Large-scale machine type communication random access backoff method based on machine learning

The invention belongs to the technical field of Internet-of-Things communication, and discloses a large-scale machine type communication random access backoff method based on machine learning, which comprises the following steps: S1, dynamically dividing virtual small cells, and classifying MTCD services; S2, establishing a Markov decision process for a backoff access problem; S3, obtaining MTCD service quality parameters, and constructing a reward function; S4, constructing an evaluation network Critic and a strategy network Actor; S5, generating a backoff access decision by using the policy network Actor; S6, minimizing the TD error, updating neural network parameters of the evaluation network Critic, and transmitting the TD error to the strategy network Actor; S7, obtaining a TD error as a dominant function to update neural network parameters of the strategy network Actor, and adjusting the strategy to improve the return; S8, performing N times of loop iteration, and training and updating network parameter