Heterogeneous resource allocation method and system for network modal generation
By using a cooperative game theory model and a multi-agent reinforcement learning algorithm, the rigidity problem of resource allocation in multimodal networks was solved, the dynamic allocation of heterogeneous resources was realized, resource utilization and system efficiency were improved, and the successful generation of new network modes was ensured.
Patent Information
- Application Number
- CN202511560550.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-29
- Publication Date
- 2026-02-17
AI Technical Summary
In multimodal network environments, traditional static or semi-static resource allocation strategies cannot be dynamically adjusted, resulting in low resource utilization, idle resources, and resource bottlenecks for critical services. Furthermore, the lack of adaptive resource allocation capabilities fails to meet differentiated business needs, leading to resource competition conflicts.
By employing a cooperative game theory model and a multi-agent reinforcement learning algorithm, a resource allocation strategy is constructed using asymmetric Nash bargaining theory. The contribution of each network mode is calculated, an optimization objective that balances collective and individual interests is established, and interactive training is performed using a multi-agent reinforcement learning algorithm to achieve dynamic allocation of heterogeneous resources.
It achieves fairness and rationality in the allocation of multimodal network resources, dynamically adapts to complexity, improves the overall resource utilization rate, ensures the successful generation of new network modes and the preservation of the interests of existing modes, and enhances the overall efficiency of the system.
Smart Images
Figure CN121547415A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of network resource management and allocation technology, and in particular to a heterogeneous resource allocation method and system oriented towards network modality generation. Background Technology
[0002] In a multimodal network environment, multiple network modes need to share heterogeneous resources (such as computing, storage, and bandwidth) in the cloud network infrastructure. However, due to the significant differences in service requirements among different modes, these resources often compete fiercely in actual operation. Traditional network mode generation methods typically employ static or semi-static resource allocation strategies, heavily relying on pre-configured rules or manual intervention. The insufficient programmability of the underlying infrastructure prevents dynamic adjustments based on real-time service load, network status, or service quality requirements. This rigid resource management mechanism not only makes it difficult to achieve global optimization of cross-modal resource allocation but also results in low resource utilization, with some resources remaining idle for extended periods while critical services face resource bottlenecks.
[0003] The lack of adaptive resource allocation capabilities for multimodal networks makes it difficult to provide differentiated resource guarantees for different services, resulting in poor service adaptability and an inability to meet the increasingly complex application scenarios. Furthermore, the fragmented management of heterogeneous resources and the lack of a unified collaborative scheduling framework further exacerbate resource contention and limit the overall performance of cloud network infrastructure. With the application and improvement of programmable technology, multimodal network platforms can achieve unified representation and integrated programming scheduling of underlying cloud network resources. To adapt to differentiated service needs, they need to support the rapid generation of network modes.
[0004] Resource allocation for generating new network modes is complex when resources are limited. New network modes compete with other network modes for limited heterogeneous resources, which causes negative impacts. However, successfully generating network modes can bring new sources of revenue and improve the overall efficiency of the system. Therefore, resource allocation becomes a complex game problem. Summary of the Invention
[0005] To address the heterogeneous resource competition and complex resource allocation game during the generation of new network modalities, this invention proposes a heterogeneous resource allocation method and system for network modality generation. By utilizing a cooperative game model and a multi-agent reinforcement learning algorithm, it handles the complex resource allocation game relationships during the generation of new network modalities, outputs effective heterogeneous resource allocation strategies for each network modality in a multimodal network environment, achieves successful generation of new network modalities, enhances collective interests while taking into account individual interests.
[0006] To achieve the above objectives, the technical solution adopted is:
[0007] This invention provides a heterogeneous resource allocation method for network modality generation, comprising the following steps:
[0008] Step 1: Construct a cooperative game model for generating new network modes in a multimodal network environment. The cooperative game model is based on asymmetric Nash bargaining theory and is used to characterize the resource allocation game relationship when multiple network modes coexist.
[0009] Step 2: Calculate the contribution of each network mode to the generation of the new network mode;
[0010] Step 3: Based on the cooperative game model and contribution level, establish a resource allocation optimization objective that takes into account both collective and individual interests;
[0011] Step 4: Use a multi-agent reinforcement learning algorithm to interactively train the agents corresponding to each network modality to obtain heterogeneous resource allocation strategies;
[0012] Step 5: Based on the heterogeneous resource allocation strategy, realize the dynamic allocation of heterogeneous resources in the multimodal network environment to ensure the successful generation of new network modes.
[0013] According to the heterogeneous resource allocation method for network modality generation of the present invention, the process of establishing the cooperative game model in step 1 further includes: identifying the network modality subjects participating in the game, wherein the network modality subjects include existing network modalities and new network modalities to be generated; defining the utility function of each network modality, wherein the expression of the utility function is: Among them U i Let i be the utility of the i-th network mode after participating in cooperation. For the service revenue after adjusting for the i-th network mode, X i For the revenue compensation of the i-th network mode, The cost of relinquishing resources for the i-th network mode is determined; using an asymmetric Nash bargaining model, the optimization objective of resource allocation is transformed into a Nash product, expressed as: The constraint is C1: in Let be the bargaining chip for the i-th network mode, i.e., the payoff in the non-cooperative state, and m be the total number of network modes participating in the cooperative game. Constraint C1 guarantees that the payoff of network modes participating in the cooperative game is no less than that in the non-cooperative state.
[0014] According to the heterogeneous resource allocation method for network modality generation of the present invention, the resource relinquishment cost is further... It consists of losses related to node task migration, flow rerouting, and service quality degradation caused by reduced resources during network convergence.
[0015] According to the heterogeneous resource allocation method for network modality generation of the present invention, the specific calculation method of the contribution degree in step 2 is as follows: for the j-th existing network modality, its contribution degree w j = The amount of resources relinquished by the j-th existing network mode / the total amount of resources required by the nodes and links of the new network mode; for the new network mode, its contribution w new = Unallocated remaining heterogeneous resources / Total resources required by new network mode nodes and links, and the sum of the contributions of all network modes satisfies the normalization condition.
[0016] According to the heterogeneous resource allocation method for network modality generation of the present invention, the expression for the resource allocation optimization objective in step 3 is further as follows: The constraint is C1: Among them, w i The contribution of the i-th network mode.
[0017] According to the heterogeneous resource allocation method for network modality generation of the present invention, the multi-agent reinforcement learning algorithm in step 4 adopts a multi-agent proximal policy optimization algorithm, wherein the agent corresponds one-to-one with the network modality, including new network modality agents and existing network modality agents.
[0018] According to the heterogeneous resource allocation method for network modality generation of the present invention, the observable environment state of the new network modality agent further includes: the topology of the underlying physical network, the heterogeneous resource distribution information in the physical network, the topology requirements of the new network modality, and the amount of heterogeneous resource requests; the observable environment state of the existing network modality agent includes: its own network topology, its own heterogeneous resource occupancy, and the node overlap information and link overlap information with the new network modality.
[0019] According to the heterogeneous resource allocation method for network modality generation of the present invention, the action of the agent is further defined as follows: the action of the new network modality agent is the heterogeneous resource quota requirement corresponding to its nodes and links; the action of the existing network modality agent is the heterogeneous resource capacity that it needs to give up for its overlapping nodes and links with the new network modality.
[0020] According to the heterogeneous resource allocation method for network modality generation of the present invention, the reward function of the agent is further set as: reward of the new network modality agent = contribution of the new network modality × (utility after cooperation - utility without cooperation); Where w i For the contribution of the i-th existing network mode, U i For the benefits of cooperation, This is a non-cooperative utility.
[0021] Furthermore, the present invention also provides a heterogeneous resource allocation system for network modality generation, comprising:
[0022] The cooperative game model construction module is used to construct a cooperative game model for the generation of new network modes in a multimodal network environment. The cooperative game model is constructed based on asymmetric Nash bargaining theory and is used to characterize the resource allocation game relationship when multiple network modes coexist.
[0023] The contribution calculation module is used to calculate the contribution of each network mode to the generation of new network modes;
[0024] The optimization objective establishment module is used to establish a resource allocation optimization objective that takes into account both collective and individual interests based on the cooperative game model and contribution level.
[0025] The agent training module is used to interactively train agents corresponding to each network mode using a multi-agent reinforcement learning algorithm to obtain heterogeneous resource allocation strategies.
[0026] The resource dynamic allocation module is used to dynamically allocate heterogeneous resources in a multimodal network environment according to the heterogeneous resource allocation strategy, so as to ensure the successful generation of new network modes.
[0027] The beneficial effects achieved by adopting the above technical solution are:
[0028] 1. Achieving fairness and rationality in multimodal network resource allocation: This invention employs an asymmetric Nash bargaining model to construct a cooperative game decision-making mechanism, which fully adapts to the asymmetric characteristics among network modes (participants). By accurately calculating the contribution of each network mode to the generation of a new network mode—that is, based on the proportion of resources waived by existing network modes relative to the resources required by the nodes and links of the new mode, and the degree of support provided by the unallocated remaining heterogeneous resources for the generation of the new mode—cooperative benefits are allocated. This mechanism ensures fairness in the distribution of benefits, avoids the imbalance caused by single-dimensional decision-making, and significantly improves the rationality of the resource allocation mechanism.
[0029] 2. Dynamically Adapting to the Complexity and Dynamics of Multimodal Network Environments: This invention relies on Multi-Agent Reinforcement Learning (MARL) and Software-Defined Networking (SDN) to construct a working framework. On the one hand, it collects partial observable information from each network modality through agents, achieving accurate perception of the network environment state. On the other hand, it employs the MARL algorithm of Multi-Agent Proximal Policy Optimization (MAPPO) to complete the interactive training of agents corresponding to each network modality. This effectively addresses the complexity of resource allocation games in multimodal networks, as well as the dynamic changes in business load and network state. It breaks through the rigid limitations of traditional static / semi-static resource allocation strategies, outputting accurate and effective resource allocation strategies for logically isolated network modalities.
[0030] 3. Balancing collective and individual interests to ensure successful generation of new network modalities: This invention takes "jointly optimizing the overall resource efficiency of multimodal networks, ensuring the successful generation of new network modalities, and improving the quality of business services" as its core optimization objectives. Through the construction of a cooperative game model, it ensures that the interests of each participating network modality are not lower than those in the non-cooperative state after resource allocation adjustments (satisfying individual rationality). At the same time, it can efficiently schedule idle computing, storage, and other resources of network modalities during business downtime to new modalities by dynamically adjusting heterogeneous resource allocation strategies. This significantly improves the overall resource utilization rate and overall system efficiency of multimodal networks, achieving the dual objectives of "successful generation of new network modalities" and "no loss of interests of existing modalities". Attached Figure Description
[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.
[0032] Figure 1 This is an overall framework diagram of the heterogeneous resource allocation method for network modality generation according to an embodiment of the present invention. Detailed Implementation
[0033] The exemplary solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art.
[0034] To ensure the generation of new network modes, multiple network modes cooperate in heterogeneous resource allocation. A dynamic resource allocation mechanism is introduced, allocating idle computing power or storage space from network modes during periods of low traffic to new network modes. Through cooperative game theory, the resource utilization and overall efficiency of the multimodal network are significantly improved. Figure 1 As shown, this embodiment of the invention discloses a heterogeneous resource allocator for network modality generation based on asymmetric Nash bargaining theory, comprising the following steps:
[0035] Step S101: Construct a cooperative game model for generating new network modes in a multimodal network environment. The cooperative game model is constructed based on asymmetric Nash bargaining theory and is used to characterize the resource allocation game relationship when multiple network modes coexist.
[0036] Based on asymmetric Nash bargaining theory, a cooperative game model is constructed that includes the following elements:
[0037] ① Participants: Identify the network modal entities participating in the game. These include existing network modalities and new network modalities to be generated.
[0038] ② Utility function
[0039] In a multimodal network environment, multiple participating network modes in a cooperative game dynamically adjust their heterogeneous resource allocation strategies during the symbiotic evolution process to successfully generate new network modes. They weigh the potential benefits of cooperation against the costs of resource allocation, aiming to optimize and enhance collective interests while ensuring their own interests (i.e., individual rationality). For m network modes in a multimodal network environment, the overall optimization objective for resource allocation to successfully generate new network modes can be expressed as:
[0040]
[0041] in, and These represent the revenue that the i-th network mode obtains from services before and after resource allocation adjustments for the generation of the new network mode, respectively. X represents the cost of resource allocation adjustments made by the i-th network mode to generate the new network mode. Due to the reduction of heterogeneous resources in some nodes and links, a series of problems such as task migration of nodes, rerouting of flows, and network convergence are caused, which reduce the quality of service and cause resource allocation costs. i C1 represents the compensation received by the i-th network mode for relinquishing its network resources. C2 represents that all network modes coexist in a multimodal network environment and receive benefits from services, and C3 represents that the benefits of each network mode are not affected by the adjustment of resource allocation.
[0042] For the i-th network mode, its specific utility U, when participating in cooperative game theory... i This mainly includes business revenue, revenue compensation, and the cost of relinquishing resources, which can be expressed as:
[0043]
[0044] Among them, U i Let i be the utility of the i-th network mode after participating in cooperation. For the service revenue after adjusting for the i-th network mode, X i For the revenue compensation of the i-th network mode, The cost of relinquishing resources for the i-th network mode.
[0045] ③ The goal of cooperative game
[0046] By utilizing cooperative game theory based on Nash bargaining, the optimization objective of resource allocation (Formula 1) can be transformed into the form of a Nash product, which yields:
[0047]
[0048] in, The bargaining chip represents the i-th network mode participating in the cooperative game, i.e., the payoff obtained when negotiations break down and the game reverts to the non-cooperative state. Constraint C1 indicates that the payoff obtained from participating in the cooperative game cannot be lower than the payoff in the non-cooperative state, and also represents the individual rationality in Nash bargaining cooperative games.
[0049] The relationship between Formula 1 and Formula 3: Formula 1 is the starting point of the optimization problem, defining the "efficiency" objective of resource allocation, namely, maximizing the total net benefit of the system's collective. However, merely maximizing efficiency may lead to uneven distribution of benefits, making it difficult to guarantee the rationality and fairness of each individual participating in the cooperative network modality. Therefore, Formula 3 introduces the concept of cooperative game theory based on Formula 1, shifting the optimization objective from maximizing the total benefit to maximizing the product of the cooperative dividends (i.e., excess utility) of all participants. This shift is fundamental: Formula 3, through the mathematical form of a series of products, internalizes the social contract of "fair distribution" as the core optimization objective of the system. It not only inherits the initial intention of Formula 1 to improve collective benefits but also strengthens the protection of individual interests, ensuring that the total benefit generated by cooperation can be distributed in a way that is relatively satisfactory to all participants.
[0050] Step S102: Calculate the contribution of each network mode to the generation of the new network mode.
[0051] To ensure fair benefit distribution, the contribution of each network mode to the successful generation of a new network mode is calculated. Specifically, for the j-th existing network mode, its contribution w is calculated. j = The amount of resources relinquished by the j-th existing network mode / the total amount of resources required by the nodes and links of the new network mode; for the new network mode, its contribution w new = Unallocated remaining resources / Total resources required for new network mode nodes and links, and the sum of the contributions of all network modes satisfies the normalization condition, that is, the sum of the contributions of all network modes satisfies the normalization requirement that "the sum of contributions is 1".
[0052] Step S103: Based on the cooperative game model and contribution level, establish a resource allocation optimization objective that takes into account both collective and individual interests.
[0053] The contribution weights calculated in step S102 are introduced into the Nash product optimization objective in step S101 (Formula 3) to form the final resource allocation optimization objective considering fair benefit distribution, which is expressed as follows:
[0054]
[0055] Among them, w iThe contribution of the i-th network mode to cooperative game play is determined by constraint C1, which guarantees that the benefit of the network mode participating in cooperative game play is no less than that of the non-cooperative state. This ensures that cooperation is "profitable" for each participant. The utility of each mode after cooperation cannot be worse than that of non-cooperation.
[0056] Step S104: Use a multi-agent reinforcement learning algorithm to interactively train the agents corresponding to each network modality to obtain heterogeneous resource allocation strategies.
[0057] The present invention employs the MARL algorithm of Multi-Agent Proximal Policy Optimization (MAPPO) to achieve interactive training of agents related to various network modalities. The states, actions, and rewards of the agents are as follows:
[0058] (1) State: For an agent with a newly generated network modality, its observable environmental state includes: the topology of the underlying physical network, the distribution information of heterogeneous resources in the physical network, the topology requirements of the network modality to be generated, and the amount of heterogeneous resource requests; for an agent with an existing network modality, its observable environmental state includes its own network topology, its own heterogeneous resource occupancy, and whether the nodes and links overlap with the new network modality.
[0059] (2) Actions: Actions are the heterogeneous resource allocation strategies adopted by multiple agents for the generation of new network modalities. For agents of newly generated network modalities, their actions are the specific quotas required by their nodes and links; for agents of existing network modalities, their actions are the heterogeneous resource capacity that their overlapping nodes and links need to give up.
[0060] (3) Rewards: For agents in newly generated network modalities, the reward is the product of their contribution and the difference in utility before and after cooperation; for agents in existing network modalities, the reward is...
[0061] Step S105: Based on the heterogeneous resource allocation strategy, realize the dynamic allocation of heterogeneous resources in the multimodal network environment to ensure the successful generation of new network modes.
[0062] The heterogeneous resources mentioned in this article include computing resources, storage resources, and bandwidth resources. The dynamic allocation process is implemented by the software-defined networking (SDN) controller, which distributes and executes resource allocation policies. The SDN controller works in conjunction with the network mode orchestrator and the physical network infrastructure manager to complete the unified scheduling of heterogeneous resources.
[0063] Corresponding to the above method, embodiments of the present invention also disclose a heterogeneous resource allocation system for network modality generation, the system comprising:
[0064] The cooperative game model construction module is used to construct a cooperative game model for the generation of new network modes in a multimodal network environment. The cooperative game model is constructed based on asymmetric Nash bargaining theory and is used to characterize the resource allocation game relationship when multiple network modes coexist.
[0065] The contribution calculation module is used to calculate the contribution of each network mode to the generation of new network modes;
[0066] The optimization objective establishment module is used to establish a resource allocation optimization objective that takes into account both collective and individual interests based on the cooperative game model and contribution level.
[0067] The agent training module is used to interactively train agents corresponding to each network mode using a multi-agent reinforcement learning algorithm to obtain heterogeneous resource allocation strategies.
[0068] The resource dynamic allocation module is used to dynamically allocate heterogeneous resources in a multimodal network environment according to the heterogeneous resource allocation strategy, so as to ensure the successful generation of new network modes.
[0069] Unless otherwise specifically stated, the relative steps, numerical expressions, and values of the components and steps described in these embodiments do not limit the scope of the invention.
[0070] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0071] The units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations are not considered to be beyond the scope of this invention.
[0072] Those skilled in the art will understand that all or part of the steps in the above methods can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk. Optionally, all or part of the steps in the above embodiments can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiments can be implemented in hardware or as a software functional module. This invention is not limited to any particular combination of hardware and software.
[0073] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A heterogeneous resource allocation method for network modality generation, characterized in that, The method comprises the following steps: Step 1, constructing a cooperative game model for new network mode generation in a multi-modal network environment, wherein the cooperative game model is constructed based on asymmetric Nash bargaining theory and is used to represent resource allocation game relationship when multiple network modes coexist; Step 2, calculating the contribution degree of each network mode to the generation of the new network mode; Step 3, establishing a resource allocation optimization target considering both collective interests and individual interests based on the cooperative game model and the contribution degree; Step 4, using a multi-agent reinforcement learning algorithm to interactively train agents corresponding to each network mode to obtain a heterogeneous resource allocation strategy; Step 5, realizing dynamic allocation of heterogeneous resources in the multi-modal network environment according to the heterogeneous resource allocation strategy to ensure successful generation of the new network mode.
2. The network modal generation oriented heterogeneous resource allocation method according to claim 1, characterized in that, The establishment process of the cooperation game model in step 1 includes: clearly defining the network modal subject participating in the game, the network modal subject including an existing network modal and a new network modal to be generated; defining the utility function of each network modal, the expression of the utility function being: Wherein U i is the utility of the i th network modal after participating in cooperation, is the adjusted business income of the i th network modal, X i is the income compensation of the i th network modal, is the resource giving-up cost of the i th network modal; using an asymmetric Nash bargaining model, the optimization goal of resource allocation is converted into the form of Nash product, expressed as: The constraint condition is C1: Wherein is the negotiation weight of the i th network modal, that is, the income in the non-cooperation state, m is the total number of network modals participating in the cooperation game, and the constraint condition C1 ensures that the interests of the network modals participating in the cooperation game are not lower than those in the non-cooperation state.
3. The network modal generation oriented heterogeneous resource allocation method according to claim 2, characterized in that, The resource giving-up cost The related loss is caused by the task migration of the node, the flow re-routing, and the quality of service degradation caused by the network convergence due to the resource reduction.
4. The method for heterogeneous resource allocation oriented to network modal generation according to claim 1, characterized in that, The contribution degree described in step 2 is specifically calculated as follows: for the jth existing network mode, the contribution degree w j = the resource amount given up by the jth existing network mode / the total resource amount required by the new network mode nodes and links; For the new network mode, its contribution degree w new = unallocated remaining isomeric resource amount / total resource amount required by new network mode nodes and links, and the sum of the contribution degrees of all network modes satisfies the normalization condition.
5. The network modal generation oriented heterogeneous resource allocation method according to claim 2, characterized in that, The expression of the resource allocation optimization objective in step 3 is as follows: The constraint condition is C1: Wherein, w i is the contribution degree of the i-th network mode.
6. The network modal generation oriented heterogeneous resource allocation method according to claim 1, characterized in that, The multi-agent reinforcement learning algorithm used in step 4 is a multi-agent proximal policy optimization algorithm, and the agents correspond one-to-one to the network modes, including a new network mode agent and an existing network mode agent.
7. The network modal generation oriented heterogeneous resource allocation method according to claim 6, characterized in that, The observable environment state of the new network mode agent includes the topology of the underlying physical network, the distribution information of the heterogeneous resources in the physical network, the topology demand of the new network mode, and the request amount of the heterogeneous resources; and the observable environment state of the existing network mode agent includes the network topology structure of itself, the occupation situation of the heterogeneous resources, the node overlap information with the new network mode, and the link overlap information.
8. The network modal generation oriented heterogeneous resource allocation method according to claim 6, characterized in that, The action of the agent is defined as follows: the action of the new network mode agent is the heterogeneous resource quota demand corresponding to its nodes and links; and the action of the existing network mode agent is the heterogeneous resource capacity to be given up for the overlapping nodes and links with the new network mode.
9. The network modal generation oriented heterogeneous resource allocation method according to claim 6, wherein, The reward function of the intelligent agent is set as: the reward of the new network mode intelligent agent = the contribution degree of the new network mode × (the post-cooperation utility - the non-cooperation utility); wherein w i is the contribution degree of the i th existing network mode, U i is the post-cooperation utility, is the non-cooperation utility.
10. A heterogeneous resource allocation system oriented to network modal generation, characterized by, The method comprises the following steps: a cooperative game model construction module for constructing a cooperative game model for new network mode generation in a multi-modal network environment, wherein the cooperative game model is constructed based on asymmetric Nash bargaining theory and is used to represent resource allocation game relationship when multiple network modes coexist; a contribution degree calculation module for calculating the contribution degree of each network mode to the generation of the new network mode; an optimization target establishment module for establishing a resource allocation optimization target considering both collective interests and individual interests based on the cooperative game model and the contribution degree; an agent training module for using a multi-agent reinforcement learning algorithm to interactively train agents corresponding to each network mode to obtain a heterogeneous resource allocation strategy; a resource dynamic allocation module for realizing dynamic allocation of heterogeneous resources in the multi-modal network environment according to the heterogeneous resource allocation strategy to ensure successful generation of the new network mode.