CN113438173B - Routing and spectrum allocation method, device, storage medium and electronic equipment - Google Patents

Abstract

The present invention relates to a routing and spectrum allocation method, device, storage medium and electronic device. The method includes: obtaining a service connection request, obtaining a candidate path and the corresponding link, according to the candidate path and the corresponding link path, establish a path link relationship matrix, calculate the number of idle frequency slots, idle frequency slots, the number of idle frequency slots, and the average size of idle frequency slots on each candidate path, according to the number of frequency slots to be occupied. and the idle frequency slot, obtain a set of available frequency slots, in the set of available frequency slots, obtain the starting position and size of the available frequency slots, obtain a one-dimensional spectrum state distribution vector, and connect the service connection request, The path link relationship matrix and the spectrum state distribution vector are input to the trained routing and spectrum allocation model, and the routing and spectrum allocation results corresponding to the service connection request are obtained, which improves the utilization rate of spectrum resources.

Description

Routing and spectrum allocation method, apparatus, storage medium, and electronic device

technical field

The present invention relates to the technical field of optical fiber communication, and in particular, to a routing and spectrum allocation method, device, storage medium and electronic device.

Background technique

In recent years, research on applying deep reinforcement learning to routing and spectrum allocation in optical networks has received much attention. For machine learning problems, the data and features determine the upper limit of performance, and models and algorithms only approach this upper limit. In deep reinforcement learning based routing and spectrum allocation, the input state represents the structure and features of the data, so the state representation is very critical.

However, most of the current deep reinforcement learning-based routing and spectrum allocation methods do not consider whether the frequency slot resource distribution of the link in the optical network is concentrated and whether the frequency slot resource state is congested in their state representation, which leads to the problem of congested frequency slot resources in the optical network. In the process of routing spectrum allocation, frequency slot resource utilization is low and network blocking rate is high.

SUMMARY OF THE INVENTION

Based on this, the purpose of the present invention is to provide a routing and spectrum allocation method, device, medium and electronic device, which have the advantages of improving the utilization rate of spectrum resources and reducing the network blocking rate.

According to a first aspect of the embodiments of the present application, a routing and spectrum allocation method is provided, including the following steps:

Acquiring a service connection request and the frequency slot resource status of the current optical network, where the service connection request includes a source node, a destination node and a spectrum width;

According to the service connection request and the frequency slot resource state of the current optical network, obtain multiple candidate paths between the source node and the destination node and the service connection request needs to be performed on each candidate path The number of frequency slots occupied; wherein, each of the candidate paths consists of one or more links;

establishing a path link relationship matrix according to the candidate path and the corresponding link;

Calculate the number of idle frequency slots, idle frequency slots, the number of idle frequency slots, and the average size of idle frequency slots on each candidate path according to the state of each frequency slot on the candidate path;

Obtain a set of available frequency slots according to the number of frequency slots to be occupied and the idle frequency slots;

In the set of available frequency slots, obtain the starting position and size of the available frequency slot at the frontmost position on the frequency slot axis;

Splicing the number of idle frequency slots, the number of idle frequency slots, the size of the set of available frequency slots, the starting position and size of the available frequency slots, and the average size of the idle frequency slots , obtain a one-dimensional spectral state distribution vector;

The service connection request, the path link relationship matrix, and the spectrum state distribution vector are input into the trained routing and spectrum allocation model to obtain a route and spectrum allocation result corresponding to the service connection request.

According to a second aspect of the embodiments of the present application, a routing and spectrum allocation apparatus is provided, including:

a request acquisition module, configured to acquire a service connection request and a frequency slot resource state of the current optical network, where the service connection request includes a source node, a destination node and a spectrum width;

A path obtaining module is configured to obtain a plurality of candidate paths between the source node and the destination node and the service connection request in each path according to the service connection request and the frequency slot resource state of the current optical network. The number of frequency slots to be occupied on the candidate path; wherein, each candidate path consists of one or more links;

a matrix establishment module, configured to establish a path link relationship matrix according to the candidate path and the corresponding link;

A frequency slot calculation module, configured to calculate the number of idle frequency slots, idle frequency slots, the number of idle frequency slots, and idle frequency slots on each candidate path according to the state of each frequency slot on the candidate path average size;

a frequency slot obtaining module, configured to obtain a set of available frequency slots according to the number of frequency slots to be occupied and the idle frequency slots;

a position obtaining module, configured to obtain, in the set of available frequency slots, the starting position and size of the available frequency slot that is at the frontmost position on the frequency slot axis;

A vector acquisition splicing module, configured to combine the number of free frequency slots, the number of free frequency slots, the size of the set of available frequency slots, the starting position and size of the available frequency slots, and the free The average size of the frequency slot is spliced to obtain a one-dimensional spectrum state distribution vector;

A result obtaining module is used to input the service connection request, the path link relationship matrix, and the spectrum state distribution vector into the trained routing and spectrum allocation model to obtain the route and spectrum corresponding to the service connection request Assign results.

According to a third aspect of the embodiments of the present application, an electronic device is provided, including: a processor and a memory; wherein, the memory stores a computer program, and the computer program is adapted to be loaded by the processor and execute any of the above A routing and spectrum allocation method as described.

According to a fourth aspect of the embodiments of the present application, there is provided a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the routing and spectrum allocation method described in any one of the above.

The embodiment of the present application acquires a service connection request and the frequency slot resource status of the current optical network, where the service connection request includes a source node, a destination node and a spectrum width, according to the service connection request and the frequency slot resource of the current optical network state, to obtain multiple candidate paths between the source node and the destination node and the number of frequency slots that the service connection request needs to occupy on each candidate path; wherein, each candidate path consists of a or multiple links, according to the candidate path and the corresponding link, establish a path link relationship matrix, according to the status of each frequency slot on the candidate path, calculate the idle on each candidate path The number of frequency slots, the number of idle frequency slots, the number of idle frequency slots, and the average size of the idle frequency slots. In the set of available frequency slots, obtain the starting position and size of the available frequency slot at the frontmost position on the frequency slot axis, and calculate the number of free frequency slots, the number of free frequency slots, and the available frequency slots. The size of the slot set, the starting position and size of the available frequency slots, and the average size of the idle frequency slots are spliced to obtain a one-dimensional spectrum state distribution vector, and the service connection request, the path chain The route relationship matrix and the spectrum state distribution vector are input to the trained routing and spectrum allocation model, and the routing and spectrum allocation results corresponding to the service connection request are obtained. The present invention fully considers whether the frequency slot resource distribution of the link in the optical network is concentrated and whether the frequency slot resource state is not The problem of congestion is further reduced, and the network congestion rate of routing and spectrum allocation model for routing and spectrum allocation is reduced, and the utilization rate of spectrum resources is improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not limiting of the present application.

For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings.

Description of drawings

1 is a schematic flowchart of a routing and spectrum allocation method of the present invention;

Fig. 2 is the schematic flow chart of S20 calculating the visual saliency map in the routing and spectrum allocation method of the present invention;

3 is a schematic flowchart of S22 in the routing and spectrum allocation method of the present invention;

4 is a schematic diagram of an optical network topology in the routing and spectrum allocation method of the present invention;

5 is a schematic flowchart of S30 in the routing and spectrum allocation method of the present invention;

6 is a schematic flowchart of S40 in the routing and spectrum allocation method of the present invention;

7 is a schematic diagram of a link frequency slot state in the routing and spectrum allocation method of the present invention;

8 is a schematic flowchart of S50 in the routing and spectrum allocation method of the present invention;

9 is a schematic flowchart of S80 in the routing and spectrum allocation method of the present invention;

10 is a structural block diagram of a routing and spectrum allocation device of the present invention;

11 is a structural block diagram of the path obtaining module 92 of the routing and spectrum allocation device of the present invention;

12 is a structural block diagram of the frequency slot number calculation unit 924 of the routing and spectrum allocation device of the present invention;

13 is a block diagram of the structure of the routing and spectrum allocation device matrix building module 93 of the present invention;

14 is a structural block diagram of the frequency slot calculation module 94 of the routing and spectrum allocation device of the present invention;

15 is a structural block diagram of the frequency slot obtaining module 95 of the routing and spectrum allocation device of the present invention;

FIG. 16 is a structural block diagram of the result obtaining module 98 of the routing and spectrum allocation apparatus of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

It should be clear that the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

Terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the embodiments of the present application. As used in the embodiments of this application and the appended claims, the singular forms "a," "the," and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It will also be understood that the term "and/or" as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with this application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application, as recited in the appended claims. In the description of this application, it should be understood that the terms "first", "second", "third", etc. are only used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence, nor can understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific situations.

Also, in the description of the present application, unless otherwise specified, "a plurality" means two or more. "And/or", which describes the association relationship of the associated objects, means that there can be three kinds of relationships, for example, A and/or B, which can mean that A exists alone, A and B exist at the same time, and B exists alone. The character "/" generally indicates that the associated objects are an "or" relationship.

Referring to FIG. 1, an embodiment of the present invention provides a routing and spectrum allocation method, including the following steps:

S10. Acquire a service connection request and the frequency slot resource status of the current optical network, where the service connection request includes a source node, a destination node and a spectrum width.

An optical network is a wide area network, a metropolitan area network or a newly built large-scale local area network that uses optical fibers as the main transmission medium. Optical fibers are connected through nodes, and multiple adjacent nodes are connected to form links of the optical network. The frequency slot is the unit that stores and transmits data in the link. The fixed bandwidth of each frequency slot is 12.5Gbps. The frequency slot resource is the number of frequency slots that can be accommodated on all links in the optical network. The state of the spectrum resource includes the occupancy of the frequency slot. state and idle state. The occupied state of the frequency slot refers to the state of the frequency slot that has been allocated during the service connection process, and the idle state refers to the state of the frequency slot that has not been allocated during the service connection process.

A service connection request is a request to establish optical fiber communication from one node of an optical network to another node to transmit service data. The data input node is the source node, the data receiving node is the destination node, and the data transmission rate is the spectrum width.

In the embodiment of the present application, after the optical network has been running for a period of time, during the dynamic establishment and removal of service connections, frequency slots are continuously allocated, released, and reused, so that some frequency slots in the optical network are in an occupied state, and some frequency slots are in an occupied state. The slot is in an idle state. Therefore, when establishing a service connection request, the frequency slot resource state of the current optical network should be obtained. The source and destination nodes are generated from the current optical network through a Poisson process, and the spectral bandwidth is set to a multiple of 12.5 between 25 and 100 in Gbps. Wherein, the Poisson process is to select two random nodes in the optical network topology, and make the two nodes mutually exclusive.

S20. According to the service connection request and the frequency slot resource state of the current optical network, obtain multiple candidate paths between the source node and the destination node and the service connection request in each candidate path The number of frequency slots to be occupied; wherein, each candidate path consists of one or more links.

In this embodiment of the present application, from the source node to the destination node, there are multiple other nodes in the middle, links are formed between adjacent nodes, and multiple links are combined to form a path. Therefore, there are multiple links from the source node to the destination node. paths, these paths are used as candidate paths for the service connection request of the present application, wherein each candidate path consists of one or more links. The number of frequency slots that the service connection request needs to occupy on each of the candidate paths can be calculated according to the difference in the selection positions of the source node and the destination node and the difference in the spectrum width in the specific service connection request.

S30. Establish a path link relationship matrix according to the candidate paths and the corresponding links.

The path link relationship matrix is used to represent the connection relationship between multiple links on each candidate path in the optical network topology.

S40. Calculate the number of idle frequency slots, idle frequency slots, the number of idle frequency slots, and the average size of idle frequency slots on each candidate path according to the state of each frequency slot on the candidate path.

In this embodiment of the present application, the state of each frequency slot includes an occupied state and an unoccupied state (idle state). On the frequency slot axis, the occupied state can be represented by 0, and the idle state can be represented by 1. By accumulating the frequency slots of 1 on the frequency slot axis of each candidate path, each of the The number of free frequency slots on the candidate path.

The idle frequency slot slot consists of one or more consecutive adjacent idle frequency slots. When acquiring the idle frequency slot, the state representation corresponding to the adjacent frequency slot on the frequency slot axis can be obtained by AND operation. Among them, the calculation rule of AND operation is: 1&1=1, 1&0=0, 0&0=0, when the frequency slot is in the occupied state, it is represented by 0, and when the frequency slot is in the idle state, it is represented by 1, and multiple consecutive adjacent frequency slots are In the idle state, the AND operation result of these consecutive adjacent frequency slots is 1. When the next adjacent frequency slot is in the occupied state, the AND operation result is 0, and the adjacent idle frequency slots before the occupied state are regarded as one idle state. frequency slots, in this way, the number of free frequency slots on each of the candidate paths is calculated.

The average size of the idle frequency slots is the ratio of the number of idle frequency slots to the number of idle frequency slots. If this value is large, it means that the distribution of idle frequency slots is sparse, the link traffic distribution is concentrated, and the network status is not congested. . On the contrary, if this value is small, it means that the idle frequency slots are densely distributed, the link traffic distribution is scattered, and the network status is congested.

S50. Obtain a set of available frequency slots according to the number of frequency slots to be occupied and the idle frequency slots.

In this embodiment of the present application, there may be multiple idle frequency slots, and if the number of idle frequency slots corresponding to the idle frequency slot is greater than or equal to the number of frequency slots to be occupied, the idle frequency slot In order to satisfy the available frequency slots of the service connection request, a plurality of the available frequency slots constitute a set of available frequency slots.

S60. In the set of available frequency slots, obtain the starting position and size of the available frequency slot at the frontmost position on the frequency slot axis.

Optionally, the first-fit method can be used to obtain the starting position and size of the available frequency slot slot at the frontmost position on the frequency slot axis. Among them, the First-Fit method is a kind of continuous physical memory allocation method, which connects the free memory blocks according to the address from small to large. When allocating memory, search backwards from the head of the linked list, that is, from low address to high address. Once a memory block that can meet the requirements is found, the memory block is allocated. The starting position and size are characteristic information of available frequency slots that meet bandwidth requirements.

S70. Calculate the number of idle frequency slots, the number of idle frequency slots, the size of the set of available frequency slots, the starting position and size of the available frequency slots, and the average of the idle frequency slots Size splicing to obtain a one-dimensional spectral state distribution vector.

In this embodiment of the present application, the number of free frequency slots, the number of free frequency slots, the size of the set of available frequency slots, the starting position and size of the available frequency slots, and the free The average size of the frequency slots is spliced into a one-dimensional spectrum state distribution vector, and the six elements in the spectrum state distribution vector corresponding to these six values are used to represent the spectrum distribution state of the link.

S80. Input the service connection request, the path link relationship matrix, and the spectrum state distribution vector into the trained routing and spectrum allocation model, and obtain a routing and spectrum allocation result corresponding to the service connection request.

Before inputting the service connection request, the path link relationship matrix, and the spectrum state distribution vector into the trained routing and spectrum allocation model, this embodiment of the present application further trains the routing and spectrum allocation model. The routing and spectrum allocation model is composed of multiple agents running in parallel, and the routing and spectrum allocation model is trained based on the asynchronous dominant action evaluation method of deep reinforcement learning, wherein each agent receives the state representation of the optical network. , and then select a candidate path from the multiple candidate paths from the source node to the destination node for routing and spectrum allocation. If the routing and spectrum allocation are successful, the receiving reward value is positive. If the path selection and spectrum allocation are unsuccessful, the receiving reward value is burden. In this embodiment of the present application, the state representation includes a service connection request, the path link relationship matrix, and the spectrum state distribution vector. Specifically, each agent includes a policy network and a value network. The policy network receives the state representation, and how to process it to generate an action. The action acts on the optical network and is used to select a candidate path from the optical network for routing. And spectrum allocation, receive reward value after allocation is complete. The value network evaluates the reward value. If the data transmission in the service connection request is successful from the source node to the destination node, the routing and spectrum allocation is successful; if the data transmission in the service connection request fails, the routing and spectrum allocation fails, and the reward value is -1.

和价值网络

，设置了所述策略网络的权重参数

和所述价值网络的权重参数

，

是时刻t接收的光网络状态表示，

是时刻t策略网络根据

产生的动作，更新权重参数

和

是针对每一段时间进行的，每一段时间包括

个连续的动作(时间步长

至

)，权重参数由

向

，

向

更新。经过一段时间的训练后，参数由下式给出的小批量(批量大小为

)进行梯度上升或下降更新。其中，更新公式为：policy network for each agent

and value network

, which sets the weight parameters of the policy network

and the weight parameters of the value network

,

is the optical network state representation received at time t,

is the time t policy network according to

Generated actions that update weight parameters

and

is performed for each period, each period includes

consecutive actions (time steps

to

), the weight parameter is given by

Towards

,

Towards

renew. After a period of training, the parameters are given by the mini-batch (batch size is

) for gradient ascent or descent updates. Among them, the update formula is:

是优势函数的估计值，由下式给出in,

is an estimate of the advantage function, given by

是策略分布的熵，

控制熵正则项的强度，

和

是学习率，

是折扣因子。经过循环迭代训练，对所述策略网络和所述价值网络进行优化，获得路由和频谱分配模型。

is the entropy of the policy distribution,

controls the strength of the entropy regularizer,

and

is the learning rate,

is the discount factor. After cyclic and iterative training, the policy network and the value network are optimized to obtain a routing and spectrum allocation model.

Applying the embodiment of the present invention, by acquiring the service connection request and the frequency slot resource status of the current optical network, the service connection request includes the source node, the destination node and the spectrum width, according to the service connection request and the frequency of the current optical network. slot resource status, obtain multiple candidate paths between the source node and the destination node and the number of frequency slots that the service connection request needs to occupy on each candidate path; wherein, each candidate path It consists of one or more links. According to the candidate path and the corresponding link, a path link relationship matrix is established, and according to the state of each frequency slot on the candidate path, the The number of idle frequency slots, the number of idle frequency slots, the number of idle frequency slots, and the average size of idle frequency slots, and the set of available frequency slots is obtained according to the number of frequency slots to be occupied and the idle frequency slots, In the set of available frequency slots, obtain the starting position and size of the available frequency slot at the frontmost position on the frequency slot axis, and combine the number of idle frequency slots, the number of idle frequency slots, the The size of the set of available frequency slots, the starting position and size of the available frequency slots, and the average size of the idle frequency slots are concatenated to obtain a one-dimensional spectrum state distribution vector, and the service connection request, the The path link relationship matrix and the spectrum state distribution vector are input into the trained routing and spectrum allocation model to obtain the routing and spectrum allocation results corresponding to the service connection request. The present invention fully considers whether the frequency slot resource distribution of the link in the optical network is concentrated and whether the frequency slot resource state is not The problem of congestion reduces the network congestion rate of routing and spectrum allocation in the routing and spectrum allocation model, and improves the utilization of spectrum resources.

In an optional embodiment, please refer to FIG. 2, the step S20 includes S21-S22, and the details are as follows:

S21. According to the service connection request and the frequency slot resource state of the current optical network, calculate a plurality of candidate paths between the source node and the destination node by the shortest path method;

In the embodiment of the present application, multiple candidate paths between the source node and the destination node are calculated by the shortest path method. The shortest path method is specifically the Dijkstra method, which starts from a certain node and follows a directed graph with weights. Among the paths taken by an edge to reach another node, the path with the smallest sum of weights on each edge is called the shortest path.

S22. Calculate the number of frequency slots to be occupied by the service connection request on each candidate path according to the distance of each candidate path and the spectrum width.

In the embodiment of the present application, due to the different distances from the source node to the destination node for different service connection requests, the spectrum width requirements are different, and the number of frequency slots to be occupied by the calculation is also different. Specifically, the distance of each candidate path is compared with a preset distance interval to obtain a distance interval corresponding to each candidate path. For the same distance interval, the number of frequency slots is calculated by the ratio of the spectrum width to the spectrum fixed bandwidth.

In an optional embodiment, please refer to FIG. 3 , the step S22 includes S221-S222, and the details are as follows:

S221. Obtain all the links of each of the candidate paths, traverse the nodes of each of the links, obtain the distance of each of the links according to a preset node topology distance table, and sum the distances, get the distance of each of the candidate paths;

S222. Compare the distance of each of the candidate paths with a preset distance interval, and calculate the service connection request in each The number of frequency slots to be occupied on the candidate path; the method of calculating the number of frequency slots is:

表示为第

条所述候选路径的距离，

，

为一个频隙槽的固定带宽，大小为12.5Gbps，

为所述频谱宽度，

为所述业务连接请求在第

条所述候选路径上需占用的频隙数。in,

expressed as the

the distance of the candidate paths,

,

is the fixed bandwidth of a frequency slot, the size is 12.5Gbps,

is the spectral width,

request for the service connection in the

The number of frequency slots to be occupied on the candidate path.

In this embodiment of the present application, for the same spectrum width, the farther the distance is, the more frequency slots need to be occupied, and the closer the distance is, the fewer frequency slots need to be occupied. For service connection requests within the same distance range, the larger the spectrum width, the more frequency slots to be occupied, and the smaller the spectrum width, the less the number of frequency slots to be occupied. According to the distance interval and the spectrum width, the number of frequency slots that the service connection request needs to occupy on each of the candidate paths is calculated. As shown in Figure 4, there are three nodes in the optical network, node 1 is the source node, node 2 is the destination node, node 3 is the intermediate node, the total number K of candidate paths from the source node to the destination node is 2, and the corresponding The total number L of the above-mentioned links is 6, corresponding to the sequence numbers ① to ⑥. For the first candidate path, the distance of link ① is 500km and the spectrum width is 50Gbps, then the calculated number of frequency slots is 2; for the second candidate path, the distance of link ③ is 750km, and the distance of link ⑤ is 750km. is 750km, the second candidate path is 1500km, the spectrum width is 50Gbps, and the calculated number of frequency slots is 3.

In an optional embodiment, referring to FIG. 5 , the step S30 includes S31 to S33, and the details are as follows:

S31. Obtain the total number K of the candidate paths and the total number L of the links;

以及所述链路在每条所述候选路径k对应的链路编号

，其中，

，

，

表示所述候选路径k的第

条链路；S32. Obtain the number of the link in the total link

and the link number corresponding to each candidate path k of the link

,in,

,

,

represents the first path of the candidate path k

link;

表示候选路径

以及链路

，若链路

对应为候选路径

中的链路

，将所述路径链路关系矩阵的元素

设置为

；否则，将所述元素

设置为0。S33. Build a path link relationship matrix R with a size of K×L, wherein the elements of the path link relationship matrix

Indicates a candidate path

and link

, if the link

corresponding to the candidate path

link in

, the elements of the path link relationship matrix

Set as

; otherwise, the element

Set to 0.

In the embodiment of the present application, as shown in FIG. 4 , for the first candidate path, namely link ①, link ① is the first link of the first candidate path, and for the second candidate path, namely From link ③ to link ⑤, the link ③ and the link ⑤ are the first link and the second link of the second candidate path, respectively, so that the path link relationship matrix R is constructed as :

In an optional embodiment, referring to FIG. 6 , the step S40 includes S41-S42, and the details are as follows:

S41. Calculate the comprehensive state of each frequency slot on each candidate path according to the occupancy status of each frequency slot in each link on the candidate path; wherein, if the frequency slot is on the candidate path is occupied in one or more links on the candidate path, the comprehensive state of the frequency slot on the candidate path is occupied, and if the frequency slot is not occupied in the links on the candidate path, the the comprehensive state of the frequency slot on the candidate path is idle;

S42. According to the comprehensive state of each frequency slot on the candidate path, obtain the number of idle frequency slots, the number of idle frequency slots, and the average size of idle frequency slots on each of the candidate paths; wherein, the idle The frequency slot is composed of one or more consecutive idle frequency slots, and the average size of the idle frequency slot is the ratio of the number of idle frequency slots to the number of idle frequency slots.

In the embodiment of the present application, referring to FIG. 7 , on the frequency slot axis of the link ① of the first candidate path, there are 6 empty frequency slots (blank squares) and 4 occupied frequency slots (gray squares). Because the first candidate path has only link ①, the comprehensive state of each frequency slot of the first candidate path is the state of each frequency slot on link ①. Specifically, it includes 3 idle frequency slots, the first idle frequency slot includes 2 idle frequency slots, the second idle frequency slot includes 3 idle frequency slots, and the third idle frequency slot includes 1 idle frequency slot. frequency slot, the average size of the idle frequency slot slot is 2. On the frequency slot axis of link ③ of the second candidate path, there are 6 empty frequency slots and 4 occupied frequency slots; on the frequency slot axis of link ⑤, there are 7 empty frequency slots and 3 occupied frequency slots . Since the frequency slots labeled 7, 8 and 9 on the frequency slot axis are occupied in link ③ and link ⑤, the second candidate path labeled 0, 1, 2, 4 and 5 The frequency slot is in the idle state, and the rest are in the occupied state. The number of idle frequency slots is 5, and the number of idle frequency slots is 2. The first idle frequency slot includes 3 idle frequency slots, and the second idle frequency slot includes 2 slots. Idle frequency slot, the average size of idle frequency slot slot is 2.5.

In an optional embodiment, please refer to FIG. 8 , the step S50 includes S51 to S53, and the details are as follows:

S51. Compare the maximum value of the number of idle frequency slots corresponding to each of the idle frequency slots on each of the candidate paths with the number of frequency slots to be occupied;

S52. If the maximum value of the number of idle frequency slots is greater than or equal to the number of frequency slots to be occupied, determine the idle frequency slot as an available frequency slot that satisfies the service connection request;

S53. Obtain a set of available frequency slots that satisfy the service connection request according to the available frequency slots.

In the embodiment of the present application, for the first candidate path, since the number of frequency slots to be occupied by the service connection request is 2, the number of idle frequency slots of the first idle frequency slot and the second idle frequency slot is 2 respectively and 3, therefore, both the first idle frequency slot and the second idle frequency slot are determined as available frequency slots, and these two idle frequency slots constitute a set of available frequency slots. For the second candidate path, since the number of frequency slots to be occupied by the service connection request is 3, and the number of idle frequency slots in the first idle frequency slot is 3, the first idle frequency slot is determined as available. frequency slot.

In an optional embodiment, referring to FIG. 9 , the step S80 includes S81 to S83, and the details are as follows:

和

，将所述路径链路关系矩阵K×L转换为一维行向量M；其中，所述一维行向量M的行维度为1，列维度为K×L；S81. Convert the source node and the destination node in the service connection request into two one-dimensional row vectors

and

, convert the path link relationship matrix K×L into a one-dimensional row vector M; wherein, the row dimension of the one-dimensional row vector M is 1, and the column dimension is K×L;

和

、所述一维行向量M、以及所述频谱状态分布向量进行向量拼接，得到状态向量S；S82. Convert the one-dimensional row vector

and

, the one-dimensional row vector M and the spectral state distribution vector carry out vector splicing to obtain a state vector S;

S83. Input the state vector S into the trained routing and spectrum allocation model to obtain routing and spectrum allocation results.

和

，一维行向量上的每个元素代表光网络的一个节点，源节点和目的节点为1，其他节点为0。也即，

，

。将所述路径链路关系矩阵K×L转换为一维行向量

。对于第一条候选路径，所述频谱状态分布向量

。对于第二条候选路径，所述频谱状态分布向量

。将所述向量

、

、

、

和

拼接，得到状态向量S。In the embodiment of the present application, the source node and the destination node are respectively modeled as a one-dimensional row vector

and

, each element on the one-dimensional row vector represents a node of the optical network, the source node and the destination node are 1, and the other nodes are 0. That is,

,

. Convert the path link relationship matrix K×L into a one-dimensional row vector

. For the first candidate path, the spectral state distribution vector

. For the second candidate path, the spectral state distribution vector

. the vector

,

,

,

and

Splicing to get the state vector S.

Corresponding to the above method embodiments, please refer to FIG. 10 , an embodiment of the present invention provides a routing and spectrum allocation device 9 based on deep reinforcement learning, including:

a request obtaining module 91, configured to obtain a service connection request and the frequency slot resource state of the current optical network, where the service connection request includes a source node, a destination node and a spectrum width;

The path obtaining module 92 is configured to obtain, according to the service connection request and the frequency slot resource state of the current optical network, a plurality of candidate paths between the source node and the destination node and the service connection request at each time. The number of frequency slots to be occupied on the candidate paths; wherein, each candidate path consists of one or more links;

a matrix establishment module 93, configured to establish a path link relationship matrix according to the candidate path and the corresponding link;

A frequency slot calculation module 94, configured to calculate the number of idle frequency slots, idle frequency slots, idle frequency slots, and idle frequency slots on each candidate path according to the state of each frequency slot on the candidate path the average size of the slot;

a frequency slot obtaining module 95, configured to obtain a set of available frequency slots according to the number of frequency slots to be occupied and the idle frequency slots;

a position obtaining module 96, configured to obtain, in the set of available frequency slots, the starting position and size of the available frequency slot that is at the frontmost position on the frequency slot axis;

A vector obtaining module 97, configured to obtain the number of free frequency slots, the number of free frequency slots, the size of the set of available frequency slots, the starting position and size of the available frequency slots, and the free The average size of the frequency slot is spliced to obtain a one-dimensional spectrum state distribution vector;

The result obtaining module 98 is configured to input the service connection request, the path link relationship matrix, and the spectrum state distribution vector into the trained routing and spectrum allocation model, and obtain the route and the corresponding service connection request. Spectrum allocation results.

Optionally, please refer to FIG. 11 , the path obtaining module 92 includes:

A path calculation unit 922, configured to calculate a plurality of candidate paths between the source node and the destination node by the shortest path method according to the service connection request and the frequency slot resource state of the current optical network;

The frequency slot number calculation unit 924 is configured to calculate, according to the distance of each candidate path and the spectrum width, the number of frequency slots to be occupied by the service connection request on each candidate path.

Optionally, please refer to FIG. 12 , the frequency slot number calculation unit 924 includes:

The link obtaining unit 926 is configured to obtain all the links of each of the candidate paths, traverse the nodes of each of the links, obtain the distance of each of the links according to the preset node topology distance table, and set the The distances are summed to obtain the distance of each of the candidate paths;

A distance dividing unit 928, configured to divide the distance of each candidate path into a preset number of distance intervals, and calculate the service connection request on each candidate path according to the distance interval and the spectrum width The number of frequency slots to be occupied.

Optionally, please refer to FIG. 13, the matrix establishment module 93 includes:

a quantity obtaining unit 932, configured to obtain the total quantity K of the candidate paths and the total quantity L of the links;

以及所述链路在每条所述候选路径k对应的链路编号

，其中，

，

，

表示所述候选路径k的第

条链路；A number acquiring unit 934, configured to acquire the number of the link in the total link

and the link number corresponding to each candidate path k of the link

,in,

,

,

represents the first path of the candidate path k

link;

表示候选路径

以及链路

，若链路

对应为候选路径

中的链路

，将所述路径链路关系矩阵的元素

设置为

；否则，将所述元素

设置为0。A matrix construction unit 936, configured to construct a path link relationship matrix R with a size of K×L, wherein the elements of the path link relationship matrix

Indicates a candidate path

and link

, if the link

corresponding to the candidate path

link in

, the elements of the path link relationship matrix

Set as

; otherwise, the element

Set to 0.

Optionally, please refer to FIG. 14 , the frequency slot calculation module 94 includes:

The state calculation unit 942 is configured to calculate the comprehensive state of each frequency slot on each candidate path according to the occupancy state of each frequency slot in each link on the candidate path; wherein, if the frequency slot If one or more links on the candidate path are occupied, the comprehensive state of the frequency slot on the candidate path is occupied, if the frequency slot is not used in any link on the candidate path is occupied, the comprehensive state of the frequency slot on the candidate path is idle;

The idle frequency slot obtaining unit 944 is configured to obtain the number of idle frequency slots, the number of idle frequency slots, and the average of idle frequency slots on each candidate path according to the comprehensive state of each frequency slot on the candidate path size; wherein, the idle frequency slot is composed of one or more consecutive idle frequency slots, and the average size of the idle frequency slot is the ratio of the number of idle frequency slots to the number of idle frequency slots.

Optionally, please refer to FIG. 15 , the frequency slot obtaining module 95 includes:

A frequency slot number comparison unit 952, configured to compare the maximum value of the number of idle frequency slots corresponding to each of the idle frequency slots on each of the candidate paths with the number of frequency slots to be occupied;

A frequency slot determination unit 954, configured to determine the idle frequency slot as an available frequency slot satisfying the service connection request if the maximum value of the number of idle frequency slots is greater than or equal to the number of frequency slots to be occupied groove;

The frequency slot set obtaining unit 956 is configured to obtain, according to the available frequency slots, a set of available frequency slots that satisfy the service connection request.

Optionally, please refer to FIG. 16, the result obtaining module 98 includes:

和

，将所述路径链路关系矩阵K×L转换为一维行向量M；其中，所述一维行向量M的行维度为1，列维度为K×L；A vector conversion unit 982, configured to convert the source node and the destination node in the service connection request into two one-dimensional row vectors

and

, convert the path link relationship matrix K×L into a one-dimensional row vector M; wherein, the row dimension of the one-dimensional row vector M is 1, and the column dimension is K×L;

和

、所述一维行向量M、以及所述频谱状态分布向量进行向量拼接，得到状态向量S；The vector splicing unit 984 is used to combine the one-dimensional row vector

and

, the one-dimensional row vector M and the spectral state distribution vector carry out vector splicing to obtain a state vector S;

The state input unit 986 is configured to input the state vector S into the trained routing and spectrum allocation model to obtain the routing and spectrum allocation results.

Applying the embodiment of the present invention, by acquiring the service connection request and the frequency slot resource status of the current optical network, the service connection request includes the source node, the destination node and the spectrum width, according to the service connection request and the frequency of the current optical network. slot resource status, obtain multiple candidate paths between the source node and the destination node and the number of frequency slots that the service connection request needs to occupy on each candidate path; wherein, each candidate path It consists of one or more links. According to the candidate path and the corresponding link, a path link relationship matrix is established, and according to the state of each frequency slot on the candidate path, the The number of idle frequency slots, the number of idle frequency slots, the number of idle frequency slots, and the average size of idle frequency slots, and the set of available frequency slots is obtained according to the number of frequency slots to be occupied and the idle frequency slots, In the set of available frequency slots, obtain the starting position and size of the available frequency slot at the frontmost position on the frequency slot axis, and combine the number of idle frequency slots, the number of idle frequency slots, the The size of the set of available frequency slots, the starting position and size of the available frequency slots, and the average size of the idle frequency slots are concatenated to obtain a one-dimensional spectrum state distribution vector, and the service connection request, the The path link relationship matrix and the spectrum state distribution vector are input into the trained routing and spectrum allocation model to obtain the routing and spectrum allocation results corresponding to the service connection request. The present invention fully considers whether the frequency slot resource distribution of the link in the optical network is concentrated and whether the frequency slot resource state is not Congestion reduces the network congestion rate of routing and spectrum allocation model for routing and spectrum allocation, and improves the utilization rate of spectrum resources.

The present application also provides an electronic device, comprising: a processor and a memory; wherein, the memory stores a computer program, and the computer program is adapted to be loaded by the processor and execute the method steps of the above embodiments.

The present application further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the method steps of the foregoing embodiments.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the invention patent. It should be pointed out that, for those skilled in the art, some modifications and improvements can be made without departing from the concept of the present invention, and the present invention is also intended to include these modifications and modifications.

Claims (10)

1. A method of routing and spectrum allocation, comprising:

acquiring a service connection request and a frequency slot resource state of a current optical network, wherein the service connection request comprises a source node, a destination node and a spectrum width;

acquiring a plurality of candidate paths between the source node and the destination node and the number of frequency slots required to be occupied by the service connection request on each candidate path according to the service connection request and the frequency slot resource state of the current optical network; wherein each of the candidate paths consists of one or more links;

establishing a path link relation matrix according to the candidate paths and the corresponding links;

according to the state of each frequency slot on the candidate paths, calculating the number of idle frequency slots, the number of idle frequency slots and the average size of the idle frequency slots on each candidate path;

obtaining an available frequency slot set according to the frequency slot number to be occupied and the idle frequency slot;

obtaining the initial position and the size of the available frequency slot which is positioned most front on the frequency slot axis in the available frequency slot set;

splicing the number of the idle frequency slots, the number of the idle frequency slot slots, the size of the available frequency slot set, the initial position and the size of the available frequency slot and the average size of the idle frequency slot to obtain a one-dimensional spectrum state distribution vector;

and inputting the service connection request, the path link relation matrix and the spectrum state distribution vector into a trained routing and spectrum allocation model to obtain a routing and spectrum allocation result corresponding to the service connection request.

2. The method according to claim 1, wherein the obtaining a plurality of candidate paths between the source node and the destination node and the number of frequency slots that the service connection request needs to occupy on each of the candidate paths according to the service connection request and the frequency slot resource status of the current optical network comprises:

calculating a plurality of candidate paths between a source node and a destination node by a shortest path method according to the service connection request and the frequency slot resource state of the current optical network;

and calculating the frequency slot number of the service connection request on each candidate path according to the distance of each candidate path and the spectrum width.

3. The method according to claim 2, wherein the calculating the number of frequency slots required to be occupied by the service connection request on each of the candidate paths according to the distance of each of the candidate paths and the spectrum width comprises:

acquiring all links of each candidate path, traversing nodes of each link, acquiring the distance of each link according to a preset node topology distance table, and summing the distances to obtain the distance of each candidate path;

comparing the distance of each candidate path with a preset distance interval, and calculating the frequency slot number of the service connection request on each candidate path according to the distance interval of the distance of each candidate path and the frequency spectrum width; the frequency slot number is calculated in the following way:

wherein,

is shown as

The distance of the candidate path is determined,

k represents a total number of the plurality of candidate paths,

is a fixed bandwidth of one slot, with a size of 12.5Gbps,

for the said width of the frequency spectrum,

requesting for said service connection at

And the number of frequency slots required to be occupied on the candidate path.

4. The method of claim 1, wherein said constructing a path-link relationship matrix from said candidate paths and corresponding said links comprises:

acquiring the total number K of the candidate paths and the total number L of the links;

obtaining the number of the link in the total link

And the link number corresponding to each candidate path k of the link

Wherein

，

，

represents the candidate path k

A link;

constructing a path link relation matrix R with the size of KxL, wherein elements of the path link relation matrix

Representing candidate paths

And a link

If the link is

Corresponds to a candidate path

In (1)

Elements of the path-link relation matrix

Is arranged as

(ii) a Otherwise, the element is processed

Is set to 0.

5. The method according to claim 1, wherein said calculating the number of free frequency slots, the average size of free frequency slots, and the average size of free frequency slots on each of the candidate paths according to the state of each frequency slot on the candidate paths comprises:

calculating the comprehensive state of each frequency slot on each candidate path according to the occupation state of each frequency slot in each link on the candidate path; wherein if the frequency slot is occupied in one or more links on the candidate path, the integrated state of the frequency slot on the candidate path is occupied, and if the frequency slot is not occupied in all links on the candidate path, the integrated state of the frequency slot on the candidate path is idle;

acquiring the number of idle frequency slots, the number of idle frequency slot slots and the average size of the idle frequency slot slots on each candidate path according to the comprehensive state of each frequency slot on the candidate path; the idle frequency slot is composed of one or more continuous idle frequency slots, and the average size of the idle frequency slot is the ratio of the number of the idle frequency slots to the number of the idle frequency slots.

6. The method according to claim 1, wherein said obtaining a set of available frequency slot slots according to the number of frequency slots to be occupied and the idle frequency slot slots comprises:

comparing the maximum value of the number of idle frequency slots corresponding to each idle frequency slot on each candidate path with the number of frequency slots to be occupied;

if the maximum value of the idle frequency slot number is larger than or equal to the frequency slot number to be occupied, determining the idle frequency slot as an available frequency slot meeting the service connection request;

and obtaining an available frequency slot set meeting the service connection request according to the available frequency slot.

7. The method according to claim 1, wherein the inputting the service connection request, the path link relation matrix, and the spectrum state distribution vector into a trained routing and spectrum allocation model to obtain a routing and spectrum allocation result corresponding to the service connection request comprises:

converting the source node and the destination node in the service connection request into two one-dimensional row vectors

And

converting the path link relation matrix into a one-dimensional row vector M; the one-dimensional row vector M has a row dimension of 1 and a column dimension of K × L; k is the total number of the candidate paths, and L is the total number of the links;

the one-dimensional row vector is processed

And

vector splicing is carried out on the one-dimensional row vector M and the frequency spectrum state distribution vector to obtain a state vector S;

and inputting the state vector S into a trained routing and spectrum allocation model to obtain a routing and spectrum allocation result.

8. A routing and spectrum allocation apparatus, comprising:

the request acquisition module is used for acquiring a service connection request and the frequency slot resource state of the current optical network, wherein the service connection request comprises a source node, a destination node and a spectrum width;

a path obtaining module, configured to obtain, according to the service connection request and a frequency slot resource state of the current optical network, multiple candidate paths between the source node and the destination node and a number of frequency slots that the service connection request needs to occupy on each candidate path; wherein each of the candidate paths consists of one or more links;

a matrix establishing module, configured to establish a path link relationship matrix according to the candidate path and the corresponding link;

the frequency slot calculation module is used for calculating the number of idle frequency slots, the average size of the idle frequency slots and the average size of the idle frequency slots on each candidate path according to the state of each frequency slot on each candidate path;

a frequency slot obtaining module, configured to obtain an available frequency slot set according to the number of frequency slots to be occupied and the idle frequency slots;

a position obtaining module, configured to obtain, in the available frequency slot set, a starting position and a size of an available frequency slot that is located most forward on a frequency slot axis;

a vector obtaining module, configured to splice the number of idle frequency slots, the size of the available frequency slot set, the initial position and size of the available frequency slot, and the average size of the idle frequency slots to obtain a one-dimensional spectrum state distribution vector;

and the result obtaining module is used for inputting the service connection request, the path link relation matrix and the spectrum state distribution vector into a trained routing and spectrum allocation model to obtain a routing and spectrum allocation result corresponding to the service connection request.

9. An electronic device, comprising: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the routing and spectrum allocation method according to any of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the routing and spectrum allocation method according to any one of claims 1 to 7.

Images