Research On Spectrum Sensing Algorithm Of Cognitive V2X Based On Deep Learning

Vehicle is one of the convenient travel tools.With the improvement of people’s living standards,the penetration rate of vehicles is getting higher,and the number of vehicles on the road is increasing every year,resulting in more traffic congestion and traffic accidents.Vehicle to everything(V2X)communication technology can effectively expand vehicle perception capabilities,thereby improving road traffic efficiency and reducing traffic accidents.The safe information transmitted in V2 X requires extremely high latency and reliability.Therefore,in the current situation where the spectrum resources of V2 X are very scarce,efficient use of the spectrum outside the 5G mobile communication band has become one of the key technologies of V2 X.Cognitive radio recognizes spectrum holes in licensed spectrum through spectrum sensing and allows unauthorized users to access it,thus becoming one of the ways to efficiently utilize the current limited spectrum resources.In this paper,we first introduce the relevant theories of spectrum sensing technology in V2 X.Then in view of the limitations of the existing spectrum sensing technology,a spectrum sensing algorithm based on deep learning is proposed for V2 X,which is optimized according to the communication requirements.The specific work is as follows:1.A novel single-user spectrum sensing algorithm based on Res Ne Xt,which is an improved residual network is proposed.According to the vehicle to base station(V2B)communication channel model defined in the Third Generation Partnership Project(3GPP),three fading factors including path loss,shadow effect and small-scale fading under the complex channel of vehicles moving at high speed are considered to establish a more realistic wireless environment,where the primary user signal transmitter communicate with the secondary user.Then,the modulated signal data sampled on this communication channel is processed and fed into the neural network for offline training.For different V2 X conmmunication environment,detection performance of the well-trained neural network is compared with that of the existing spectrum sensing methods based on deep learning.The simulation results show that the detection performance of our proposed algorithm under the urban environment reaches 97% when the signal-to-noise ratio is-13 d B,which is 5% higher than that of the single-user spectrum sening algorithm based on Res Net with shorter detect time.2.A quantized cooperative spectrum sensing algorithm based on deep learning is proposed.To solve the problems of single-user spectrum sensing technology,a multi-user cooperative spectrum sensing algorithm based on deep learning is introduced.At the same time,due to the limited capacity of the communication channel for the vehicle to upload local spectrum sensing information to the fusion center,the local spectrum sensing information is quantized.The simulation results show that the proposed quantization algorithm can achieve the optimal detection performance while making full use of the limited vehicle communication channel capacity.