Cooperative spectrum sharing intelligent contract design method based on deep neural network

Abstract

The invention belongs to the technical field of wireless communication, and particularly relates to a cooperative spectrum sharing intelligent contract design method based on a deep neural network. Aimed at the double asymmetric dynamic network information problems of information and behavior hiding before and after contract signing, on the basis of a mathematical model description method for researching a multi-stage flow unloading capacity type and a flow unloading service behavior of a small base station, a multi-stage private information discrimination mechanism is explored to avoid a dynamic reverse selection problem. A multi-stage private behavior incentive strategy is designed to avoid a dynamic morality risk behavior. The method aims to solve the problem of non-convex optimizationof a multi-stage dynamic contract model in a double asymmetric dynamic network information scene. Private information and reputation information of a small base station are utilized, a multi-stage intelligent contract model based on deep learning is explored by establishing a multi-layer neural network structure framework, and a multi-stage dynamic contract optimization design strategy under a dual-information asymmetric scene is studied, so that multi-stage cooperative spectrum sharing implementation is guaranteed.

Description

technical field [0001] The invention belongs to the technical field of wireless cooperative communication, and in particular relates to a multi-stage cooperative spectrum sharing contract method under the condition of dynamic asymmetric information. Background technique [0002] Due to the complexity of the cognitive wireless network environment and the dynamic nature of network information, dynamic network information cannot be accurately obtained, and the information and behavior of Small Base Stations (SBSs) are often unverifiable. Therefore, in the cooperative spectrum sharing contract mechanism, the problems of dynamic adverse selection and dynamic moral hazard often appear at the same time. For example, before each phase of signing, Macro Base Stations (MBSs) do not know the capability information of SBSs in the current stage, that is, the problem of dynamic reverse selection; after each phase of signing, MBSs do not know whether SBSs are actively participating in traf...

AI Technical Summary

Problems solved by technology

[0003] In addition, the optimization problem of cooperative spectrum sharing contract mechanism is a non-convex optimization problem

Most of the existing methods use strategies such as approximation, restricting the solution domain, or simplifying to convex optimization subproblems, which will lead to some performance loss

Although some heuristic optimization methods can obtain approximate optimal solutions, due to their own limitations, they cannot satisfactorily solve large, complex, dynamic non-convex optimization problems

At the same time, most of the existing methods solve the above optimization problems through iterative strategies, the calculation cost is too high and cannot meet the real-time requirements of the dynamic environment.

In addition, most of the traditional optimization algorithms based on machine learning are based on the construction of deep neural networks to approximate iterative optimization algorithms, but such methods can only solve relatively simple iterative problems.

Although researchers have exploratoryly applied deep neural networks to some areas of communication, most of them focus on signal modeling and detection rather than optimization algorithms

Images