The invention relates to a
distributed computing unloading method and device based on deep
reinforcement learning. The method comprises the steps that a calculation unloading framework is set, a communication model and a calculation model are established according to the calculation unloading framework, the communication model is used for calculating the
signal-to-
noise-
interference ratio of
terminal equipment, the calculation model is used for conducting local calculation and edge calculation on the
terminal equipment, and the
terminal equipment is calculated based on the calculation unloading framework, the communication model and the calculation model. And modeling a calculation unloading problem into a Markov
decision process, and carrying out optimization iteration solution on the Markov
decision process by utilizing a depth deterministic strategy gradient
algorithm of the double-Critic network to obtain an unloading decision. Due to the fact that the depth deterministic strategy gradient
algorithm of the double Critic networks is used for conducting optimization iteration solution, the double Critic networks are respectively fitted, the complexity of fitting of a single Critic network is reduced, the convergence speed of the Critic networks is improved, and the overall convergence speed of the model is greatly improved.