Resource allocation method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a resource allocation method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a service request of a target user; constructing a white list corresponding to the target user according to the target model and the service request; the white list comprises a service sub-time slot corresponding to the service request and a first resource quantity corresponding to each service sub-time slot; and according to the target model, the service sub-time slots and the first resource quantity corresponding to each service sub-time slot, analyzing a multi-modal signal instruction and a feedback demand of the service request, and determining a target service channel corresponding to a multi-modal signal contained in each service sub-time slot in the service request and a second resource quantity corresponding to each target service channel. By adopting the method, accurate resource allocation of the service request can be realized, and the accuracy of edge cloud computing resource scheduling is improved.

Description

Technical Field

[0001] The present application relates to the field of cloud computing technology, and in particular to a resource allocation method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the development of cloud computing technology, the applications or computing services provided by cloud center nodes can be decomposed into several parts and distributed to the user's local nodes for processing, forming a distributed processing and storage architecture of the edge cloud to meet the computing needs of applications with low latency requirements.

[0003] In traditional technology, after receiving a user's service request, the edge cloud computing server first analyzes and identifies the specific requirements of the service request. Then, the edge cloud computing server treats the service request as a whole and allocates fixed scheduling resources according to the pre-set scheduling strategy and pre-set resource blocks.

[0004] However, the current traditional technology, due to the large granularity of resource blocks relative to business requests, the fixed scheduling resources allocated according to resource blocks cause resource waste, resulting in poor accuracy in edge cloud computing resource scheduling. Summary of the invention

[0005] Based on this, it is necessary to provide a resource allocation method, apparatus, computer device, computer-readable storage medium and computer program product to address the above technical issues.

[0006] In a first aspect, the present application provides a resource allocation method, comprising:

[0007] Obtain the target user's business request;

[0008] Constructing a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes the service sub-timeslot corresponding to the service request and the first resource amount corresponding to each of the service sub-timeslots;

[0009] According to the target model, the service sub-time slots and the first resource amount corresponding to each of the service sub-time slots, the multimodal signal instructions and feedback requirements of the service request are analyzed to determine the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels.

[0010] In one embodiment, before obtaining the service request of the target user, the method further includes:

[0011] Obtain historical service requests of each target user in the target area;

[0012] The preset resource information corresponding to each target area is determined according to the first component of the target model and the historical business request; the preset resource information includes edge nodes, initial cache resource types corresponding to each edge node, initial cache resource sizes, initial cache resource request times, initial cache resource acquisition times, initial cache resource release times, and initial cache resource release objects.

[0013] In one embodiment, the constructing a whitelist corresponding to the target user according to the target model and the service request includes:

[0014] The interactive scene included in the business request is segmented according to the second component of the target model to obtain a plurality of business sub-timeslots, and a first resource amount corresponding to each of the business sub-timeslots is determined based on the second component; the first resource amount includes a total amount of instruction recognition computing power and a total amount of rendering power corresponding to the business sub-timeslots;

[0015] A whitelist corresponding to the target user is constructed based on the service sub-timeslots and the total instruction recognition computing power and the total rendering power corresponding to each of the service sub-timeslots.

[0016] In one of the embodiments, the analyzing the multimodal signal instruction and feedback requirement of the service request according to the target model, the service sub-time slot and the first resource amount corresponding to each of the service sub-time slots, and determining the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels includes:

[0017] If the type of the service request is a target type, decomposing a multimodal signal included in the service request according to the third component of the target model, the service sub-timeslots, and the first resource amount corresponding to each of the service sub-timeslots to obtain a target modal signal;

[0018] Based on the fourth component, the service sub-time slots and the first resource amounts corresponding to each of the service sub-time slots, the target modal signal instructions and feedback requirements of the service request are analyzed to determine the target service channels corresponding to each of the target modal signals and the second resource amounts corresponding to each of the target service channels.

[0019] In one of the embodiments, if the type of the service request is a target type, decomposing the multimodal signal contained in the service request according to the third component of the target model, the service sub-timeslot, and the first resource amount corresponding to each of the service sub-timeslots to obtain a target modal signal includes:

[0020] If the type of the service request is a target type, determining, in the multimodal signal included in the service request, a resource strength corresponding to each modal signal in the multimodal signal according to the third component of the target model, the service sub-timeslot, and the first resource amount corresponding to each service sub-timeslot;

[0021] A target modal signal is determined in each of the modal signals according to the resource strength.

[0022] In one of the embodiments, based on the fourth component, the service sub-time slot and the first resource amount corresponding to each of the service sub-time slots, analyzing the target modal signal instruction and feedback requirement of the service request, determining the target service channel corresponding to each of the target modal signals and the second resource amount corresponding to each of the target service channels, includes:

[0023] Based on the fourth component, the service sub-timeslot and the first resource amount corresponding to each of the service sub-timeslots, the target modal signal instruction and feedback requirement of the service request are analyzed to obtain the initial resource amount of each of the target modal signals in each of the service sub-timeslots;

[0024] The real-time minimum transmission unit of each of the service sub-timeslots is calculated based on the initial resource amount, and the target service channel corresponding to each of the target modal signals and the second resource amount corresponding to each of the target service channels are determined based on the real-time minimum transmission unit.

[0025] In one of the embodiments, after analyzing the multimodal signal instruction and feedback requirement of the service request according to the target model, the service sub-time slot and the first resource amount corresponding to each of the service sub-time slots, and determining the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels, the method further includes:

[0026] The resource occupancy inventory corresponding to the target user is obtained, and the resource occupancy inventory and resource occupancy time are updated to the whitelist corresponding to the target user.

[0027] In one embodiment, the method further comprises:

[0028] If the resource occupation time corresponding to the service request reaches the initial cache resource release time, determine whether there is a same-domain request with the same service type as the service request within the same-domain range corresponding to the service request;

[0029] If the same-domain request exists, the resource occupancy stock of the target user is scheduled to the target address corresponding to the same-domain request.

[0030] In one of the embodiments, if the same-domain request exists, scheduling the resource usage stock of the target user to the target address corresponding to the same-domain request includes:

[0031] If the same-domain request exists, the address pointer of the target user is added to the whitelist of the user to which the same-domain request belongs; the address pointer of the target user in the whitelist of the user to which the same-domain request belongs is used to build a dedicated tunnel;

[0032] The empty segment of the message frame structure of the service request is marked according to a preset special value to obtain a target message frame structure; the target message frame structure is used to guide the user to which the same domain request belongs to access the resource occupation stock.

[0033] In a second aspect, the present application further provides a resource allocation device, including:

[0034] A first acquisition module, used to acquire a target user's service request;

[0035] A construction module, configured to construct a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes a service sub-timeslot corresponding to the service request and a first resource amount corresponding to each service sub-timeslot;

[0036] The first allocation module is used to analyze the multimodal signal instructions and feedback requirements of the service request according to the target model, the service sub-time slots and the first resource amount corresponding to each service sub-time slot, and determine the target service channel corresponding to the multimodal signal contained in each service sub-time slot in the service request and the second resource amount corresponding to each target service channel.

[0037] In one embodiment, the device further comprises:

[0038] The second acquisition module is used to acquire the historical service requests of each target user in the target area;

[0039] The second allocation module is used to determine the preset resource information corresponding to each target area according to the first component of the target model and the historical business request; the preset resource information includes edge nodes, the initial cache resource type corresponding to each edge node, the initial cache resource size, the initial cache resource request time, the initial cache resource acquisition time, the initial cache resource release time and the initial cache resource release object.

[0040] In one of the embodiments, the construction module is specifically used to segment the interaction scene contained in the business request according to the second component of the target model to obtain multiple business sub-timeslots, and determine the first resource amount corresponding to each of the business sub-timeslots based on the second component; the first resource amount includes the total amount of instruction recognition computing power and the total amount of rendering corresponding to the business sub-timeslot;

[0041] A whitelist corresponding to the target user is constructed based on the service sub-timeslots and the total instruction recognition computing power and the total rendering power corresponding to each of the service sub-timeslots.

[0042] In one of the embodiments, the first allocation module is specifically configured to decompose the multimodal signal included in the service request according to the third component of the target model, the service sub-timeslot and the first resource amount corresponding to each of the service sub-timeslots to obtain a target modal signal if the type of the service request is a target type;

[0043] Based on the fourth component, the service sub-time slots and the first resource amounts corresponding to each of the service sub-time slots, the target modal signal instructions and feedback requirements of the service request are analyzed to determine the target service channels corresponding to each of the target modal signals and the second resource amounts corresponding to each of the target service channels.

[0044] In one of the embodiments, the first allocation module is specifically configured to determine, if the type of the service request is a target type, a resource strength corresponding to each modal signal in the multimodal signal included in the service request according to the third component of the target model, the service sub-timeslot and the first resource amount corresponding to each service sub-timeslot;

[0045] A target modal signal is determined in each of the modal signals according to the resource strength.

[0046] In one of the embodiments, the first allocation module is specifically used to analyze the target modal signal instruction and feedback requirement of the service request based on the fourth component, the service sub-timeslot and the first resource amount corresponding to each of the service sub-timeslots, and obtain the initial resource amount of each of the target modal signals in each of the service sub-timeslots;

[0047] The real-time minimum transmission unit of each of the service sub-timeslots is calculated based on the initial resource amount, and the target service channel corresponding to each of the target modal signals and the second resource amount corresponding to each of the target service channels are determined based on the real-time minimum transmission unit.

[0048] In one embodiment, the device further comprises:

[0049] The updating module is used to obtain the resource occupancy inventory corresponding to the target user, and update the resource occupancy inventory and resource occupancy time to the whitelist corresponding to the target user.

[0050] In one embodiment, the device further comprises:

[0051] A determination module, configured to determine whether there is a same-domain request with the same service type as the service request within the same-domain range corresponding to the service request if the resource occupation time corresponding to the service request reaches the initial cache resource release time;

[0052] The scheduling module is used to schedule the resource occupancy stock of the target user to the target address corresponding to the same-domain request if the same-domain request exists.

[0053] In one of the embodiments, the scheduling module is specifically used to add the address pointer of the target user to the whitelist of the user to which the same-domain request belongs if the same-domain request exists; the address pointer of the target user in the whitelist of the user to which the same-domain request belongs is used to build a dedicated tunnel;

[0054] The empty segment of the message frame structure of the service request is marked according to a preset special value to obtain a target message frame structure; the target message frame structure is used to guide the user to which the same domain request belongs to access the resource occupation stock.

[0055] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0056] Obtain the target user's business request;

[0057] Constructing a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes the service sub-timeslot corresponding to the service request and the first resource amount corresponding to each of the service sub-timeslots;

[0058] According to the target model, the service sub-time slots and the first resource amount corresponding to each of the service sub-time slots, the multimodal signal instructions and feedback requirements of the service request are analyzed to determine the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels.

[0059] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0060] Obtain the target user's business request;

[0061] Constructing a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes the service sub-timeslot corresponding to the service request and the first resource amount corresponding to each of the service sub-timeslots;

[0062] According to the target model, the service sub-time slots and the first resource amount corresponding to each of the service sub-time slots, the multimodal signal instructions and feedback requirements of the service request are analyzed to determine the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels.

[0063] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0064] Obtain the target user's business request;

[0065] Constructing a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes the service sub-timeslot corresponding to the service request and the first resource amount corresponding to each of the service sub-timeslots;

[0066] According to the target model, the service sub-time slots and the first resource amount corresponding to each of the service sub-time slots, the multimodal signal instructions and feedback requirements of the service request are analyzed to determine the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels.

[0067] The resource allocation method, device, computer equipment, computer readable storage medium and computer program product obtain a service request, build a whitelist corresponding to the service request according to the first component of the target model; the whitelist contains the service sub-time slot corresponding to the service request and the first resource amount corresponding to each service sub-time slot; according to the second component of the target model, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, determine the target service channel corresponding to the multimodal signal contained in the service request and the second resource amount corresponding to each target service channel; and allocate resources to the service request based on the target service channel and the second resource amount corresponding to each target service channel. By adopting this method, a whitelist is constructed through a target model, and a more fine-grained resource analysis of the service request is realized according to the service sub-time slot, and the service sub-time slot adapted to the service request and the first resource amount corresponding to each service sub-time slot are obtained, and the multimodal signal in the service request is analyzed based on the target model, and the target service channel and the second resource amount required are dynamically determined based on the service sub-time slot and the first resource amount corresponding to each of the service sub-time slots, so as to realize accurate resource allocation for the service request and improve the accuracy of edge cloud computing resource scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 An application environment diagram of a resource allocation method in an embodiment;

[0070] Figure 2 A schematic diagram of a flow chart of a resource allocation method in an embodiment;

[0071] Figure 3 A functional architecture diagram of resource allocation by an edge cloud server in one embodiment;

[0072] Figure 4 A schematic diagram of a process for configuring preset resource information in an embodiment;

[0073] Figure 5 A schematic diagram of allocating preset resource information in an embodiment;

[0074] Figure 6 A schematic diagram of a process for building a whitelist in one embodiment;

[0075] Figure 7 A schematic diagram of building a whitelist in one embodiment;

[0076] Figure 8 A schematic diagram of a process for dividing a target service channel and a second resource amount in one embodiment;

[0077] Fig. 9 A schematic diagram of a flow chart of a step of determining a target modal signal in one embodiment;

[0078] Fig.10 It is a schematic diagram of a process of dividing a target service channel and a second resource amount in another embodiment;

[0079] Fig.11 A schematic diagram of dividing a target service channel and a second resource amount in one embodiment;

[0080] Fig.12 A schematic diagram of a process of requesting resource sharing in the same domain in an embodiment;

[0081] Fig.13 A schematic diagram of establishing a same-domain user negotiation mechanism and a resource-directed protocol in one embodiment;

[0082] Fig.14 A schematic diagram of a dynamic resource rapid release mechanism in one embodiment;

[0083] Fig.15 A schematic diagram of a process for marking a message frame structure in an embodiment;

[0084] Fig.16 A schematic diagram of an example of a resource allocation method in an embodiment;

[0085] Fig.17 is a structural block diagram of a resource allocation device in an embodiment;

[0086] Fig.18 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0088] Edge cloud is a distributed processing and storage architecture that decomposes the applications or computing services provided by the original central node into several parts and distributes them to edge cloud nodes for processing, so as to meet the needs of emerging applications such as low latency and large data traffic, such as extended reality based on tactile interaction, smart manufacturing, autonomous driving and other popular scenarios. However, in highly interactive services, such as 5G tactile somatosensory games 3v3, stronger real-time computing and rendering capabilities are required to complete the simulation and reproduction of tactile signals. Insufficient edge cloud resources and computing power will lead to tactile signal "artifacts" and tactile feedback hysteresis.

[0089] In traditional technology, after receiving the user's business request, the edge cloud computing server first analyzes and identifies the specific needs of the business request. Then, the edge cloud computing server calculates the optimal scheduling path for each node according to the pre-set scheduling strategy. After broadcasting, it needs to summarize the calculations, calculate redundancy, and waste over-saturated resources. In addition, according to the pre-set resource blocks, the business request is treated as a whole and fixed scheduling resources are allocated.

[0090] However, the scheduling of edge cloud computing resources usually uses resource types and resource blocks as the granularity of requesting resource allocation. Resource scheduling is to reasonably and effectively adjust, monitor, analyze and use various resources (for example, CPU, GPU, memory, etc.). The resource blocks based on traditional technologies are the smallest allocable units of these resource types within a certain range. The resource allocation method based on resource blocks in traditional technologies is relatively coarse-grained, mainly based on preset scheduling strategies and fixed resource pools to achieve rapid response and simplified resource management. However, this coarse-grained allocation method may have certain limitations in resource utilization and dynamic adjustment. In addition, resource scheduling is highly correlated with the immediacy of actual user operations. Traditional technologies cannot meet the needs of users for accurate on-demand scheduling, which results in poor accuracy in edge cloud computing resource scheduling.

[0091] The resource allocation method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the edge cloud server 104 through the network. The data storage system can store the data that the edge cloud server 104 needs to process. The data storage system can be integrated on the edge cloud server 104, and the data storage system includes resource storage, and the resource storage includes a preset resource pool, a dynamic resource pool, and an elastic resource application whitelist. The edge cloud server 104 obtains the service request initiated by the terminal 102 of the target user; the edge cloud server 104 builds a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes the service sub-time slot corresponding to the service request and the first resource amount corresponding to each service sub-time slot; the edge cloud server 104 analyzes the multimodal signal instructions and feedback requirements of the service request according to the target model, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, and determines the target service channel corresponding to the multimodal signal contained in each service sub-time slot in the service request and the second resource amount corresponding to each target service channel. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. The edge cloud server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. The edge cloud server manages AI models and dispatches edge cloud computing resources through the edge cloud management platform. The edge cloud service includes an AI (Artificial Intelligence) model, which is used to decide the allocation of preset resource pools and dynamic resource pools. The edge cloud server 104 also includes applications that provide computing services for the terminal 102, such as games, audio and video, monitoring, cloud conferencing, and security applications.

[0092] In an exemplary embodiment, Figure 2 As shown, a resource allocation method is provided, which is applied to Figure 1 The edge cloud server in the example is used to illustrate, including the following steps 202 to 206. Among them:

[0093] Step 202: Obtain the service request of the target user.

[0094] Among them, the service request includes a tactile interaction signal. Tactile interaction is the interaction between computer applications and tactile information control based on skin feeling or motion, including perception, reproduction and feedback. The interaction instructions contained in the tactile interaction signal are generally complex, requiring high precision and ultra-low latency tactile feedback.

[0095] In the present application embodiment, Figure 3 As shown, Figure 3 This is the main functional module architecture of dynamic resource scheduling, where the terminal of the target user can be User1 or User2, and the edge cloud can be a mobile edge cloud server in the edge node. The target user initiates a service request to the edge cloud server through the terminal, and the service request is used to apply for subsequent dynamic resources. The edge cloud server receives the service request initiated by the terminal of the target user, and then obtains the service request of the target user.

[0096] In a specific embodiment, the target user needs to receive real-time tactile feedback while watching a virtual reality (VR) game to enhance the sense of immersion. This scenario involves complex multimodal signal transmission, and the service request obtained by the edge cloud computing server includes visual signals, auditory signals, and tactile interaction signals.

[0097] Step 204: construct a whitelist corresponding to the target user according to the target model and the service request.

[0098] The whitelist includes service sub-time slots corresponding to the service request and first resource amounts corresponding to each service sub-time slot.

[0099] In the embodiment of the present application, the target model may be an AI model, for example, the AI ​​model may be a neural network model or a large-scale pre-trained model (large model) to handle complex business logic and resource management tasks. Figure 3 As shown in the figure, the edge cloud server analyzes and processes the business request through the AI ​​model, and then builds a whitelist corresponding to the target user. The whitelist has multiple functions. First, it serves as a security mechanism to authenticate users and isolate other business users; second, it serves as a real-time monitoring and dynamically updated database to record and update the verification tag information of the resource release flow between the same business users; third, it divides users into different levels.

[0100] Specifically, the whitelist can be pre-built when the edge cloud computing server is initialized or when the user registers, and then dynamically updated according to the business requests and resource usage of the target user in the actual application. For example, during initialization, a preliminary whitelist is created, which contains the basic information of all registered users. Each time a new user registers, the basic information of the new user and the permissions of the business type that the user has are added to the whitelist. The basic information includes the user's unique identifier (e.g., user ID, MAC address, etc.).

[0101] The mobile edge cloud server monitors the behavior and resource usage of all users in real time. When a user initiates a new request or resource usage changes, the relevant information is recorded immediately. At the same time, the terminal of the target user regularly reports its resource usage, such as reporting resource inventory information at the end of each time window. After receiving the user's report information, the edge cloud server updates the corresponding record in the whitelist.

[0102] The edge cloud server receives a service request (Service-X-Buffer Request, requesting an operation to adjust or configure the buffer of service Service-X) from a target user. The service request includes the target user's identity information, service type (e.g., VR game), required resource type and size, etc. The edge cloud server first verifies the target user's identity information to ensure that the service request is from an authorized user, and checks whether the target user has the authority to access the corresponding service type in the service request to prevent unauthorized users from accessing services with high resource requirements. Then, the edge cloud server inputs the service request into the AI ​​model, analyzes the service request through the AI ​​model, and divides the service request into multiple service sub-time slots, each of which may correspond to a specific operation or scene. For example, a service request for synchronous transmission of tactile interaction signals, visual signals, and auditory signals is divided into three service sub-time slots: sub-time slot 1 (initial scene loading), sub-time slot 2 (dynamic scene), and sub-time slot 3 (game task switching). The AI ​​model allocates the corresponding first resource amount to each service sub-time slot based on the mapping relationship between the sub-time slot and the required resources learned during the training process.

[0103] Step 206, analyze the multimodal signal instructions and feedback requirements of the service request according to the target model, the service sub-time slots and the first resource amount corresponding to each service sub-time slot, and determine the target service channel corresponding to the multimodal signal contained in each service sub-time slot in the service request and the second resource amount corresponding to each target service channel.

[0104] In the embodiment of the present application, the edge cloud computing server further analyzes the multimodal signal instructions and feedback requirements of the multimodal signal in the service request through the target model, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, analyzes and determines the type of modal signal instructions of each modal signal, and determines the feedback requirements of each modal signal, and finally outputs the number of service channels corresponding to each modal signal in the service request, and the second resource amount corresponding to each service channel. For example, when a user applies for a weak interactive service (such as a simple interactive instruction or does not require immediate tactile feedback), the multimodal signal can be directly analyzed and calculated without splitting the multimodal signal (audio-visual touch), and the minimum unit of the real-time multimodal signal Multi-Signal-Unit is calculated. When users apply for strong interactive services (such as complex interactive instructions or tactile feedback that requires high precision or ultra-low latency), the tactile interaction signal can be separated, and even the audio-visual and tactile three-dimensional signals can be independent of each other. The minimum unit of real-time single-channel signal (a certain modal signal) (S-Signal-Unit1, S-Signal-Unit2...) is calculated, and each single-channel signal is divided into 2 to N channels for transmission (the tactile interaction signal transmission requires high requirements, and multiple tactile sub-channels can be divided for parallel transmission). Monitor each service sub-time slot. When the task of each service sub-time slot ends, dynamically release the idle channel and wait for the next time slot (re-applied by the original target user or other users of the same service and the same second layer) to activate the call. When the service session ends, release all channels.

[0105] In the above resource allocation method, a whitelist is constructed through a target model to achieve a more fine-grained resource analysis of business requests according to business sub-slots, and business sub-slots adapted to the business requests and the first resource amounts corresponding to each business sub-slot are obtained. Based on the target model, the multimodal signals in the business requests are analyzed, and the target business channels and the second resource amounts required by them are dynamically determined based on the business sub-slots and the first resource amounts corresponding to each of the business sub-slots, so as to achieve accurate resource allocation for business requests and improve the accuracy of edge cloud computing resource scheduling.

[0106] In an exemplary embodiment, before dynamically scheduling resources for a service request, it is necessary to first perform preset resource scheduling, such as Figure 4 As shown, before step 202, the method further includes steps 402 to 404. Among them:

[0107] Step 402: Obtain historical service requests of each target user in the target area.

[0108] In an embodiment of the present application, the target area may be a user area divided by the edge cloud server according to user distribution characteristics, or may be a target area divided according to residential areas, industrial areas, and the like. Different target areas have different user types. For example, business requests in residential areas are mostly related to games and audio and video, while business requests in industrial areas may be mostly cloud conferences, smart manufacturing, and the like. Alternatively, the types of business requests corresponding to residential areas in different areas are also different, and thus the edge cloud computing resource requirements required by target users in different areas are different. Therefore, the edge cloud server can allocate corresponding preset resources to different target areas. At the same time, target users in different areas may undergo changes such as migration, and the edge cloud server can schedule preset resources for each target area according to a preset period. Therefore, the edge cloud server collects the historical business requests of each target user contained in each target area as training samples for the target model.

[0109] Step 404: Determine preset resource information corresponding to each target area according to the first component of the target model and the historical service request.

[0110] The preset resource information includes edge nodes, initial cache resource types corresponding to each edge node, initial cache resource sizes, initial cache resource request time, initial cache resource acquisition time, initial cache resource release time, and initial cache resource release objects.

[0111] The first component of the target model is a preset resource allocation component, which is used to predict the preset resource information of the target users in each target area.

[0112] In an embodiment of the present application, the edge cloud server uses historical business requests as training samples of the first component of the target model, and uses the optimal preset resource information adapted in each target area as a sample label, that is, the sample label corresponding to each historical business request used as a training sample is the optimal preset resource information, and then the first component of the target model is trained according to the historical business requests and the optimal preset resource information. Each historical business request is processed in a loop through the first component to be trained, and the initial preset resource information corresponding to the historical business request of the current round is obtained in each loop, and the current loss value of the first component is calculated based on the initial preset resource information and the optimal preset resource information, until the first loss value of the first component meets the preset training condition during the loop processing, then the first component at this time is determined to be the first component that has completed training.

[0113] In the application of the first component of the target model by the edge cloud computing server, the edge cloud server analyzes the business requests initiated by each target user in the current cycle according to the trained first component according to a preset cycle, and determines the threshold resource information corresponding to each target area in the next cycle.

[0114] In a specific embodiment, Figure 5 As shown in the figure, based on the business front-end survey and business back-end statistics of each business in the edge cloud computing server, the historical business requests of the target users are used as training sample data, and an AI model is built to predict the cloud resource requirements Service-X-Buffer Request of each business. Service-X-Buffer Request provides users with the node location for obtaining the initial cache resources (the AI ​​model analyzes the busy area coverage and the neighborhood resource access range to determine the edge node location M), the initial cache resource type (bandwidth, CPU, GPU computing power, etc.), the initial cache resource size, the initial cache resource request time, the initial cache resource acquisition time, the initial cache resource release time, and the initial cache resource release object (which can be released back to the edge cloud total pool, or released to adjacent users of the same business through a dedicated protocol and a business-exclusive tunnel).

[0115] In this embodiment, by collecting the historical business requests of each target user in the target area and using the first component of the target model (pre-set resource allocation component) for analysis and training, the edge cloud server can accurately predict the preset resource requirements of different target areas, including edge node location, resource type, size, time and other information, thereby optimizing resource allocation and management, improving service response speed and user experience. Through dynamic scheduling and intelligent prediction, efficient utilization and flexible allocation of resources are ensured, responding to changes and migration of user needs, and improving the computing performance and adaptability of edge cloud servers in various environments.

[0116] In an exemplary embodiment, Figure 6 As shown, step 204 includes steps 602 to 604. Among them:

[0117] Step 602: segment the interaction scenario included in the service request according to the second component of the target model to obtain a plurality of service sub-time slots, and determine a first resource amount corresponding to each service sub-time slot based on the second component.

[0118] Among them, the first resource amount includes the total amount of instruction recognition computing power and the total amount of rendering corresponding to the business sub-time slot.

[0119] Among them, the second component is a service sub-time slot division component for segmenting the interactive scene according to the service request initiated by the target user in the target model.

[0120] In an embodiment of the present application, the edge cloud server predicts the changes in the interaction scenarios of the current target user and the division of possible business sub-time slots according to the business type and user behavior pattern of the business request through the second component of the target model, and obtains multiple business sub-time slots. Furthermore, the edge cloud server analyzes the types and quantities of instructions that may be involved in each sub-time slot through the second component of the target model. For example, the T1 time slot may involve instructions such as starting a game and loading resources, and identifies and classifies instructions, calculates the execution time and required computing power of each instruction, and obtains the total computing power required for all instructions in each business sub-time slot. At the same time, the edge cloud server analyzes the type of rendering tasks that need to be performed in each sub-time slot through the second component of the target model, evaluates the complexity of the rendering tasks through the second component of the target model, calculates the resource requirements of each rendering task, and obtains the total rendering amount of all rendering tasks in each business sub-time slot.

[0121] In a specific embodiment, Figure 7 As shown, the edge cloud server determines the preset resources of each edge node according to the preset resource information through the whitelist call, and the preset resources are used for subsequent refined allocation. Then, the edge cloud server monitors the resource usage of node M in real time and reports and updates it to the whitelist in real time. Then, the second component of the target model divides the interaction scene in the current business request according to the resource usage updated in real time by each user in the whitelist, for example, analyzing the type and number of instructions that may be involved in each sub-time slot.

[0122] Step 604: construct a whitelist corresponding to the target user based on the service sub-time slots and the total amount of instruction recognition computing power and the total amount of rendering power corresponding to each service sub-time slot.

[0123] Among them, Figure 7 As shown, the whitelist is the basis for the edge cloud server to schedule resources according to the dynamic resource pool.

[0124] In an embodiment of the present application, the edge cloud server is based on the business sub-time slots output by the second component of the target model and the total amount of instruction recognition computing power and rendering corresponding to each business sub-time slot, and the dynamic business resource size required for each business sub-time slot, including CPU, GPU, memory, etc., is comprehensively calculated, and then the white list corresponding to the target user is constructed. In addition, the edge cloud server generates the weight coefficient of the target user according to the dynamic business resource size of each business sub-time slot, so as to classify different target users (classes include A, B, C, etc.). The time slot interval Tx is set as an adjustable parameter of the edge cloud server. The user level affects the priority of user resource scheduling, the number of available sub-tunnels, and the priority of the next reuse. Specifically, the AI ​​model predicts the changes in the interaction scene and the possible sub-time slot division according to the business type and user behavior pattern, and divides the interaction scene into multiple sub-time slots (such as T1, T2, T3...) based on the prediction results. Each sub-time slot corresponds to a different task or interaction fragment, has specific resource requirements, and analyzes the type and number of instructions that may be involved in each sub-time slot. For example, the T1 time slot may involve instructions such as starting a game and loading resources. The computing power requirements of all instructions in each sub-time slot are accumulated to obtain the total instruction recognition computing power of each time slot, and the type and complexity of rendering tasks required in each sub-time slot are analyzed. For example, the T1 time slot may be the initial scene rendering, and the T2 time slot may be the dynamic scene rendering. The complexity of the rendering task is evaluated through the AI ​​model, the resource requirements of each rendering task are calculated, the resource requirements of all rendering tasks in each sub-time slot are accumulated, and the total rendering amount of each time slot is obtained. According to the total instruction recognition computing power and the total rendering amount, the size of dynamic business resources required for each sub-time slot is comprehensively calculated, including CPU, GPU, memory, etc. At the same time, the AI ​​model combines the real-time system load and resource availability to dynamically adjust the resource demand forecast. Based on the resource demand and time slot weight of each sub-time slot, the weight coefficient of each user is calculated, and the level threshold is set according to the distribution of the weight coefficient. For example, users with a weight coefficient greater than 0.8 are classified as Class A, users with a weight coefficient between 0.5 and 0.8 are classified as Class B, and users with a weight coefficient less than 0.5 are classified as Class C. Based on the calculated user weight coefficients, users are divided into different levels (such as Class A, B, and C), and the levels affect the priority and strategy of resource scheduling. Finally, the time slot interval Tx is an adjustable parameter of the edge cloud server, which is set according to business requirements and the performance of the edge cloud server. A smaller time slot interval can improve the precision of resource scheduling.

[0125] In a specific embodiment, taking the edge cloud server processing the VR game service request initiated by the target user A as an example, the business sub-time slot division and resource demand analysis and the construction of the white list corresponding to the target user A are illustrated. First, the edge cloud server analyzes and processes the VR game service request through the second component of the target model, and the output result is to divide the VR game scene into three business sub-time slots, including T1 (initial scene loading), T2 (dynamic scene), and T3 (task switching). At the same time, the total amount of instruction recognition computing power analysis includes T1 instruction analysis: start the game, load resources, and the computing power requirement is 2 CPU cores; T2 instruction analysis: move, shoot, and the computing power requirement is 4 CPU cores; T3 instruction analysis: task switching, the computing power requirement is 1 CPU core; total computing power: T1 is 2 CPU cores, T2 is 4 CPU cores, and T3 is 1 CPU core. The total amount of rendering calculation includes T1 rendering analysis: initial scene rendering, rendering requirement is 4 GPU cores; T2 rendering analysis: dynamic scene rendering, rendering requirement is 6 GPU cores; T3 rendering analysis: task switching rendering, rendering requirement is 2 GPU cores. Then, the edge cloud server generates a user weight coefficient based on the resource demand per time slot. For example, the user weight coefficient is 0.3 for T1, 0.5 for T2, and 0.2 for T3. Then the weight coefficient of target user A is (2×0.3+4×0.5+1×0.2) / (2+4+1)=0.46 / 7=0.657. Finally, the edge cloud server classifies target user A as a Class B user based on the weight coefficient, and sets the time slot interval Tx to 0.5 seconds based on business requirements and system performance, thereby completing the construction of a whitelist for target user A.

[0126] In this embodiment, the interactive scenarios in the service requests are segmented into the first resource amount at the granularity of the service sub-time slots, and a whitelist is constructed for each target user to prioritize users, decide on the transmission form and mode of audio-visual and touch multimodal signals, refine resources and precisely schedule them, and achieve adaptive transmission between the third layer and the second layer.

[0127] In an exemplary embodiment, Figure 8 As shown, step 206 includes steps 802 to 804. Among them:

[0128] Step 802: If the type of the service request is a target type, the multimodal signal included in the service request is decomposed according to the third component of the target model, the service sub-slot and the first resource amount corresponding to each service sub-slot to obtain a target modal signal.

[0129] Among them, the third component is used to decompose the multimodal signal in the business request and determine the target modal signal in the multimodal signal. The third component is an AI component composed of a neural network or a large model (for example, BERT or Transformer, etc.).

[0130] In an embodiment of the present application, the target type can be determined by whether the service request carries a target identifier, which indicates that the type of the current service request is a strong interaction type. That is, if the service request carries a target identifier, it means that the type of the current service request is the target type. Then, the interaction instructions in the service request are complex or require high precision, or have tactile feedback requirements with ultra-low latency.

[0131] The edge cloud server inputs the service sub-time slot of the current service request, the first resource amount corresponding to each service sub-time slot, and the multimodal signal in the service request into the third component, and after the third component performs analysis and processing, outputs and obtains the target modal signal in the service request that requires more fine-grained dynamic resource allocation. Because the multimodal signals in the service request may have signals with small resource requirements and low latency requirements, for example, the sound signal in the interface switching scene, the sound signal has a small correlation with the action in the scene or the object in other visual scenes, while in the scene that requires instant rendering, the sound signal has a high correlation with the action and the object, and the sound feedback corresponding to the action of the operation needs to be fed back in time during the user operation. Therefore, the third component predicts the target modal signal through the scene to be realized corresponding to the current service request, and realizes the screening of the target modal signal in the service request at the current stage through the third component.

[0132] Step 804, based on the fourth component, the service sub-slots and the first resource amount corresponding to each service sub-slot, analyze the target modal signal instructions and feedback requirements of the service request, and determine the target service channel corresponding to each target modal signal and the second resource amount corresponding to each target service channel.

[0133] In the embodiment of the present application, the edge cloud server performs target modal signal instruction analysis and feedback demand analysis according to the fourth component, determines the instructions that need to be executed in the current time slot, such as visual rendering, audio playback, tactile feedback, etc., and determines the feedback that each target modal signal needs to receive in the current time slot, such as user interaction confirmation, status update, etc. Under the target type of service request, the fourth component divides the number of target service channels required for the target modal signal according to the instruction analysis results and feedback demand analysis results of the target modal signal, and allocates the corresponding second resource amount to each target service channel according to each service sub-time slot and the corresponding second resource amount, so as to ensure that the total amount of the second resource amount of each target service channel is compatible with the first resource amount. For example, if the first resource amount required for a certain service sub-time slot is only 100MB of bandwidth, then under the current service sub-time slot, the sum of the second resource amounts to be allocated for all target service channels cannot exceed 100MB.

[0134] In an optional embodiment, if the type of the service request is not a target type, that is, the service request does not carry a target identifier, the multimodal signal can be processed as a whole, divided into a unified service channel, and fixed scheduling resources are allocated to the unified service channel according to pre-set resource blocks.

[0135] In this embodiment, the target modal signal of the multimodal request in the service request is first analyzed to determine the target modal signal that requires independent dynamic resource allocation, and then the target modal signal is divided into a target service channel at the granularity of the service sub-time slot. Parallel signal transmission is performed through multiple target service channels. This can improve the accuracy of resource division for service requests by the edge cloud server, accurately divide resources according to the requirements of the service request, improve the processing performance of the edge cloud server for service requests, reduce the delay of multimodal signal transmission, and avoid the occurrence of tactile signal "artifacts", tactile feedback hysteresis and other phenomena.

[0136] In an exemplary embodiment, Fig. 9 As shown, step 802 includes steps 902 to 904. Among them:

[0137] Step 902, if the type of the service request is a target type, determine the resource strength corresponding to each modal signal in the multimodal signal included in the service request according to the third component of the target model, the service sub-slot and the first resource amount corresponding to each service sub-slot.

[0138] In an embodiment of the present application, when a service request carries a target identifier, the edge cloud server analyzes and maps the multimodal signal according to the third component of the target model, and outputs the resource intensity corresponding to each modal signal in the multimodal signal. Among them, the training samples used for training by the third component are service requests, service sub-time slots corresponding to each service request, and the first resource amount corresponding to each service sub-time slot, and the pre-set resource intensity is used as a sample label. Specifically, the third component analyzes the instruction complexity and feedback requirements of each modal signal according to the service sub-time slots divided for service requests in the whitelist and the first resource amount corresponding to each service sub-time slot, comprehensively analyzes the computing resources and transmission resources of each service sub-time slot with the multimodal signal in the service request, and predicts and outputs the resource intensity of the requirements corresponding to each modal signal.

[0139] Step 904: determine a target modal signal among the modal signals according to the resource strength.

[0140] In the embodiment of the present application, the edge cloud server determines the signal with resource strength greater than the signal strength threshold of Yisu River as the target modal signal according to the resource strength corresponding to each modal signal output by the third component and the preset signal strength threshold. Optionally, in different time slots, the edge cloud server can dynamically adjust different target modal signals to ensure that the modal signal with the highest resource strength is always given priority.

[0141] In this embodiment, the multimodal signal in the service request is decomposed, and the resource intensity of each modal signal is calculated based on the first resource amount of the service sub-slot. According to the resource intensity, the modal signal with the highest resource intensity is selected as the target modal signal to ensure that within the limited resource constraints, the signal with the highest demand for system resources is processed first, thereby improving the overall efficiency and response speed of the edge cloud server.

[0142] In an exemplary embodiment, Fig.10 As shown, step 804 includes steps 1002 to 1004. Among them:

[0143] Step 1002, based on the fourth component, the service sub-slot and the first resource amount corresponding to each service sub-slot, analyze the target modal signal instruction and feedback requirements of the service request to obtain the initial resource amount of each target modal signal in each service sub-slot.

[0144] Among them, the fourth component is used to analyze the target modal signal instructions and feedback requirements of the target modal signal, and calculate the real-time minimum resource size, that is, the real-time minimum transmission unit. The real-time minimum transmission unit is used as the basis for dividing the size and number of business channels during business sessions.

[0145] In the embodiment of the present application, the edge cloud server inputs the target modal signal, the service sub-time slot and the first resource amount corresponding to each service sub-time slot in the service request to the fourth component. Among them, the sample data for training the fourth component can be the target modal signal, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, and the sample label can be the pre-set service channel corresponding to each target modal signal and the resource amount corresponding to each service channel.

[0146] The edge cloud server comprehensively analyzes the target modal signal instructions of the target modal signal in each business sub-time slot according to the fourth component, as well as the instruction complexity corresponding to the target modal signal instructions, and analyzes the feedback requirements of the target modal signal, that is, the bandwidth and delay required to transmit the target modal signal, and finally calculates the initial resource amount of the target modal signal in each business sub-time slot.

[0147] Step 1004, calculate the real-time minimum transmission unit of each service sub-time slot based on the initial resource amount, and determine the target service channel corresponding to each target modal signal and the second resource amount corresponding to each target service channel based on the real-time minimum transmission unit.

[0148] In an embodiment of the present application, the edge cloud server calculates the real-time minimum transmission unit of each service sub-time slot based on the initial resource amount, that is, the real-time minimum transmission unit of the target modal signal in each service sub-time slot obtained after decomposing the multimodal signal in the service request. For example, the trial minimum transmission unit for the multimodal signal can be determined based on the instruction complexity, feedback requirements and the total amount of resources required for each time slot, or mapped according to the initial resource amount according to the fifth component of the target model. After determining the real-time minimum transmission unit of each target modal signal in each service sub-time slot, the edge cloud server determines the number of target service channels required for each target modal signal based on the real-time minimum transmission unit, and determines the resource amount of each target service channel, such as Fig.11As shown, the edge cloud server can divide the target business channel of the tactile signal into two according to the real-time minimum transmission unit of the tactile signal, and divide the target business channels of the visual signal and the auditory signal into one respectively according to the real-time minimum transmission unit corresponding to the visual signal and the real-time minimum transmission unit corresponding to the auditory signal. At the same time, the edge cloud server determines the second resource amount of each target service channel based on the real-time minimum transmission unit. For example, the fifth component maps the second resource amount required for each target service channel based on the real-time minimum transmission unit analysis. For example, if there is no target modal signal in the service sub-slot T1, that is, the edge cloud server processes the multimodal signal as a whole and divides it into a unified service channel, then the resource amount of the service channel can be determined; if the visual signal, auditory signal and tactile signal are all divided into target modal signals with independent resource allocation in the service sub-slot T2, then each target modal signal is divided into an independent service channel. If the number of target service channels for the visual signal, auditory signal and tactile signal are respectively divided into 1, 1, and 2, then the second resource amount of the auditory channel, the second resource amount of the visual channel, the second resource amount corresponding to tactile channel 1, and the second resource amount corresponding to tactile channel 2 can be further determined.

[0149] In this embodiment, by dividing the target modal signal into target service channels at the granularity of service sub-time slots and performing parallel signal transmission through multiple target service channels, the accuracy of resource division for service requests by the edge cloud server can be improved, resources can be accurately divided according to the requirements of service requests, the processing performance of the edge cloud server for service requests can be improved, accurate resource allocation for service requests can be achieved, and the accuracy of edge cloud computing resource scheduling can be improved.

[0150] In an exemplary embodiment, after step 206, the method further includes step 208. Wherein:

[0151] Step 208: Obtain the resource occupancy inventory corresponding to the target user, and update the resource occupancy inventory and resource occupancy time to the whitelist corresponding to the target user.

[0152] In an embodiment of the present application, the edge cloud server monitors the behavior and resource usage of all users in real time, and obtains the user's resource occupancy inventory in real time through the target user's unique identifier (for example, user ID). Then, the edge cloud server obtains the whitelist information of the target user, and updates the resource occupancy inventory and resource occupancy time to the whitelist record. Among them, the mobile edge cloud server is provided with a monitoring function, and the user information in the whitelist is updated in real time. For example, V1X-T1 in the whitelist information (U1, S1, V1X-T1) of user User1 indicates the dynamic resource occupancy inventory of the user in the T1 phase of the service S sub-time slot. The user needs to report the real-time inventory information regularly in order to build a same-domain user negotiation mechanism and a resource-oriented protocol.

[0153] In this embodiment, by updating the resource usage inventory of each target user in the whitelist, a foundation is laid for the subsequent construction of a same-domain user negotiation mechanism and a resource-directed protocol, thereby reducing computing power costs and scheduling time.

[0154] In an exemplary embodiment, Fig.12 As shown, the method further includes steps 1202 to 1204. Among them:

[0155] Step 1202: If the resource occupation time corresponding to the service request reaches the initial cache resource release time, determine whether there is a same-domain request with the same service type as the service request within the same-domain range corresponding to the service request.

[0156] In the embodiment of the present application, users in the same service and domain can share resources without applying for dynamic resources from the mobile edge cloud server for scheduling, thus saving computing power costs and scheduling time. Fig.13 As shown, Fig.14 As shown in the figure, the same domain refers to the scope of all users under the jurisdiction of a mobile edge cloud server. When the service session of user User1 in the same domain ends, another user User2 (or User3, User4) in the same domain is using the same service. The resources released by User1 can be directly reused through the dedicated tunnel (D2D) established in the second layer, instead of the traditional scheduling scheme in which User1 releases resources back to the edge cloud node server in traditional technology, and then the edge cloud server distributes the resources to User2 who applied for the resources. Specifically, the edge cloud server identifies and confirms the user group in the same area through the whitelist mechanism and user authentication. Users in the same domain can have similar geographical locations, network environments and service requirements. The edge cloud server divides users in the same domain into the same group to form a collaboration unit between users. The edge cloud server collects the resource demand data of each user in each user group in the same domain, including bandwidth, computing power, storage, etc., and obtains the user's resource usage in real time through the dynamic resource pool monitoring mechanism, and then defines the same-domain user negotiation mechanism and resource orientation protocol, defines how to share and collaborate resources between users, including resource request, resource release, resource sharing and resource priority mechanism, and formulates the resource negotiation process, including the sending, receiving and processing of resource requests, so that users can send resource requests and the edge cloud server can respond according to the negotiation protocol and resource availability.

[0157] After the edge cloud server completes the allocation of dynamic resources for the business request initiated by the current target user, it monitors the resource occupancy time corresponding to the business request initiated by the target user in real time. If the resource occupancy time corresponding to the business request reaches the initial cache resource release time, it means that this part of the resource occupancy stock is about to be released. The edge cloud server is first determined whether there is a same-domain request with the same business type as the business request within the same domain scope corresponding to the business request.

[0158] In a specific embodiment, when the business session corresponding to the business request completes the allocation of dynamic resources, a timer is set for each session to record the start time of the session, and the initial cache resource release time determined when allocating the threshold resource information is used as the time threshold. If the timer reaches the time threshold, the timer is triggered, indicating that the resource occupancy time corresponding to the business request reaches the initial cache resource release time.

[0159] Step 1204: If there is a same-domain request, the resource usage stock of the target user is scheduled to the target address corresponding to the same-domain request.

[0160] In the embodiment of the present application, if there is a same-domain request, the edge cloud server no longer releases the target user's resource occupancy inventory back to the edge cloud server, but dispatches the part of the resource occupancy inventory to the target address of the user terminal of the same-domain request, so as to realize the reuse of the target user's resource occupancy inventory. Optionally, if there are multiple same-domain requests, the edge cloud server determines the release address of the current target user's resource occupancy inventory according to the user level of the users to whom other same-domain requests belong.

[0161] In this embodiment, by introducing a resource sharing mechanism for users in the same service and domain in the edge cloud server, using a whitelist mechanism and user authentication to identify and confirm user groups in the same area, obtaining user resource usage in real time through a dynamic resource pool monitoring mechanism, and defining a same-domain user negotiation mechanism and a resource-directed protocol, resource sharing and collaboration between users are achieved. When the resource occupancy time corresponding to a service request reaches the initial cache resource release time, if there is a same-domain request, the edge cloud server dispatches the target user's resource occupancy inventory to the target address corresponding to the same-domain request, thereby avoiding the redundant process of releasing resources back to the edge cloud server and then reallocating them, significantly reducing computing power costs and scheduling time.

[0162] In an exemplary embodiment, Fig.15 As shown, step 1204 includes steps 1502 to 1504. Among them:

[0163] Step 1502: If there is a same-domain request, add the address pointer of the target user to the whitelist of the user to which the same-domain request belongs.

[0164] Among them, the address pointer of the target user in the whitelist of the user to which the same domain request belongs is used to build a dedicated tunnel.

[0165] In an embodiment of the present application, if there is a same-domain request, the edge cloud server selects the mapping relationship based on the real-time resource usage, and adds the MAC (Media Access Control Address) address pointer of the target user whose resources are about to be released to the whitelist information of the nearby users in the same domain who need resources, thereby establishing a large Layer 2 exclusive tunnel (Device-to-Device, D2D) between users in the same domain. The exclusive tunnel is used to directly reuse the resource occupation stock that is about to be released, without the need for resource scheduling through the edge cloud server.

[0166] Step 1504: Mark the empty segment of the message frame structure of the service request according to a preset special value to obtain a target message frame structure.

[0167] The target message frame structure is used to guide the user to which the request belongs in the same domain to access the resource usage inventory.

[0168] In the embodiment of the present application, the edge cloud server selects an empty segment (unused field) in the frame structure of the service message and marks it with a preset special value as a protocol flag for accessing the exclusive tunnel. The mark is used to identify that this message is transmitted through the D2D tunnel without adding an additional message header. For example, suppose an empty segment is selected in the extended header of the IPv6 message and a special value (for example, 0xFFFF) is assigned to mark that this message is transmitted through the D2D tunnel, and the mark only occupies one byte to avoid additional transmission load.

[0169] In a specific embodiment, in a certain 5G tactile interaction scenario, there are users User1 and User2, User1's service session is about to end, and User2 is using the same service and needs resources. The edge cloud server monitors the information of users User1 and User2 in the whitelist in real time, and the current resource occupancy stock of User1 is (U1, S1, V1X-T1). User1 reports its resource usage regularly, and the edge cloud server records the resources to be released. The edge cloud server adds the MAC address pointer of User1 to the whitelist information of User2, that is, it is ready to reuse User1's resources for User2. Then, the edge cloud server establishes a large Layer 2 exclusive tunnel (D2D) between User1 and User2. After the tunnel is established, User1's resources can be directly reused by User2 without scheduling through the edge cloud node server. At the same time, in the frame structure of the service message, an empty segment is selected to mark the preset special value, that is, an empty segment is selected in the extension header of the IPv6 message and assigned a special value (0xFFFF) to indicate that this message is transmitted through the D2D tunnel. Then, User2 directly accesses the occupied resource stock that is about to be released by User1 through the marked message, thereby realizing resource reuse.

[0170] In this embodiment, by optimizing resource sharing and transmission mechanisms, the resource scheduling overhead of the intermediate links is reduced, resource utilization efficiency and response speed are improved, thereby effectively reducing computing power costs and shortening scheduling time. At the same time, the empty segment of the message frame structure of the service request is marked according to a preset special value, without adding an additional message header, thereby reducing the occupancy of byte resources.

[0171] In an exemplary embodiment, Fig.16 As shown, an example of a resource scheduling method is provided, including:

[0172] Step 1, AI-driven scheduling of edge cloud sub-nodes: The user initiates a tactile interactive game service request, the edge cloud server (node ​​cluster) center confirms the service request and responds, establishes a service connection, retrieves the front-end market and back-end platform data of the service based on the service type of the request information, performs AI modeling analysis, predicts the total cloud resource demand Service-Buffer-Request for the service, determines the subsequent scheduled sub-nodes (mobile edge cloud servers), initial cache resource types (bandwidth, CPU, GPU computing power, etc.), initial cache resource size, initial cache resource usage time, initial cache resource release time, and release objects (released back to the preset resource pool of edge cloud resource storage or released to adjacent users of the same service through a tunnel via a dedicated protocol).

[0173] Step 2, create a whitelist to achieve service isolation and adaptive transmission: Create a whitelist based on the pre-scheduling information Service-Buffer-Request. First, authenticate the user to form a service isolation protection mechanism. Second, segment the scene and calculate the minimum real-time transmission unit for each time slot to prioritize users. Decide whether to transmit the audio-visual and touch multi-modal signals as a whole or split the signals into multiple parallel channels for transmission (the latter is usually selected for tactile interaction services). Perform channel division (tactile interaction usually splits the signals into 1 channel each for audio-visual and 1 channel or multiple channels for tactile) to achieve adaptive transmission. Third, monitor the information reported by the scheduling sub-node in real time to obtain the same-domain resource application.

[0174] Step 3: User negotiation enables resource sharing for users in the same domain: During busy hours, traffic in the same domain is congested. After the whitelist monitors and obtains the same domain resource request, it performs resource scheduling address mapping, establishes a service-specific tunnel, completes the assignment mark in the protocol message frame of the same domain user, and realizes fast reuse of the underlying resources in the same domain without sending it through the dynamic resource pool. By controlling the whitelist address mapping and the in-frame assignment parameters, the resource release object (resource pool or user in the same domain) is controlled to achieve adaptive resource scheduling. Resources are released in a step-by-step manner until the service ends.

[0175] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0176] Based on the same inventive concept, the embodiment of the present application also provides a resource allocation device for implementing the resource allocation method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more resource allocation device embodiments provided below can refer to the limitations on the resource allocation method above, and will not be repeated here.

[0177] In an exemplary embodiment, Fig.17 As shown, a resource allocation device 1700 is provided, comprising: a first acquisition module 1701, a construction module 1702 and a first allocation module 1703, wherein:

[0178] The first acquisition module 1701 is used to acquire the service request of the target user;

[0179] A construction module 1702 is used to construct a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes a service sub-time slot corresponding to the service request and a first resource amount corresponding to each service sub-time slot;

[0180] The first allocation module 1703 is used to analyze the multimodal signal instructions and feedback requirements of the service request according to the target model, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, and determine the target service channel corresponding to the multimodal signal contained in each service sub-time slot in the service request and the second resource amount corresponding to each target service channel.

[0181] In one embodiment, the apparatus 1700 further includes:

[0182] The second acquisition module is used to acquire the historical service requests of each target user in the target area;

[0183] The second allocation module is used to determine the preset resource information corresponding to each target area according to the first component of the target model and the historical business request; the preset resource information includes edge nodes, the initial cache resource type corresponding to each edge node, the initial cache resource size, the initial cache resource request time, the initial cache resource acquisition time, the initial cache resource release time and the initial cache resource release object.

[0184] In one embodiment, the construction module 1702 is specifically used to segment the interaction scene included in the business request according to the second component of the target model to obtain multiple business sub-time slots, and determine the first resource amount corresponding to each business sub-time slot based on the second component; the first resource amount includes the total amount of instruction recognition computing power and the total amount of rendering corresponding to the business sub-time slot;

[0185] A whitelist corresponding to the target user is constructed based on the business sub-timeslots and the total amount of instruction recognition computing power and rendering power corresponding to each business sub-timeslot.

[0186] In one embodiment, the first allocation module 1703 is specifically used to decompose the multimodal signal included in the service request according to the third component of the target model, the service sub-time slot and the first resource amount corresponding to each service sub-time slot to obtain the target modal signal if the type of the service request is a target type;

[0187] Based on the fourth component, the service sub-slots and the first resource amounts corresponding to each service sub-slot, the target modal signal instructions and feedback requirements of the service request are analyzed to determine the target service channels corresponding to each target modal signal and the second resource amounts corresponding to each target service channel.

[0188] In one embodiment, the first allocation module 1703 is specifically used to determine, if the type of the service request is a target type, the resource strength corresponding to each modal signal in the multimodal signal included in the service request according to the third component of the target model, the service sub-time slot and the first resource amount corresponding to each service sub-time slot;

[0189] A target modal signal is determined among each modal signal according to resource strength.

[0190] In one embodiment, the first allocation module 1703 is specifically used to analyze the target modal signal instruction and feedback requirements of the service request based on the fourth component, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, and obtain the initial resource amount of each target modal signal in each service sub-time slot;

[0191] The real-time minimum transmission unit of each service sub-time slot is calculated based on the initial resource amount, and the target service channel corresponding to each target modal signal and the second resource amount corresponding to each target service channel are determined based on the real-time minimum transmission unit.

[0192] In one embodiment, the apparatus 1700 further includes:

[0193] The update module is used to obtain the resource occupancy inventory corresponding to the target user, and update the resource occupancy inventory and resource occupancy time to the whitelist corresponding to the target user.

[0194] In one embodiment, the apparatus 1700 further includes:

[0195] A determination module, configured to determine whether there is a same-domain request with the same service type as the service request within the same-domain range corresponding to the service request if the resource occupation time corresponding to the service request reaches the initial cache resource release time;

[0196] The scheduling module is used to schedule the resource usage stock of the target user to the target address corresponding to the same-domain request if there is a same-domain request.

[0197] In one embodiment, the scheduling module is specifically used to add the address pointer of the target user to the white list of the user to which the same-domain request belongs if there is a same-domain request; the address pointer of the target user in the white list of the user to which the same-domain request belongs is used to build a dedicated tunnel;

[0198] The empty segment of the message frame structure of the service request is marked according to a preset special value to obtain a target message frame structure; the target message frame structure is used to guide the user to which the request belongs in the same domain to access the resource occupation stock.

[0199] Each module in the resource allocation device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0200] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.18 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a resource allocation method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0201] Those skilled in the art will understand that Fig.18 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0202] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0203] Obtain the target user's business request;

[0204] A whitelist corresponding to the target user is constructed according to the target model and the service request; the whitelist includes a service sub-time slot corresponding to the service request and a first resource amount corresponding to each service sub-time slot;

[0205] According to the target model, the service sub-time slots and the first resource amount corresponding to each service sub-time slot, the multimodal signal instructions and feedback requirements of the service request are analyzed to determine the target service channel corresponding to the multimodal signal contained in each service sub-time slot in the service request and the second resource amount corresponding to each target service channel.

[0206] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0207] Obtain historical service requests of each target user in the target area;

[0208] The preset resource information corresponding to each target area is determined according to the first component of the target model and the historical business requests; the preset resource information includes the edge node, the initial cache resource type corresponding to each edge node, the initial cache resource size, the initial cache resource request time, the initial cache resource acquisition time, the initial cache resource release time and the initial cache resource release object.

[0209] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0210] The interactive scene included in the business request is segmented according to the second component of the target model to obtain multiple business sub-time slots, and the first resource amount corresponding to each business sub-time slot is determined based on the second component; the first resource amount includes the total amount of instruction recognition computing power and the total amount of rendering corresponding to the business sub-time slot;

[0211] A whitelist corresponding to the target user is constructed based on the business sub-timeslots and the total amount of instruction recognition computing power and rendering power corresponding to each business sub-timeslot.

[0212] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0213] If the type of the service request is a target type, decomposing the multimodal signal included in the service request according to the third component of the target model, the service sub-timeslot, and the first resource amount corresponding to each service sub-timeslot to obtain a target modal signal;

[0214] Based on the fourth component, the service sub-slots and the first resource amounts corresponding to each service sub-slot, the target modal signal instructions and feedback requirements of the service request are analyzed to determine the target service channels corresponding to each target modal signal and the second resource amounts corresponding to each target service channel.

[0215] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0216] If the type of the service request is a target type, determining, in the multimodal signal included in the service request, a resource strength corresponding to each modal signal in the multimodal signal according to the third component of the target model, the service sub-timeslot, and the first resource amount corresponding to each service sub-timeslot;

[0217] A target modal signal is determined among each modal signal according to resource strength.

[0218] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0219] Based on the fourth component, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, the target modal signal instruction and feedback requirement of the service request are analyzed to obtain the initial resource amount of each target modal signal in each service sub-time slot;

[0220] The real-time minimum transmission unit of each service sub-time slot is calculated based on the initial resource amount, and the target service channel corresponding to each target modal signal and the second resource amount corresponding to each target service channel are determined based on the real-time minimum transmission unit.

[0221] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0222] Obtain the resource usage inventory corresponding to the target user, and update the resource usage inventory and resource usage time to the whitelist corresponding to the target user.

[0223] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0224] If the resource occupation time corresponding to the service request reaches the initial cache resource release time, determine whether there is a same-domain request with the same service type as the service request within the same-domain range corresponding to the service request;

[0225] If there is a same-domain request, the target user's resource usage inventory is scheduled to the target address corresponding to the same-domain request.

[0226] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0227] If there is a same-domain request, the address pointer of the target user is added to the whitelist of the user to which the same-domain request belongs; the address pointer of the target user in the whitelist of the user to which the same-domain request belongs is used to build a dedicated tunnel;

[0228] The empty segment of the message frame structure of the service request is marked according to a preset special value to obtain a target message frame structure; the target message frame structure is used to guide the user to which the request belongs in the same domain to access the resource occupation stock.

[0229] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0230] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0231] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0232] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0233] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0234] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A resource allocation method, characterized in that: The method comprises: Obtain the target user's business request; Constructing a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes the service sub-timeslot corresponding to the service request and the first resource amount corresponding to each of the service sub-timeslots; According to the target model, the service sub-time slots and the first resource amount corresponding to each of the service sub-time slots, the multimodal signal instructions and feedback requirements of the service request are analyzed to determine the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels.

2. The method according to claim 1, characterized in that Before obtaining the service request of the target user, the method further includes: Obtain historical service requests of each target user in the target area; The preset resource information corresponding to each target area is determined according to the first component of the target model and the historical business request; the preset resource information includes edge nodes, initial cache resource types corresponding to each edge node, initial cache resource sizes, initial cache resource request times, initial cache resource acquisition times, initial cache resource release times, and initial cache resource release objects.

3. The method according to claim 1, characterized in that The step of constructing a whitelist corresponding to the target user according to the target model and the service request includes: The interactive scene included in the business request is segmented according to the second component of the target model to obtain a plurality of business sub-timeslots, and a first resource amount corresponding to each of the business sub-timeslots is determined based on the second component; the first resource amount includes a total amount of instruction recognition computing power and a total amount of rendering power corresponding to the business sub-timeslots; A whitelist corresponding to the target user is constructed based on the service sub-timeslots and the total amount of instruction recognition computing power and the total amount of rendering corresponding to each of the service sub-timeslots.

4. The method according to claim 1, characterized in that: The step of analyzing the multimodal signal instruction and feedback requirement of the service request according to the target model, the service sub-timeslot, and the first resource amount corresponding to each of the service sub-timeslots, and determining the target service channel corresponding to the multimodal signal contained in each of the service sub-timeslots in the service request and the second resource amount corresponding to each of the target service channels includes: If the type of the service request is a target type, decomposing a multimodal signal included in the service request according to the third component of the target model, the service sub-timeslots, and the first resource amount corresponding to each of the service sub-timeslots to obtain a target modal signal; Based on the fourth component, the service sub-time slots and the first resource amounts corresponding to each of the service sub-time slots, the target modal signal instructions and feedback requirements of the service request are analyzed to determine the target service channels corresponding to each of the target modal signals and the second resource amounts corresponding to each of the target service channels.

5. The method according to claim 4, characterized in that If the type of the service request is a target type, decomposing the multimodal signal included in the service request according to the third component of the target model, the service sub-timeslot, and the first resource amount corresponding to each of the service sub-timeslots to obtain a target modal signal, including: If the type of the service request is a target type, determining, in the multimodal signal included in the service request, a resource strength corresponding to each modal signal in the multimodal signal according to the third component of the target model, the service sub-timeslot, and the first resource amount corresponding to each service sub-timeslot; A target modal signal is determined in each of the modal signals according to the resource strength.

6. The method according to claim 4, characterized in that The analyzing the target modal signal instruction and feedback requirement of the service request based on the fourth component, the service sub-time slot and the first resource amount corresponding to each service sub-time slot, and determining the target service channel corresponding to each target modal signal and the second resource amount corresponding to each target service channel, includes: Based on the fourth component, the service sub-timeslot and the first resource amount corresponding to each of the service sub-timeslots, the target modal signal instruction and feedback requirement of the service request are analyzed to obtain the initial resource amount of each of the target modal signals in each of the service sub-timeslots; The real-time minimum transmission unit of each of the service sub-timeslots is calculated based on the initial resource amount, and the target service channel corresponding to each of the target modal signals and the second resource amount corresponding to each of the target service channels are determined based on the real-time minimum transmission unit.

7. The method according to claim 1, characterized in that After analyzing the multimodal signal instruction and feedback requirement of the service request according to the target model, the service sub-time slots, and the first resource amount corresponding to each of the service sub-time slots, and determining the target service channel corresponding to the multimodal signal contained in each of the service sub-time slots in the service request and the second resource amount corresponding to each of the target service channels, the method further includes: The resource occupancy stock corresponding to the target user is obtained, and the resource occupancy stock and resource occupancy time are updated to the whitelist corresponding to the target user.

8. The method according to claim 7, characterized in that The method further comprises: If the resource occupation time corresponding to the service request reaches the initial cache resource release time, determine whether there is a same-domain request with the same service type as the service request within the same-domain range corresponding to the service request; If the same-domain request exists, the resource occupancy stock of the target user is scheduled to the target address corresponding to the same-domain request.

9. The method according to claim 8, characterized in that If the same-domain request exists, scheduling the resource usage stock of the target user to the target address corresponding to the same-domain request includes: If the same-domain request exists, the address pointer of the target user is added to the whitelist of the user to which the same-domain request belongs; the address pointer of the target user in the whitelist of the user to which the same-domain request belongs is used to build a dedicated tunnel; The empty segment of the message frame structure of the service request is marked according to a preset special value to obtain a target message frame structure; the target message frame structure is used to guide the user to which the same domain request belongs to access the resource occupation stock.

10. A resource allocation device, characterized in that: The device comprises: A first acquisition module, used to acquire a target user's service request; A construction module, configured to construct a whitelist corresponding to the target user according to the target model and the service request; the whitelist includes a service sub-timeslot corresponding to the service request and a first resource amount corresponding to each service sub-timeslot; The first allocation module is used to analyze the multimodal signal instructions and feedback requirements of the service request according to the target model, the service sub-time slots and the first resource amount corresponding to each service sub-time slot, and determine the target service channel corresponding to the multimodal signal contained in each service sub-time slot in the service request and the second resource amount corresponding to each target service channel.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images