Cognitive Spectrum Scheduling Method for Internet of Vehicles Based on DNN and MCTS

The application of Internet of Vehicles technology has led to an increase in the demand for vehicle scheduling, but the available computational resources and spectrum in current wireless networks are limited and scarce. Therefore, in order to improve the efficiency of vehicle scheduling, a cognitive spectrum scheduling method based on Deep Neural Network and Monte Carlo Tree Search for vehicle networking has been proposed. It evaluates the priority of MCTS and provides spectrum resource scheduling solutions, and uses DNN for offline training to obtain an environment model. In the simulation experiment for this method, the CUR of the proposed method increased by a maximum of 19.3% compared to other methods used for comparison. Its ALC is 20.4% higher than other methods. Its convergence time is lower than all methods used for comparison, with a maximum difference of 59.3%. The mean MAE of the proposed method is 0.793, and the mean RMSE is 0.628. The results of MAE and RMSE demonstrate that the proposed method in the experiment exhibits the lowest errors, both in training and testing processes. The proposed method provides a certain technical foundation for the cognitive spectrum scheduling of the Internet of Vehicles.