q-learning For High-speed Railways

The expanding demand for wireless services has necessitated cognitive radio technology which is a cutting-edge research area that overcomes the limitations of the conventional static spectrum allocation policy.The rapid development of high-speed railway puts forward a growing demand for an improved wireless communication technology.So it is urgent to apply cognitive radio into high speed railway.The wireless communication in high speed railway is different from the others.In high speed railway,when the speed of the train is up to 350 kilometers per hour,there unavoidably arises some issues,such as Doppler shift,fast cell switching and the penetration loss.As relevant tests have shown,there is a frequent occurrence of low Qos of wireless services and even dropped calls due to these issues.This paper gives an effort toward developing a rail-road-specific spectrum management that meets the needs of future wireless communication systems for railways.In this paper we try to propose a novel concept of cognitive base station(CBS)which works as a spectrum assigner.It learns from feedback received through interactions with an external environment based on Q-Learning approach and assigns available spectrum to the passengers in the range of coverage.Simulation results show that after autonomous learning,our proposed spectrum management scheme can significantly improve the vehicular communication by decreasing frequency hopping.Aimed at rapid and accurate spectrum evaluation,learning the future states of the spectrum is necessary.In this paper,we model the primary users and secondary users with the sharing channels as a two state hidden Markov model(HMM).Then the prediction information can be used together with spectrum sensing results for spectrum management.Simulation results show the modified spectrum management can further decrease frequency hopping and unsuccessful transmission.Cooperative learning is a mechanism within a multi-agent system that can improve the quality of learning.Our proposed CBS-based wireless communication environment can be seen as a multi-agent system.In this paper,we develop the Q-Learning-based on belief allocation to study an optimal multi-CBS joint action policy which converges to the stable status.Experimental results show the cooperative learning based spectrum management can reduce cell-switching.