Artificial intelligence-based networking method and device for fog radio access networks

Patent metadata

Field	Value
Publication number	US-11201784-B2
Application number	US-201916718457-A
Country	US
Kind code	B2
Filing date	Dec 18, 2019
Priority date	Dec 18, 2018
Publication date	Dec 14, 2021
Grant date	Dec 14, 2021

Abstract

An artificial intelligence-based networking method for fog radio access networks, which includes: a central computing logic module receives reported data which includes measurement report data from user terminals, wireless transmission data from base stations, and operation and maintenance data from a radio access network. Based on these reported data and proper machine learning algorithms, the central computing logic module configures an operating mode of a radio access network that matches user behavior, service attributes, and radio access network performance indicators. According to the operating mode, an edge computing logic module determines whether to optimize a current configuration of an edge communication entity and allocation of radio resources, computing resources, and caching resources. With proper machine learning algorithms, the proposed networking method meets various service requirements. By configuring the radio access network flexibly, the method enables the radio access network to adapt to different application scenarios and performance objectives.

First claim

What is claimed is:

1. An artificial intelligence-based networking method for fog radio access networks (F-RANs), comprising:
receiving, by a central computing logic module, reported data which includes: measurement report data from user terminals, wireless transmission data from base stations, and operation and maintenance data from a radio access network, wherein the measurement report data relates to user behavior, the wireless transmission data relates to performance indicators of the radio access network, and the operation and maintenance data relates to service attributes;
configuring, by the central computing logic module, based on the reported data obtained during a cycle T 1 and a first machine learning algorithm, an operating mode of the radio access network that matches the user behavior, the service attributes, and the performance indicators of the radio access network;
receiving, by an edge computing logic module, information of the operating mode of the radio access network from the central computing logic module, and determining, according to the operating mode of the radio access network, during a cycle T 2 , whether a current configuration of an edge communication entity corresponding to the edge computing logic module meets a networking performance aim which is that a variation of a performance indicator regarding performance of the edge communication entity does not exceed a preset threshold, wherein the cycle T 2 is shorter than the cycle T 1 ; and
if the current configuration of the edge communication entity meets the networking performance aim, allocating, by the edge computing logic module, resources to the user terminals connected to the edge communication entity, to network the edge communication entity and the user terminals that are allocated with the resources as an F-RAN, wherein the resources comprise radio resources, computing resources, and caching resources.

2. The method according to claim 1 , wherein, during the cycle T 1 , the first machine learning algorithm is a representation learning algorithm, and the representation learning algorithm comprises:
obtaining dimension-reduced data from the reported data through data feature classification; and
analyzing latent factors existing in the dimension-reduced data and constructing a valid representation, and outputting the operating mode of the radio access network that matches the user behavior, the service attributes, and the radio access network performance indicators.

3. The method according to claim 1 , wherein:
in a case that the operating mode of the radio access network is a wide-area seamless coverage mode, high power nodes which are macro base stations are implemented, and a user terminal selects a serving macro base station according to a strongest received power, wherein the macro base stations are configured with completed protocol stack functions and multi-input multi-output (MIMO) and advanced multiple access methods;
in a case that the operating mode of the radio access network is a hotspot high capacity mode, a small base station is used to connect user terminals at a hotspot to the radio access network, if interference of the small base station is higher than a preset threshold θ1 or an average capacity of user terminal at the hotspot is higher than a preset threshold θ2, the small base station is transformed to a remote radio head (RRH), other functions of the small base station including functions on a physical layer, a media access control (MAC) layer, and a radio resource control (RRC) layer are moved to a base station unit (BBU) pool, and the BBU pool is connected to the RRH through a fronthaul link;
in a case that the operating mode of the radio access network is a massive-connection low power mode, a clustering mechanism is adopted, and in the clustering mechanism, adjacent user terminals are formed into a mesh or tree-like topology cluster;
packet traffic generated by cluster members is delivered to a selected cluster head through device-to-device or multi-hop relay, the cluster head directly accesses the radio access network, wherein the cluster head is a user terminal, of which a number of connected user terminals is more than a preset threshold θ3, and the cluster members are user terminals in the cluster except the cluster head; and
in a case that the operating mode of the radio access network is a low-latency high-reliability mode, a fog access point (F-AP), of which a transmission delay is less than a preset threshold θ4 is utilized, to enable user terminals to be connected to the radio access network with a multiple access method; or
if a transmission delay between the F-AP and a target user terminal is greater than a preset threshold θ4, while a transmission delay between the target user terminal and its neighbor user terminal is less than the preset threshold θ4, the target user terminal accesses the neighbor user terminal with a device-to-device communication method.

4. The method according to claim 1 , further comprising:
monitoring, by the central computing logic module, the measurement report data from all the user terminals in the radio access network, and checking whether an obtained quality of service and a number of active user terminals exceed respective preset thresholds, and if not, extending a duration of the cycle T 1 ; otherwise shortening the duration of the cycle T 1 .

5. The method according to claim 3 , wherein, according to the operating mode of the radio access network, determining, by the edge computing logic module, whether the configured edge communication entity meets the networking performance aim comprises:
during the cycle T 2 , monitoring, by the edge computing logic module, the performance of the edge communication entity and checking whether the variation of the performance indicator exceeds the preset threshold;
if the variation of the performance indicator exceeds the preset threshold, determining that the current configuration of the edge communication entity does not meet the networking performance aim, and that there is a need for the edge computing logic module to optimize the current configuration of the edge communication entity.

6. The method according to claim 5 , wherein:
in the case that the operating mode of the radio access network is the wide-area seamless coverage mode, the edge computing logic module utilizes a deep reinforcement learning algorithm DRL1 to optimize the edge communication entity, where transmission power and a handover parameter of the macro base stations are optimized;
in the case that the operating mode of the radio access network is the hotspot high capacity mode, the edge computing logic module utilizes a deep reinforcement learning algorithm DRL2 to optimize the edge communication entity, where a number of activated small base stations or RRHs and update of cached content are optimized;
in the case that the operating mode of the radio access network is the massive-connection low power mode, the edge computing logic module utilizes a decision tree algorithm and a second machine learning algorithm to optimize the edge communication entity, where a decision tree model is trained to select the cluster head and the second machine learning algorithm is utilized to optimize routing from the cluster members to the cluster head;
in the case that the operating mode of the radio access network is the low-latency high-reliability mode, the edge computing logic module utilizes a deep Bayesian learning algorithm to optimize the edge communication entity, wherein according to historical information on an access node of a user terminal, a future access node selection of the user terminal is predicted, and mobility-related parameters of the access node are optimized; and
the edge computing logic module checks whether the o

Assignees

• Univ Beijing Posts & Telecomm

Inventors

• Peng Mugen
• Xiang Hongyu
• Yan Shi

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• H04L41/40

  using virtualisation of network functions or resources, e.g. SDN or NFV entities · CPC title

• H04L41/16

  using machine learning or artificial intelligence · CPC title

• H04L41/5009

  Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF] · CPC title