Broadband spectrum sensing method based on reinforcement learning

Abstract

The invention discloses a broadband spectrum sensing method based on reinforcement learning in the technical field of wireless spectrum sensing. The broadband spectrum sensing method comprises the steps that S1, time-division, frequency-division and sequential broadband sensing search modeling is carried out; S2, performing optimal sequential perception search based on a Markov decision process; and S3, carrying out rapid sequential perception search based on Monte Carlo tree search. According to the method, a Markov decision process (MDP) is used for modeling decision-related elements in a broadband search problem into a'state ', an 'action', an 'award' and a'transition probability', and definition of an optimal online search strategy is completed based on the MDP. According to the method, an online search strategy is rapidly solved through a Monte Carlo tree search algorithm, an asymmetric sparse decision tree is constructed through an offline simulation method, and the most potential channel is searched for in a centralized mode through limited computing resources, so that the decision time is effectively shortened. And the broadband spectrum resource utilization efficiency of the cognitive radio node is improved.

Description

technical field [0001] The invention relates to the technical field of wireless spectrum sensing, in particular to a wideband spectrum sensing method based on reinforcement learning. Background technique [0002] With the expansion of wireless spectrum management and utilization in the future, spectrum sensing tasks will face the challenge of how to quickly complete wideband spectrum detection. Especially in military scenarios, the spectrum used includes a large dynamic range from very low frequency (such as maritime communication) to extremely high frequency (such as radar and satellite communication), and different electromagnetic signals include narrowband, pulse, spread spectrum, Complex signal patterns such as frequency hopping. Due to the bandwidth limitation of signal sampling, it is difficult for a cognitive radio node (Secondary User, SU) to achieve a perfect restoration of wireless signals with a large dynamic range and detection of the spectrum status of each cha...

AI Technical Summary

Problems solved by technology

Therefore, it is often difficult for pre-planning methods to fully utilize the real-time dynamic information in the sensing process; instead, the optimal broadband sensing search should be an online decision-making process based on all relevant information

Images