Multi-interface adaptive data unloading method based on reinforcement learning in MEC environment

Abstract

The invention discloses a multi-interface adaptive data unloading method based on reinforcement learning in an MEC environment, and the method comprises the following steps: modeling an environment composed of a plurality of pieces of user equipment, an edge base station and a server data transmission strategy into a Markov decision model in a finite state; and constructing a reward function based on a Markov decision model, training the neural network by using a multi-agent deep reinforcement learning algorithm to obtain an optimal decision action, and determining a transmission mode of data unloading. The reinforcement learning can realize self-learning of a model-free state-to-action high-dimensional mapping relationship, and a multi-interface adaptive data unloading method is obtained based on the reinforcement learning, so that the pressure of a server side is effectively relieved, the utilization rate of network resources is improved, and the service life of the server side is prolonged. The purposes of relieving network congestion, reducing end-to-end time delay and reducing data unloading transmission energy consumption are achieved.

Description

technical field [0001] The invention relates to the technical field of data transmission in a network, and more specifically relates to a multi-interface adaptive data unloading method based on reinforcement learning in an MEC environment. Background technique [0002] Today, with the rapid development of mobile Internet technology, more and more smart devices have entered people's lives. Some applications that require high time delay, such as mobile high-definition video, AR / VR, etc., have created a large number of mobile network during use. Data Flow. The pressure and challenges brought by the explosive growth of data traffic to the mobile network are enormous, and its impact is manifested in: (1) huge backhaul network link pressure; (2) low wireless coverage and low energy efficiency; (3) The end-to-end time delay is relatively large. [0003] With the development of technology, the service capability of the cloud computing center is also constantly improving, but the n...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the present invention provides a multi-interface adaptive data offloading method based on reinforcement learning in the MEC environment. When dealing with changes in the network environment in the edge environment, the offloading strategy based on static scheduling has the problem of poor real-time performance. The unloading strategy has the problem of high unloading failure rate, which leads to time delay and energy consumption in workflow scheduling

Images