CN115834371A - Space-ground converged network cross-domain SFC deployment method based on hybrid state synchronous DRL - Google Patents
Space-ground converged network cross-domain SFC deployment method based on hybrid state synchronous DRL Download PDFInfo
- Publication number
- CN115834371A CN115834371A CN202211457628.5A CN202211457628A CN115834371A CN 115834371 A CN115834371 A CN 115834371A CN 202211457628 A CN202211457628 A CN 202211457628A CN 115834371 A CN115834371 A CN 115834371A
- Authority
- CN
- China
- Prior art keywords
- network
- actor
- domain
- deployment
- deployed
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000001360 synchronised effect Effects 0.000 title claims abstract description 16
- 230000004927 fusion Effects 0.000 claims abstract description 17
- 230000006870 function Effects 0.000 claims description 29
- 238000012549 training Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000005540 biological transmission Effects 0.000 claims description 6
- 230000006855 networking Effects 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 2
- 238000004808 supercritical fluid chromatography Methods 0.000 claims 8
- 238000013507 mapping Methods 0.000 description 3
- 230000002787 reinforcement Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Images
Abstract
The invention discloses a cross-domain SFC deployment method of a space-ground fusion network based on hybrid state synchronization DRL (data logging language), which can effectively reduce huge expenses caused by real-time synchronization of the global network state of the space-ground fusion network through key parameter synchronization and is suitable for a satellite control node with limited resources. On the other hand, by virtue of the synchronization of the state information of the global network and the idea of digital twin, the virtual copy Actor-Critic network of the A3C has rich and sufficiently real sample sets, and the success rate of SFC deployment of the global Actor-Critic network and the real copy Actor-Critic network can be further improved; meanwhile, the actual copy Actor-Critic network of the A3C makes an optimal decision on cross-domain SFC deployment of the world fusion network based on real-time synchronous network key state parameters, and supports world fusion network services.
Description
Technical Field
The invention belongs to the technical field of network control, and particularly relates to a space-ground convergence network cross-domain SFC deployment method based on hybrid state synchronous DRL.
Background
The heaven and earth converged network is formed by fusing a satellite and earth network by relying on a ground network and expanding a satellite communication network, has the advantages of high bearing capacity, wide coverage range, flexible networking mode and the like, and is a key technology of the next generation of information network. However, the characteristics of large space-time scale, strong topology dynamics, heterogeneous links and the like of the space-ground converged network bring huge challenges to the network service support capability.
Deep Reinforcement Learning (DRL) is an effective decision optimization method, and has wide application in the field of network control. By introducing the intelligent plane, the DRL carries out iterative optimization on self decision based on the perceived network state and issues the decision to the control plane, thereby realizing the optimal control of the network. In particular, the cross-domain service function chain deployment (SFC) of the world convergence network can be modeled as a Markov Decision Process (MDP), and the Bellman equation is iteratively solved by the DRL to support the world convergence network service. However, the traditional DRL method faces the problems of high network state space dimension, high data set acquisition overhead, difficult cross-domain collaboration and the like.
In order to realize the optimal deployment of the SFC, the intelligent plane needs to acquire the global network state information in real time, and the complexity of the DRL algorithm is increased while huge communication overhead is brought to the world convergence network. On the other hand, in a world-wide converged network with limited resources, the acquisition of network state data is limited, so that the sample set is small in scale and difficult to support the training of the DRL.
Disclosure of Invention
In view of this, the invention provides a space-ground convergence network cross-domain SFC deployment method based on a hybrid state synchronous DRL based on an asynchronous dominant Actor-critical algorithm (A3C) in deep reinforcement learning. The A3C has a global Actor-criticic network and a plurality of copy Actor-criticic networks, and can asynchronously and parallelly realize the updating of the neural network. Specifically, the invention designs a hybrid network state synchronization mechanism, which is used for synchronizing key parameters of network states of all domains of a world fusion network in real time, training a partial replica Actor-critical network and a global Actor-critical network, synchronizing global network state information of the world fusion network when the network is idle, generating the key parameters through digital twinning, and training other replica Actor-critical networks and global Actor-critical networks, thereby realizing cross-domain SFC deployment.
A space-ground fusion network cross-domain SFC deployment method based on hybrid state synchronous DRL adopts an A3C network for training, wherein: one Actor-critical network is used as a global Actor-critical network theta G P Actor-critical networks form a real copy Actor-critical network setQ Actor-critical networks form a virtual copy Actor-critical network setP and Q are not less than 1;
determining a required heaven-earth converged network key parameter vector when deploying the ith virtual network function VNF of the heaven-earth converged networkAnd networking network state vector s of the world converged network i ;
For training optimization of the A3C network, the following two processes are carried out simultaneously:
the first process is as follows: for reality copy Actor-criticc networkIn the i-th virtual network function VNF v i When deployment is carried out, key parameters of the network are fused according to the real-time synchronization of the heaven and the earthTo obtain s i Input ofGets v i Deployment decision a i And sends it down to the control plane; control ofPlane according to a i V is to be i Deployment to respective domains according to v i Get a reward based on the deployed results of (a) i Calculating a policy gradient, updating in sequenceThe Critic network and the Actor network are used until all VNFs are deployed; when the SFC is deployed once, the deployment is updatedThe gradient of (a) is uploaded to a global Actor-critical network Θ G And update the theta sequentially G Finally copying the parameters of the Actor network and the critical network to a real copy Actor-critical network
The second process is as follows: for a virtual copy Actor-critical network, the network isIn the i-th virtual network function VNF v i When deployment is carried out, key parameters are calculated according to the global network state of intelligent plane synchronizationAnd obtaining s i Input ofGets v i Deployment decision a i In the world-wide integration network simulation software, a is executed based on the network state of each domain i V is to be i Deployment to corresponding Domain, v i Simulating the execution situation in the domain to obtain the return, and obtaining the return based on the return and a i Calculating a policy gradient, updating in sequenceThe Critic network and the Actor network until all VNFs are deployed; when it is finishedAfter one SFC deployment, updatingThe gradient of (a) is uploaded to a global Actor-critical network Θ G And update the theta sequentially G Finally copying the parameters of the Actor network and the critical network to a real copy Actor-critical network
And completing the deployment of the SFC according to the first process and the second process.
networking network state vector s of the world convergence network i Comprises the following steps:
the method comprises the following steps that a set of N network domains of a world fusion network is set to be D, a set of K inter-domain links is set to be L, a set of I virtual network functions VNF contained in the SFC is set to be V, and a set of J resource types is set to be R; the interval 1. Ltoreq. I'. Ltoreq.I comprises I, the ith VNF v i An authorized domain that e V can be deployed isFor j resource type r j E.g. the requirement of R is alpha i,j The bandwidth required for cross-domain transmission is w i Maximum time delay of the whole SFC is tau + (ii) a N-th field d n E.g. D resource r j The balance ofIt is to VNF v i Has a processing delay ofThe kth inter-domain link l k The bandwidth of E L is gamma k Time delay of τ l,k (ii) a Wherein when d n When it is an unauthorized domainGamma when no link is present k =0,τ l,k =∞;
{χ i' } 1≤i'≤I For indicating the state of the currently deployed VNF, if ^ x i' If n is not zero, then v is i' Deployed in domain d n Upper, if X i' If =0, then v is represented i' Not yet deployed, and has χ i' =0,i'≥i。
Preferably, when the A3C network is trained, in the Actor network, the set constraint condition is implemented by using an action mask, that is, a decision that does not satisfy the constraint is shielded; wherein the constraint condition comprises:
1) One VNF can only be deployed into one domain, and a physical link must exist between two consecutively deployed VNFs;
2) Resources occupied by all Virtual Network Functions (VNFs) deployed in one domain cannot exceed the resource margin of the domain;
3) The bandwidth occupied by all the Virtual Network Functions (VNFs) carried by one link cannot exceed the bandwidth of the link;
4) The sum of the processing delay and the cross-domain transmission delay in all VNF domains cannot exceed the total delay.
Preferably, the return is calculated by a return function as follows:
wherein, S ∈ {0,1} is a Boolean variable, which indicates whether the service function chain is successfully deployed, S =1 successfully, and S =0 unsuccessfully; r S And R F Reward punishment of success and failure of deployment respectively;
Representing the resources within the domain required for deploying all virtual network functions VNF V in constraint 2);
c l,k representing the link bandwidth required for deploying all virtual network functions VNF V in constraint 3);
τ v,l representing the processing delay required for deploying all virtual network functions VNF V in constraint 4);
The invention has the following beneficial effects:
in the cross-domain SFC deployment method of the heaven and earth fusion network based on the hybrid state synchronization DRL, on one hand, huge expenses caused by real-time synchronization of the global network state of the heaven and earth fusion network can be effectively reduced through key parameter synchronization, and the cross-domain SFC deployment method is suitable for satellite control nodes with limited resources. On the other hand, by virtue of the synchronization of the state information of the global network and the idea of digital twin, the virtual copy Actor-Critic network of the A3C has rich and sufficiently real sample sets, and the success rate of SFC deployment of the global Actor-Critic network and the real copy Actor-Critic network can be further improved. Meanwhile, the actual copy Actor-Critic network of the A3C makes an optimal decision on cross-domain SFC deployment of the world fusion network based on real-time synchronous network key state parameters, and supports world fusion network services.
Drawings
Fig. 1 is a schematic block diagram of a space-ground convergence network cross-domain SFC deployment based on hybrid state synchronous DRL.
Detailed Description
The invention is described in detail below by way of example with reference to the accompanying drawings.
The invention aims to overcome the defects of the prior art and solve the problem of cross-domain SFC deployment under space-time large scale of a space-ground converged network, and provides a space-ground converged network cross-domain SFC deployment method based on hybrid state synchronous DRL. Compared with the SFC deployment method based on the global state synchronization and the key state synchronization, the method provided by the invention has the advantages of low synchronization overhead, high deployment success rate, sufficient training samples and the like.
The method is realized by the following technical scheme:
aiming at cross-domain service function chain deployment under large space-time scale, the invention introduces a deep reinforcement learning technology into a heaven-earth fusion network, decouples the heaven-earth fusion network from the aspects of logic and function, and constructs a logic network model based on a physical plane, a control plane and an intelligent plane, as shown in figure 1.
In the method provided by the invention, the real-time synchronous key parameter is used for updating the real copy Actor-criticic network in the A3C, and the synchronous global network state generates the virtual key parameter by a digital twin method and is used for updating the virtual copy Actor-criticic network. When the real copy Actor-critical network or the virtual copy Actor-critical network completes a round of updating, the strategy gradient of the real copy Actor-critical network or the virtual copy Actor-critical network is used for updating the global Actor-critical network. Let the global Actor-Critic network be Θ G There are P real copies of the Actor-Critic network asThere are Q virtual copy Actor-criticic networks as
For the network state, a set of N network domains of the world convergence network is D, a set of K inter-domain links is L, a set of I Virtual Network Functions (VNFs) contained in the SFC is V, a set of J resource types is R, and the ith VNfv is i An authorized domain that e V can be deployed isFor j resource type r j E.g. requirement of R is alpha i,j The bandwidth required for cross-domain transmission is w i The maximum time delay of the whole SFC is tau + . N-th field d n E.g. D resource r j The balance of(if d) n Is an unauthorized domain, then) Of VNfv i Has a processing delay ofThe kth inter-domain link l k The bandwidth of E L is gamma k Time delay of τ l,k (in the absence of a link γ) k =0,τ l,k = ∞). Thus, VNfv is being deployed i The key network parameter Delta needed for neural network training i Network topology, authorized domain, resource allowance of each domain, processing time delay, bandwidth and time delay of inter-domain link, namely:
whereinIs v is i Parameters obtained after deployment, the values of which are related to the VNF deployment algorithm used in the domain, are used
Indicating the ability to obtain directly (except forExternal) to the network. Thus, deployment VNFv i The temporal network state may be represented as a vector
Wherein, the interval 1 is not less than I' not more than I comprises I; { X i' } 1≤i'≤I For indicating the state of the currently deployed VNF, if ^ x i' If n is not zero, then v is i' Deployed in domain d n Upper, if X i' If =0, then v is represented i' Not yet deployed and has a chi i' =0,i'≥i。
For network decision, VNfv i Deployable decision is a i ∈D i . It is assumed that the shortest path from each domain to the other domains is known (which can be solved by Dijkstra). Suppose that the value of the Boolean variable b epsilon {0,1} is 0 indicates that the event is false (NO), and the value of 1 indicates that the event is true (YES).
1) Using Boolean variables, taking into account the gamut mapping constraintsDenotes v i Whether or not to be deployed to domain d n The above. There is therefore a gamut mapping constraint, namely:
the constraint indicates that a VNF can only be deployed into one domain.
In addition, two successive VNFvs i And v i+1 The following cross-domain mapping constraints must be satisfied:
where the Boolean variable ρ n,m E {0,1} represents the field d n To domain d m Whether or not a link exists, the constraint indicating that a physical link must exist between two VNFs deployed in succession. For example, when domain dn to domain d m When there is no link, the link is not present,denotes v i And v i+1 Cannot be deployed to domain d, respectively n And domain d m The above.
2) Considering the intra-domain resource constraints, for a successfully deployed SFC (i.e., all VNFVs), it will be directed to the intra-domain resources r j Is satisfied with
The constraint indicates that the resources occupied by all VNFs deployed within a domain cannot exceed the domain resource margin.
3) Considering the cross-domain link bandwidth constraint, for a successfully deployed SFC (i.e., all VNF V), the bandwidth constraint is
Wherein the Boolean variable ξ n,m,k E {0,1} represents d n To domain d m Whether or not the path of (1) contains l k The constraint means that the bandwidth occupied by all VNFs carried by a link cannot exceed the link bandwidth.
4) Considering the cross-domain link latency constraint, for a successfully deployed SFC (i.e., all VNFVs), the latency constraint is:
the constraint represents the total delay of one SFC, i.e. the sum of the processing delay in all VNF domains and the cross-domain transmission delay, cannot exceed the total delay.
In the Actor network, the constraint condition is realized by adopting an action mask, namely, the decision which does not meet the constraint is shielded.
For the return function, a weight factor is definedp l,k ,p v,l Representing the cost overhead of a unit resource (computation, storage, bandwidth, time delay, etc.) to the heaven-earth converged network, the reward function is:
wherein the Boolean variable S is equal to {0,1} and represents whether the service function chain is deployed successfully, R S And R F Respectively, successful deployment and failed deployment.
For training optimization of the A3C network, two cases will be distinguished.
First, for a real copy Actor-critical networkIn the ith VNfv to SFC i When deployment is carried out, key parameters of the network are fused according to the real-time synchronization of the heaven and the earthTo obtain s i Input ofGets v i Deployment decision a i And sent to the control plane. Control plane according to a i V is to be i Deployment to respective domains according to v i Post-deployment result derived parametersAnd calculating a reward R based on R and a i Calculating a policy gradient, updating in sequenceThe Critic network and the Actor network until all VNFVs of the SFC are deployed. When SFC is completed once, updatingThe gradient of (a) is uploaded to a global Actor-critical network Θ G And update the theta sequentially G Finally copying the parameters of the Actor network and the critical network to a real copy Actor-critical network
Second, for a virtual copy Actor-critical network, the network isIn the ith VNfv to SFC i When deployment is carried out, key parameters are calculated according to the global network state of intelligent plane synchronizationAnd obtaining s i Input ofGets v i Deployment decision a i With the aid of the idea of digital twin, in network simulation software, a is executed based on the network state of each domain i V is to be i Deploying to the corresponding domain, and randomly running the existing intra-domain VNF embedding algorithm to v i Simulating the execution condition in the domain to obtain the parametersAnd calculating a reward R based on R and a i Calculating a policy gradient, updating in sequenceAnd (4) the Critic network and the Actor network until all VNF V of the SFC are deployed. When SFC is completed once, updatingThe gradient of (a) is uploaded to a global Actor-critical network Θ G And update the theta sequentially G Finally copying the parameters of the Actor network and the criticic network to a real copy Actor-criticic networkIs composed of
The training of the two networks is carried out in parallel without mutual interference.
In summary, the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A space-ground fusion network cross-domain SFC deployment method based on hybrid state synchronous DRL is characterized in that an A3C network is adopted for training, wherein: one Actor-critical network is used as a global Actor-critical network theta G P Actor-critical networks form a real copy Actor-critical network setQ Actor-critical networks form a virtual copy Actor-critical networkP and Q are not less than 1;
determining a required heaven-earth converged network key parameter vector when deploying the ith virtual network function VNF of the heaven-earth converged networkAnd networking network state vector s of the world converged network i ;
For training optimization of the A3C network, the following two processes are carried out simultaneously:
the first process is as follows: for real copy Actor-Critic networkIn the i-th virtual network function VNF v i When deployment is carried out, the network is converged according to the real-time synchronous heaven and earthKey parameterTo obtain s i Input ofGets v i Deployment decision a i And sends it to the control plane; control plane according to a i V is to be i Deployment to respective domains according to v i Get a reward based on the deployed results of (a) i Calculating a policy gradient, updating in sequenceThe Critic network and the Actor network until all the VNFs are deployed; when one SFC deployment is finished, updatingThe gradient of (a) is uploaded to a global Actor-critical network Θ G And update the theta sequentially G Finally copying the parameters of the Actor network and the critical network to a real copy Actor-critical network
The second process is as follows: for a virtual copy Actor-critical network, the network isIn the i-th virtual network function VNF v i When deployment is carried out, key parameters are calculated according to the global network state of intelligent plane synchronizationAnd obtain s i Inputting ofGets v i Deployment decision a i In the world-wide integration network simulation software, a is executed based on the network state of each domain i V is to be i Deployment to corresponding Domain, v i Simulating the execution situation in the domain to obtain the return, and obtaining the return based on the return and a i Calculating a policy gradient, updating in sequenceThe Critic network and the Actor network until all VNFs are deployed; when one deployment is completed, updatingThe gradient of (a) is uploaded to a global Actor-Critic network Θ G And update the theta sequentially G Finally copying the parameters of the Actor network and the critical network to a real copy Actor-critical network
The deployment of SFCs (i.e., all VNFVs) is accomplished in accordance with the first procedure and the second procedure.
2. The method of claim 1, wherein the space-ground converged network cross-domain SFC deployment method based on hybrid state synchronous DRL is characterized in that the space-ground converged network key parameter vectorComprises the following steps:
networking network state vector s of the world convergence network i Comprises the following steps:
the method comprises the following steps that a set of N network domains of a world convergence network is set to be D, a set of K inter-domain links is set to be L, a set of I virtual network functions VNF contained in an SFC is set to be V, and a set of J resource types is set to be R; the interval 1. Ltoreq. I'. Ltoreq.I comprises I, the ith VNF v i An authorized domain that e V can be deployed isFor j resource type r j E.g. the requirement of R is alpha i,j The bandwidth required for cross-domain transmission is w i Maximum time delay of the whole SFC is tau + (ii) a N-th field d n E.g. D resource r j The balance ofIt is to VNF v i Has a processing delay ofThe kth inter-domain link l k The bandwidth of E L is gamma k Time delay of τ l,k (ii) a Wherein when d n When it is an unauthorized domainGamma when no link is present k =0,τ l,k =∞;
{χ i' } 1≤i'≤I For indicating the state of the currently deployed VNF, if ^ x i' If n is not zero, then v is i' Deployed in domain d n Upper, if X i' If =0, then v is represented i' Not yet deployed, and has χ i' =0,i'≥i。
3. The method for deploying SFC across heaven and earth fusion network based on hybrid state synchronous DRL according to claim 2, wherein when training the A3C network, in the Actor network, the action mask is adopted to realize the set constraint condition, i.e. the decision not meeting the constraint is shielded; wherein the constraint condition comprises:
1) One VNF can only be deployed into one domain, and a physical link must exist between two consecutively deployed VNFs;
2) Resources occupied by all Virtual Network Functions (VNFs) deployed in one domain cannot exceed the resource margin of the domain;
3) The bandwidth occupied by all the Virtual Network Functions (VNFs) carried by one link cannot exceed the bandwidth of the link;
4) The sum of the processing delay and the cross-domain transmission delay in all VNF domains cannot exceed the total delay.
4. The method for deploying SFCs across the heaven and earth converged network based on hybrid state synchronous DRL of claim 3, wherein the reward is calculated by a reward function as follows:
wherein, S ∈ {0,1} is a Boolean variable, which indicates whether the service function chain is successfully deployed, S =1 successfully, and S =0 unsuccessfully; r S And R F Respectively being reward punishment of successful deployment and failed deployment;
representing the resources within the domain required by all the VNFVs of the virtual network function in constraint 2);
c l,k representing the link bandwidth required for deploying all virtual network functions VNF V in constraint 3);
τ v,l representing the processing delay required for deploying all virtual network functions VNF V in constraint 4);
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211457628.5A CN115834371B (en) | 2022-11-21 | 2022-11-21 | Cross-domain SFC deployment method of space-earth fusion network based on hybrid state synchronous DRL |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211457628.5A CN115834371B (en) | 2022-11-21 | 2022-11-21 | Cross-domain SFC deployment method of space-earth fusion network based on hybrid state synchronous DRL |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115834371A true CN115834371A (en) | 2023-03-21 |
CN115834371B CN115834371B (en) | 2024-05-03 |
Family
ID=85529757
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211457628.5A Active CN115834371B (en) | 2022-11-21 | 2022-11-21 | Cross-domain SFC deployment method of space-earth fusion network based on hybrid state synchronous DRL |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115834371B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180026911A1 (en) * | 2016-07-25 | 2018-01-25 | Cisco Technology, Inc. | System and method for providing a resource usage advertising framework for sfc-based workloads |
CN110505099A (en) * | 2019-08-28 | 2019-11-26 | 重庆邮电大学 | A kind of service function chain dispositions method based on migration A-C study |
CN113573320A (en) * | 2021-07-06 | 2021-10-29 | 西安理工大学 | SFC deployment method based on improved actor-critic algorithm in edge network |
CN113660668A (en) * | 2021-05-15 | 2021-11-16 | 西安电子科技大学 | Seamless credible cross-domain routing system of heterogeneous converged network and control method thereof |
CN114172820A (en) * | 2021-11-26 | 2022-03-11 | 广东技术师范大学 | Cross-domain SFC dynamic deployment method, device, computer equipment and storage medium |
-
2022
- 2022-11-21 CN CN202211457628.5A patent/CN115834371B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180026911A1 (en) * | 2016-07-25 | 2018-01-25 | Cisco Technology, Inc. | System and method for providing a resource usage advertising framework for sfc-based workloads |
CN110505099A (en) * | 2019-08-28 | 2019-11-26 | 重庆邮电大学 | A kind of service function chain dispositions method based on migration A-C study |
CN113660668A (en) * | 2021-05-15 | 2021-11-16 | 西安电子科技大学 | Seamless credible cross-domain routing system of heterogeneous converged network and control method thereof |
CN113573320A (en) * | 2021-07-06 | 2021-10-29 | 西安理工大学 | SFC deployment method based on improved actor-critic algorithm in edge network |
CN114172820A (en) * | 2021-11-26 | 2022-03-11 | 广东技术师范大学 | Cross-domain SFC dynamic deployment method, device, computer equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN115834371B (en) | 2024-05-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110012516B (en) | Low-orbit satellite routing strategy method based on deep reinforcement learning architecture | |
Zhao et al. | Intelligent digital twin-based software-defined vehicular networks | |
Mousavinejad et al. | Distributed cyber attacks detection and recovery mechanism for vehicle platooning | |
Wang et al. | Synchronization of multi-layer networks: From node-to-node synchronization to complete synchronization | |
Yu et al. | Distributed adaptive control for synchronization in directed complex networks | |
WO2021036414A1 (en) | Co-channel interference prediction method for satellite-to-ground downlink under low earth orbit satellite constellation | |
Zhang et al. | Sampled-data consensus of linear time-varying multiagent networks with time-varying topologies | |
Li et al. | Resilient cooperative control for networked Lagrangian systems against DoS attacks | |
CN114124823B (en) | Self-adaptive routing method, system and equipment oriented to high dynamic network topology | |
Xu et al. | Living with artificial intelligence: A paradigm shift toward future network traffic control | |
Zhao et al. | Decentralized Dynamic Event-Triggered $\mathcal {H} _ {\infty} $ Control for Nonlinear Systems With Unreliable Communication Channel and Limited Bandwidth | |
WO2022028926A1 (en) | Offline simulation-to-reality transfer for reinforcement learning | |
CN115510316A (en) | Privacy protection cross-domain recommendation system based on federal representation learning | |
CN113660676A (en) | Base station flow prediction method, system, storage medium and equipment | |
CN115714741A (en) | Routing decision method and system based on collaborative multi-agent reinforcement learning | |
CN115834371A (en) | Space-ground converged network cross-domain SFC deployment method based on hybrid state synchronous DRL | |
Hu et al. | Learning model parameters for decentralized schedule-driven traffic control | |
CN107749801A (en) | A kind of virtual network function laying method based on population Incremental Learning Algorithm | |
Gao et al. | Data-driven cooperative output regulation of multi-agent systems under distributed denial of service attacks | |
WO2018176768A1 (en) | Network architecture of humanoid network and implementation method | |
Liu et al. | General decentralized federated learning for communication-computation tradeoff | |
Chiu et al. | Arc-based Traffic Assignment: Equilibrium Characterization and Learning | |
Huang et al. | Bipartite multi-tracking in MASs with intermittent communication | |
Zuo et al. | Distributed asynchronous consensus control of nonlinear multi-agent systems under directed switching topologies | |
CN113645055B (en) | Implementation method suitable for multi-factor routing protocol in complex battlefield environment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |