Routing decision method and system based on traffic prediction

Patent metadata

Field	Value
Publication number	US-12587941-B2
Application number	US-202318874086-A
Country	US
Kind code	B2
Filing date	Jun 7, 2023
Priority date	Jun 13, 2022
Publication date	Mar 24, 2026
Grant date	Mar 24, 2026

Abstract

Disclosed are a routing decision method and system based on traffic prediction. The method is used for obtaining each data link to be analyzed between a starting point and an ending point of a direct connection or an indirect connection for forwarding data packets to be forwarded in a target network area, and each node respectively included in each data link to be analyzed; when the target network area receives the data packets to be forwarded, training and obtaining a routing selection model corresponding to said packets for each data link to be analyzed for the direct connection or the indirect connection between the starting point and the ending point in the target network area, respectively; and applying the routing selection model to obtain an optimal forwarding path corresponding to said data packets, and forwarding said data packets.

First claim

What is claimed is:

1 . A routing decision method based on traffic prediction, used for obtaining each data link to be analyzed between a starting point and an ending point of a direct connection or an indirect connection for forwarding data packets to be forwarded in a target network area, and each node respectively included in each data link to be analyzed; wherein in a case where the target network area receives the data packets to be forwarded, based on positions of the data packets to be forward in the target network area, in view of the each data link to be analyzed for the direct connection or the indirect connection between the starting point and the ending point in the target network area, respectively, a routing selection model corresponding to the data packets to be forwarded on the each data link to be analyzed are trained and obtained through following Step A to Step D, then by applying the routing selection model, an optimal forwarding path corresponding to the data packets is obtained through following Step E, and the data packets to be forwarded are forwarded according to the optimal forwarding path;
Step A, collecting and obtaining preset data types respectively corresponding to each forwarded data packet on the data link to be analyzed within one historical time period in a direction from a current time instant to a historical time instant, obtaining, in view of the each forwarded data packet, the each data link to be analyzed that the forwarded data packets pass through in the target network area, as well as each node included in the each data link to be analyzed, thereby obtaining a global view corresponding to the forwarded data packets, that is, obtaining global views corresponding to the each forwarded data packet, respectively, and then entering Step B; Step B, constructing, in view of the each global view, a feature extraction module that is configured to extract features from the global views and further obtain link classification features corresponding to the global views, and then entering step C; Step C, collecting and obtaining preset data types corresponding to the data packets to be forwarded on the data links to be analyzed, further obtaining global views corresponding to the data packets to be forwarded, and obtaining, by utilizing the feature extraction module, link classification features corresponding to the data packets to be forwarded, that is, obtaining global views corresponding to the data packets to be forwarded and the forwarded data packets, respectively; further constructing, based on the each preset data type corresponding to the data packets to be forwarded as well as the each link classification feature corresponding to the data packets to be forwarded, a classification prediction module that is configured to perform a feature conversion on the data packets to be forwarded and obtain a classification feature matrix for the each global view, and then entering Step D; Step D, training, on a basis of the classification feature matrix obtained by the classification prediction module, a routing selection model to be trained for the data link to be analyzed, by taking forwarded data types corresponding to the each forwarded data packet on the data link to be analyzed as an input, by taking a complete forwarding path in the global view as an output, by taking link classification features corresponding to the forwarded data packets on the data link to be analyzed as training samples, to obtain a routing selection model that is configured to select the optimal forwarding path; and then entering Step E; and Step E, applying, on a basis of the link classification features corresponding to the data packets to be forwarded on the each data link to be analyzed, the routing selection module for the data packets to be forwarded, by taking the each preset data type corresponding to the data packets to be forwarded as an input, to determine the optimal forwarding path of the data packets to be forwarded, and completing the forwarding of the data packets to be forwarded.

2 . The routing decision method based on the traffic prediction according to claim 1 , wherein in Step A, in view of the each forwarded data packet, the each global view corresponding to the forwarded data packets is obtained respectively, wherein the global view includes the each data link to be analyzed that forwards the forwarded data packets and the each node contained in the each data link to be analyzed;
a global view G corresponding to the forwarded data packets respectively is characterized through an adjacency matrix according to a following formula,
G ( V, E, {f v ( u )}, { f e ( u, v )});
where V denotes a set of the nodes contained in the data link to be analyzed, E denotes a set of the data links to be analyzed corresponding to the forwarded data packets, u and v denote vertices of the global view respectively, that is, nodes of the data link to be analyzed, e denotes the data link to be analyzed, f v (u) denotes node features, f e (u, v) denotes features of the data link e to be analyzed; node features f e (u, v) of the data link e to be analyzed are obtained according to a following formula:

f
e

(

u
,
v

)

=

log
⁢

(

1
+

(

c
e

(

u
,
v

)

-

a
i
e

(

u
,
v

)

/

c
e

(

u
,
v

)

)

)

,

where c e (u, v) denotes a full load capacity of the data link e to be analyzed, and a i e (u, v) denotes a preset data type of a forwarded data packet i corresponding to the data link e to be analyzed; and
the node features f v (u) is obtained according to a following formula:

Assignees

• Univ Nanjing Posts & Telecommunications

Inventors

• Guo Yongan
• Yu Dequan
• She Hao
• Qian Qijie

Classifications

• H04L45/08

  Learning-based routing, e.g. using neural networks or artificial intelligence · CPC title

• H04L45/38

  Flow based routing · CPC title

• G06N20/00

  Machine learning · CPC title

• H04W40/02 Primary

  Communication route or path selection, e.g. power-based or shortest path routing · CPC title