Wireless network resource allocation method based on generative adversarial reinforcement learning
A technology of wireless network resources and allocation method, applied in the field of wireless network resource allocation based on generative confrontation reinforcement learning, to achieve the effect of improving wireless network performance
Patent Information
- Authority / Receiving Office
- CN · China
- Current Assignee / Owner
- Publication Date
- 2020-05-19
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The present invention relates to the field of wireless network resource allocation and reinforcement learning, and more particularly, to a wireless network resource allocation method based on generative confrontation reinforcement learning. Background technique
[0002] The 5G network will support a large number of diverse business scenarios from vertical industries, such as smart security, high-definition video, telemedicine, smart home, autonomous driving, and augmented reality, etc. These business scenarios usually have different communication requirements, such as augmented reality technology requires more Low latency, autonomous driving technology requires the network to provide higher reliability. However, traditional mobile networks are mainly designed to serve a single mobile broadband service and cannot adapt to the diversified service scenarios of 5G in the future. Building a dedicated physical network for each business scenario will inevita...
Examples
Embodiment
[0085] On the host with the configuration shown in Table 1, a simulation environment is written in Python language, and three different types of services (calling, video and ultra-reliable and low-latency services) are used as examples for testing. The resources to be allocated are wireless bandwidth, the total bandwidth is 10M, and the allocation granularity is 1M, so there are 36 allocation strategies in total, that is, the number of valid actions is 36. The discount factor γ is set to 0.9, the number of samples N for sampling the overall return distribution is 50, and the initial value of ∈ is 0.9, which decreases by 0.05 every 100 runs of the algorithm, and remains unchanged when it decreases to 0.05. buffer size N B is 10000. The G network has 3 neurons in the input layer, 512 neurons in the first hidden layer, 512 neurons in the second hidden layer, and 1800 neurons in the output layer. The D network has 50 neurons in the input layer, 256 neurons in the first hidden ...