CN1112813C - Intelligent routing method in communication network - Google Patents
Intelligent routing method in communication network Download PDFInfo
- Publication number
- CN1112813C CN1112813C CN 99104078 CN99104078A CN1112813C CN 1112813 C CN1112813 C CN 1112813C CN 99104078 CN99104078 CN 99104078 CN 99104078 A CN99104078 A CN 99104078A CN 1112813 C CN1112813 C CN 1112813C
- Authority
- CN
- China
- Prior art keywords
- switch
- module
- adjoining
- load
- load value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Images
Landscapes
- Exchange Systems With Centralized Control (AREA)
- Telephonic Communication Services (AREA)
Abstract
The present invention relates to an intelligent routing method in a communication network. Under the condition of pre-holding network loads, a call is communicated, and the loads of the whole network (the using conditions of main lines on a link) are balanced. The present invention has global and systemic viewpoints, and the call communicating rate of service can be increased. Each program control exchanger comprises an interactive intelligent agent with autonomous decision, and exchanger nodes become more intelligent. The historical data (a neural network is used for prediction.) is fully used so as to cause route selection to be reasonable and autonomous. The management of the telecommunication network is developed.
Description
The present invention relates to a kind of route selection method of communication network.
The hierarchical organization communication network in past is telecommunication system especially, provides the required information of link of selecting by routing table.When running into tie link when busy, can only select in order by fixing alternate route, might be through repeatedly switching.The routing table that this static routing method places one's entire reliance upon and determined by historical situation and business demand, it is simple, reliable, but can not reflect real time business, can not adapt to the network condition of variation.
Because the randomness of communication network load can't guarantee in advance that certain part of network or network is in service area preferably, and can only add certain control when operation always.So people begin to consider dynamic routing algorithm." dynamically " shows the way by selection course is the dependence time, is likely (function of time) that depend on real-time status.Dynamic routing depends on network node (referring to switch) and constantly measures customer service, network state and constantly calculating, can avoid current or potential congested by utilizing the vacant ability of network, it and network load dynamic equilibrium are closely related, and can improve network performance.
The basic dynamic routing method of two classes is arranged: time-dependent method and state rely on method.The basic thought of time-dependent route is: at the different fact of traffic carrying capacity of different time sections every day, compile out the routing table of distinguishing on a time period in advance.It can make the traffic carrying capacity equilibrium to a certain extent, increases network utilization.But be the order of prearranging according to historical service conditions in essence.The basic thought that state relies on route is: which bar route decision selects according to the network current state.It can avoid the big link of traffic carrying capacity, and changing for the business of failing in advance to plan has adaptability preferably, but needs bigger administration overhead to come exchange message, more effectively to optimize the use of Internet resources.It can be divided into again isolated, concentrate with three kinds that distribute, depend on how the information of current network state obtains.In the isolated method, each node only has the state information of its outlet; In the concentric method, central network processor is the state information of all links in the collection network periodically; In the distribution, all nodes that relate to intercommunication to obtain current best optional route.
The typical distribution method is that (Real-Time Network Routing RTNR), was introduced into (the AT﹠amp of AT﹠T to the real-time network route in 1991; T) switching network.When not having direct route, it checks existing alternate route and loading condiction thereof, and the minimum load route is selected.Load overflows level six etc.: light load LL1, LL2, LL3, heavy load HL keeps TR and busy (busy).During trunk line on " catching " or " release " link, all to calculate loading condiction one time.If the trunk line number on the link surpasses a certain thresholding, then be considered to light load.During selection, ingress switch is sent out message and is given the purpose switch, requires to beam back the switch that the purpose switch is thought light load, and ingress switch and purpose switch think that all the link of light load is with selected.During selection, the peak load in the link of two ends is compared, find minimum load level, occupy grade if surpass the maximum of regulation, then this calling is rejected.When selecting the route of a specific call, ingress switch carries out cyclic search to the route bitmap list with same load state, and starting point is followed by the switch of its last calling of arriving the purpose switch of using, and this can reduce the time-delay of call setup.
The advantage of this method is not need other equipment and depend on it, and its route pattern calculates in exchange in real time, and flexibility and adaptivity are good.It has reduced running cost, because reduced routing tables a large amount of in the network switch, reduced the routing table of network design and prediction and calculated, reduced the operation of downloading management routes such as new routing information, reduced in the network management automatically routing again by expense, reduced degree of congestion.Yet,, increased network load because signaling information is a lot of alternately.
After this, occurred the new approaches based on distributed artificial intelligence (mainly being based on the artificial life of behavior) again, the process of getting food based on ant is modeled and studies, and this model can be represented the distribution of calling out between node.Artificial ant the node of selecting at random between move, and stay certain vestige, this vestige is the function of its Congestion Level SPCC of meeting in source node distance and way.They select the path according to the distribution at the vestige that each node imitated.The route of calling out between node is the function that the vestige of each intermediate node distributes.
The outstanding characteristics of distribution are, make network node that disposal ability all be arranged and traffic load is moved on to the place of bearing of having the ability in the network.This comes from distributed method " distribution " characteristic, thereby it is superior relatively, only its selection principle and existing similar, the systematic and not overall viewpoint of other dynamic routing method; Adaptivity, flexibility are also not enough, lack " intelligence ".
In general, " dynamically " of these methods is not enough far away, and they are the center with the local exchange switch, if its calling can be switched on just can, and used separately algorithm has determined their comparison procedure and corresponding selection the thereof.Though the artificial life method has its intrinsic advantage, the decision-making capability of each node is very limited, and the memory in past " vestige " is difficult to precise quantification, and this method is difficult to practical application at present.
The objective of the invention is for the balancing network load, improve the call completion rate.
The present invention, be that intelligent routing method in communication network is to realize like this, this communication network comprises stored-program control exchange and link therebetween, it is characterized in that it comprises that each stored-program control exchange of following content: a. utilizes artificial neural net to carry out next load value prediction constantly based on historical load data and current load value at interval with certain hour; B. when in the communication network call request being arranged, if having the ingress switch of call request and the direct route link of purpose inter-exchange to be taken, then ingress switch is passed to some adjoining switch in the optional route with call request signal; C. adjoining switch is made comparisons load value and predefined threshold value that latest computed goes out, less than, greater than use respectively " 0 " and " 1 " expression, then the result is returned to ingress switch, ingress switch is judged and is selected: a) when the load value that has only an adjoining switch during less than threshold value, then select this adjoining switch to make repeater exchanger; B) when the load value of some adjoining switch during, then refuse this call request all greater than threshold value; C) when the load value of some adjoining switch during all less than threshold value, then carrying out further repeater exchanger in these adjoining switch selects: at first, select the little and adjoining switch that vacant trunk line is many of load value to make repeater exchanger, if the load value height occurs, vacant trunk line is many, or the few situation of low, the vacant trunk line of load value, serve as last relatively foundation with vacant trunk line then, what vacant trunk line was few does not select.Thereby finish one by ingress switch to the relaying adjoining switch again to the one-level alternate route selection course of purpose switch.
In above-mentioned communication network route selection method, be used for the intelligent agent of stored-program control exchange for realizing this method, it comprises: perception/calling module, after being used for call out generating, react to going into to exhale for the calling of this switch; Modeling/training module, it receives information from perception/calling module, is used for week about, with the historical data of past 1 year internal loading, at the adjoining switch of different directions, neural network training again; The load prediction module, it receives information from modeling/training module, at the direction of each adjoining switch, with 2 hours be interval, according to the historical data of loading with work as preload and predict, and the result is preserved with neural net; Decision-making module, it receives information respectively from load prediction module and physical communication/Lexical Analysis Module, is used to be processed into and exhales or relay request; Action/respond module, it receives information from decision-making module, is used for response request; And, physical communication/Lexical Analysis Module, it receives information from action/respond module, is used to analyze different message, and carries out the transmission of message with the external world.
This method is under the situation of the load of holding network in advance, manage to allow to call out and be switched on, and make load (the main line operating position on the link) balance of the whole network, promptly make the success of this calling not influence the possibility that calling after this is switched on as far as possible, have an overall situation with the viewpoint of system, greatly improved the call completion rate, each switch all contains an intelligent agent of making decisions on one's own and can be mutual simultaneously.Strengthened " intelligence " of switch node, used for reference " history " knowledge (utilizing neural network prediction) simultaneously better, made Route Selection rationally, independently, made the management of communication network obtain great development.
The present invention is further illustrated below in conjunction with the embodiments of the invention accompanying drawing.
Fig. 1 is the simplification topology diagram of communication network of the present invention;
Fig. 2 is the structured flowchart that the present invention is arranged in the intelligent agent of stored-program control exchange.
Communication network involved in the present invention is simplified topological structure, and as shown in Figure 1, it comprises stored-program control exchange Si and link therebetween, and this intelligent routing method comprises following content:
A. each stored-program control exchange Si carries out next load value prediction constantly with artificial neural net according to the historical data and the current load value of load at interval with certain hour;
B. when in the communication network call request being arranged, if having the ingress switch of call request and the direct route link of purpose inter-exchange, when for example S1-S6 was occupied, then ingress switch was passed to adjoining switch S3 in the optional route, S4 with call request signal;
C. adjoining switch S3, load value that S4 goes out latest computed respectively and predefined threshold value are made comparisons, less than, greater than using " 0 " and " 1 " expression respectively, then the result is returned to ingress switch, ingress switch is judged and is selected:
A) when the load value that has only an adjoining switch during, then select this adjoining switch to make repeater exchanger less than threshold value;
B) when the load value of some adjoining switch during, then refuse this time call request all greater than threshold value;
C) when the load value of some adjoining switch during all less than threshold value, then in these some adjoining switch, carry out further repeater exchanger (S1-S3-S6, S1-S4-S6) select:
At first, select the little and adjoining switch that vacant trunk line is many of load value to make repeater exchanger, if the load value height occurs, vacant trunk line is many, or the few situation of low, the vacant trunk line of load value, serve as last relatively foundation with vacant trunk line then, what vacant trunk line was few does not select.
Thereby finish a routing procedure that makes a circulation to the one-level of purpose switch again to the relaying adjoining switch by ingress switch.
The Neural network prediction method that is adopted is as follows: be provided with n historical data, get its preceding m (m＜＜n) as the input of the algorithm that is adopted, its output compares with m+1 data, adjusts weights by the gradient descent direction of their variances, makes variance become little.After having trained network with n data, can predict that the rest may be inferred by analogy to n+1 data.The training of a neural net in return process of machine exists, and each switch is preserved this actual load situation, and usefulness when predicting for next time is upgraded routing table simultaneously.
In the intelligent routing method of above-mentioned communication network, the intelligent agent of related stored-program control exchange comprises:
Perception/calling module after being used for call out generating, is reacted to going into to exhale for the calling of this switch;
Modeling/training module, it receives information from perception/calling module, is used for week about, with the historical data of past 1 year internal loading, at the adjoining switch of different directions, neural network training again;
The load prediction module, it receives information from modeling/training module, at the direction of each adjoining switch, with 2 hours be interval, according to the historical data of loading with work as preload and predict, and the result is preserved with neural net;
Decision-making module, it receives information respectively from load prediction module and physical communication/Lexical Analysis Module, is used to be processed into and exhales or relay request;
Action/respond module, it receives information from decision-making module, is used for response request; And
Physical communication/Lexical Analysis Module, it receives information from action/respond module, is used to analyze different message, and carries out the transmission of message with the external world.
Described intelligent agent contains three class processes:
1. week about, with the historical data in past 1 year, at the adjacent node direction of difference, neural network training again.This class process all has several at each node, and quantity and adjacent nodal point number are together.The network that trains is for 2 usefulness
2. to each adjacent node direction, with 2 hours be the interval, predict with neural net according to the historical data of load with when preload, result and critical value relatively, less than, greater than using " 0 " and " 1 " to represent respectively, and the result is preserved.This class process all has several at each node, and quantity and adjacent nodal point number are together.The result here is for 3 usefulness.
3. be processed into and exhale or relay request.This class process number and nodal point number are together.1 and 2 will be in the real-time load result who uses here next time.
The core content that described intelligent agent is handled the process of calling or relay request is:
while (message) // wait for the arrival of news switch (message) case ' call ': // message is for calling out inspect (route table); // inspection routing table if (direct route); // if having direct route, link that vacant inform is arranged; // go into to exhale successfully, send junction call break; If (! Route) // if no optional route failure; // call failure＜!--SIPO<dP n="5">--<dp n="d5"/>Break; If (! Direct route) // if no direct route { request; // neighborhood of nodes in the optional route is passed in request, and make respective record break; Case ' response ': // message is for replying { if (response=' 0 ') // if having in abutting connection with link meet the demands { if (response＞1) // if there is a more than adjacent link to meet the demands, then two compare that compare; // selection respective link inform; // then this link is selected, and send and call out break; If (response=' 1 ') // meet the demands, then call failure continue as if no adjacent link; Case ' relay request ': // message is relay request read (load); // obtain the load condition if (load) of respective direction route // if allow relaying, passback response message { confirm (the load Bai Fenbi ﹠amp that load percentage//respective direction forecasting institute gets by database; Spare links); // and vacant number of links break; Else continue; Case ' relay call ': // message is junction call, issues next the node inform in the optional route; // its processing calling together＜!--SIPO<dP n="6">--<dp n="d6"/>Break; Default:break; Write (record); // modification database }
Wherein, inspect, compare, read and write are each node internal actions.
Message is divided into the message of reception and the message of sending, and the former is divided into calling, relay request, junction call and replys; The latter is divided into and replys and junction call (message that is reception is again the message of sending).Relay request promptly asks to call out the message through this node relaying; The i.e. calling of junction call itself.
If receive the relay request of some adjacent nodes simultaneously, then do further to consult with game theory according to the emergency priority parameter.
Send out, receive message and use three kinds of basic primitive:
*By routing table, send out request to all neighborhood of nodes, inquiry has not idle link: request, i.e. relay request
*That returns replys: confirm
*The transmission of calling out: inform, i.e. junction call
Claims (2)
1. intelligent routing method in communication network, this communication network comprises stored-program control exchange and link therebetween, it is characterized in that, it comprises following content:
A. each stored-program control exchange utilizes artificial neural net to carry out next load value prediction constantly based on historical load data and current load value at interval with certain hour;
B. when in the communication network call request being arranged, if having the ingress switch of call request and the direct route link of purpose inter-exchange to be taken, then ingress switch is passed to some adjoining switch in the optional route with call request signal;
C. adjoining switch is made comparisons load value and predefined threshold value that latest computed goes out, less than, greater than use respectively " 0 " and " 1 " expression, then the result is returned to ingress switch, ingress switch is judged and is selected:
A) when the load value that has only an adjoining switch during, then select this adjoining switch to make repeater exchanger less than threshold value;
B) when the load value of some adjoining switch during, then refuse this call request all greater than threshold value;
C) when the load value of some adjoining switch during, then in these adjoining switch, carry out further repeater exchanger and select all less than threshold value:
At first, select the little and adjoining switch that vacant trunk line is many of load value to make repeater exchanger, if the load value height occurs, vacant trunk line is many, or the few situation of low, the vacant trunk line of load value, serve as last relatively foundation with vacant trunk line then, what vacant trunk line was few does not select
Thereby finish one by ingress switch to the relaying adjoining switch again to the one-level alternate route selection course of purpose switch.
2. intelligent agent that is used for stored-program control exchange is characterized in that it comprises:
Perception/calling module after being used for call out generating, is reacted to going into to exhale for the calling of this switch;
Modeling/training module, it receives information from perception/calling module, is used for week about, with the historical data of past 1 year internal loading, at the adjoining switch of different directions, neural network training again;
The load prediction module, it receives information from modeling/training module, at the direction of each adjoining switch, with 2 hours be interval, according to the historical data of loading with work as preload and predict, and the result is preserved with neural net;
Decision-making module, it receives information respectively from load prediction module and physical communication/Lexical Analysis Module, is used to be processed into and exhales or relay request;
Action/respond module, it receives information from decision-making module, is used for response request; And
Physical communication/Lexical Analysis Module, it receives information from action/respond module, is used to analyze different message, and carries out the transmission of message with the external world.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 99104078 CN1112813C (en) | 1999-03-19 | 1999-03-19 | Intelligent routing method in communication network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 99104078 CN1112813C (en) | 1999-03-19 | 1999-03-19 | Intelligent routing method in communication network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1268006A CN1268006A (en) | 2000-09-27 |
| CN1112813C true CN1112813C (en) | 2003-06-25 |
Family
ID=5271528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 99104078 Expired - Lifetime CN1112813C (en) | 1999-03-19 | 1999-03-19 | Intelligent routing method in communication network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1112813C (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2357785A1 (en) | 2001-09-14 | 2003-03-14 | Alcatel Canada Inc. | Intelligent routing for effective utilization of network signaling resources |
| CN1302639C (en) * | 2003-09-18 | 2007-02-28 | 上海贝尔阿尔卡特股份有限公司 | Non-level select method for preventing circulation channel selection and network congestion dispersion |
| CN103260207B (en) * | 2003-12-23 | 2016-06-01 | 艾利森电话股份有限公司 | Predictability AD-HOC |
| CN100459795C (en) * | 2005-06-15 | 2009-02-04 | 华为技术有限公司 | Method for selecting service node according to load status |
| CN101288089B (en) * | 2005-07-28 | 2014-08-20 | 西门子公司 | Load prediction based on-line and off-line training of neural networks |
| CN101378578B (en) * | 2007-08-30 | 2014-12-03 | 联想(北京)有限公司 | Method for balancing load of multi-hop wireless network based on relay |
| CN105389619B (en) * | 2014-08-31 | 2021-08-03 | 通用电气公司 | Method and system for improving connectivity within a healthcare ecosystem |
| US20180359172A1 (en) * | 2017-06-12 | 2018-12-13 | Juniper Networks, Inc. | Network path prediction and selection using machine learning |
| CN114615698B (en) * | 2020-12-09 | 2023-07-18 | 中国移动通信集团四川有限公司 | IBCF interworking gateway load adjustment method and device |
-
1999
- 1999-03-19 CN CN 99104078 patent/CN1112813C/en not_active Expired - Lifetime
Also Published As
| Publication number | Publication date |
|---|---|
| CN1268006A (en) | 2000-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP2971580B2 (en) | Communication network management | |
| Kelly | Network routing | |
| Lin et al. | Integrated shop floor control using autonomous agents | |
| Monnier et al. | A genetic algorithm for scheduling tasks in a real-time distributed system | |
| Yang et al. | Online dispatching and routing model for emergency vehicles with area coverage constraints | |
| GB2303996A (en) | Communications arrangement and method | |
| CN109872535A (en) | A kind of current prediction technique of wisdom traffic, device and server | |
| CN1112813C (en) | Intelligent routing method in communication network | |
| CN102047226A (en) | Distributed service framework | |
| Alshamsi et al. | Multiagent self-organization for a taxi dispatch system | |
| Maciejewski12 et al. | A microscopic simulation approach for optimization of taxi services | |
| Lipperts et al. | Mobile agents in telecommunications networks-a simulative approach to load balancing | |
| Lamma et al. | A distributed constraint-based scheduler | |
| JP3248692B2 (en) | Adaptive route selection control method | |
| Berman | Repositioning of two distinguishable service vehicles on networks | |
| CN100463468C (en) | System and method for measuring traffic volume distributed among areas | |
| CN101959310B (en) | Method, equipment and system for scheduling transmission circuit resources | |
| Grivas et al. | Agent technology in load balancing for network applications | |
| Darko et al. | A Proactive Dynamic-Distributed Constraint Optimization Framework for Unmanned Aerial and Ground Vehicles in Traffic Incident Management | |
| CN111259261B (en) | High-speed rail driving network collaborative alarm optimization method based on state transition prediction | |
| CN1506860A (en) | Resource locating system under network environment | |
| CN1268852A (en) | Intelligent main set of programme-controlled exchange | |
| Nonas | Optimising a rule based agent using a genetic algorithm | |
| Lai et al. | On dispatching unequally capable service technicians | |
| CN117155847A (en) | Efficient power-calculation routing method and system based on power-calculation network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C56 | Change in the name or address of the patentee |
Owner name: BELL CO.,LTD., SHANGHAI Free format text: FORMER NAME: BELL TELEPHONE EQUIPMENT CO., LTD., SHANGHAI Owner name: BEIER AERKATE CO., LTD., SHANGHAI Free format text: FORMER NAME: BELL CO.,LTD., SHANGHAI Owner name: SHANGHAI ALCATEL-LUCENT CO., LTD. Free format text: FORMER NAME: BEIER AERKATE CO., LTD., SHANGHAI |
|
| CP01 | Change in the name or title of a patent holder |
Address after: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee after: ALCATEL-LUCENT SHANGHAI BELL Co.,Ltd. Address before: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee before: Shanghai Bell Alcatel Co.,Ltd. Address after: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee after: Shanghai Bell Alcatel Co.,Ltd. Address before: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee before: Shanghai Bell Co.,Ltd. Address after: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee after: Shanghai Bell Co.,Ltd. Address before: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee before: SHANGHAI BELL TELEPHONE EQUIPM |
|
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee after: NOKIA SHANGHAI BELL Co.,Ltd. Address before: 201206 No. 188-189 Baer Road, Jinqiao, Shanghai, Pudong Patentee before: ALCATEL-LUCENT SHANGHAI BELL Co.,Ltd. |
|
| CX01 | Expiry of patent term | ||
| CX01 | Expiry of patent term |
Granted publication date: 20030625 |