Holographic communication-oriented method and system for collaborative arrangement of perception edge cache and computing resources

Abstract

The application provides a holographic communication-oriented method and system for collaborative arrangement of edge cache and computing resources, and belongs to the technical field of holographic communication. The application realizes the collaborative arrangement between the service unit dependency relationship, the edge cache effectiveness, the request processing path, the computing / backhaul resource state and the cache maintenance cost; the edge processing qualification is determined through the dependency closure and the version effectiveness, so that the accuracy of the edge node service capability judgment is improved; the edge local closed loop path and the cloud collaborative path are combined with the MTP time delay and the display default state, so that the request dynamically selects a more suitable processing path under different resource states, and the time delay fluctuation and the default risk are reduced; through the edge key processing queue, the backhaul competition queue and the cache switching intensity virtual queue, the computing congestion, the backhaul pressure and the cache maintenance intensity can be continuously fed back, so that the edge overload, the backhaul congestion or the frequent cache switching caused by short-sighted decision are avoided.

Description

Technical Field

[0001] This invention relates to the field of holographic communication technology, specifically to a method and system for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication. Background Technology

[0002] With the development of technologies such as 6G, computing networks, and extended reality, holographic communication is gradually becoming an important business model for immersive interaction. Holographic communication provides users with a near-realistic communication experience through 3D visual information, multimodal perception data, and real-time interactive feedback, and can be applied to scenarios such as remote collaboration, immersive conferencing, and smart healthcare. Holographic communication requires a continuous interactive closed loop around user actions, perspective changes, edge processing, and terminal screen synchronization, and is characterized by high bandwidth, low latency, strong computing power, and strong interactivity. Holographic content places high demands on transmission bandwidth, computing power, and caching capabilities. To ensure the smoothness and stability of the immersive experience, holographic services have strict requirements on MTP latency. If the closed-loop latency between user action changes and screen updates is too large or fluctuates significantly, it can easily cause screen desynchronization, interactive stuttering, and even dizziness and interruption of the immersive experience.

[0003] Under the existing network architecture, terminal devices are limited by power consumption, computing power, and storage capacity, making it difficult to independently complete the real-time processing of complex holographic services. While remote clouds possess strong computing and storage capabilities, cross-domain backhaul links introduce long and unstable transmission latency. Therefore, pushing caching, computing, and network resources down to edge nodes closer to the user side has become an important technical direction for reducing MTP latency and improving interaction stability. However, edge node resources are limited, making it impossible to store all service content in the long term or continuously provide sufficient computing power under high concurrency requests. In particular, holographic services are usually completed collaboratively by multiple service units, and there are dependencies between different service units. If a dependent service unit is missing or its version is outdated, even if some content has been cached on the edge node, the request cannot form a complete edge-local processing path. Therefore, the collaborative management of edge caching, computing resources, and backhaul resources for holographic communication has become a crucial issue in the development of related technologies.

[0004] Existing edge service caching, task offloading, and resource allocation joint optimization techniques typically treat services as a whole, focusing on whether services are cached, whether tasks are offloaded to the edge or cloud, and how computing and communication resources are allocated. While these techniques can reduce processing latency and system overhead in general MEC services, they struggle to address situations where multiple service units collaborate in holographic services, where service units have pre-dependencies, and where cached content may be outdated. For holographic requests, a single service hit does not equate to the edge node possessing complete processing capabilities.

[0005] Existing dependency-aware service placement and SFC / VNF orchestration technologies can describe the sequential relationships, call dependencies, and chained deployment paths between service functions, making them suitable for solving network service mapping on physical nodes, resource reservation, and low-latency forwarding issues. However, they typically focus more on service instance placement and service chain path selection, rarely translating dependencies into criteria for determining whether edge caches have effective service capabilities. Simultaneously, there is a lack of a mechanism for establishing the selection between edge and cloud paths for MTP closed-loop latency in holographic communication.

[0006] Existing online optimization and queue-aware resource control technologies with two time scales can handle issues such as inconsistent service placement and task processing update frequencies, random request arrivals, and queue stability. However, their queues are typically used to describe ordinary task backlogs, service migration loads, or computing resource stability. In holographic communication scenarios, cache maintenance itself affects service continuity and long-term operational costs. Frequent loading, refreshing, or removing of service units incurs additional overhead and weakens edge closed-loop service capabilities. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication, so as to solve at least one of the technical problems existing in the background art.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication, comprising:

[0010] Based on the established edge processing eligibility determination model and MTP latency, queue feedback and long-term optimization objectives, edge processing eligibility is obtained for each holographic request in each time slot within the decision frame according to dependency closure determination.

[0011] Select request paths based on edge processing eligibility, determine time slot-level path orchestration; allocate edge computing resources for edge path requests, allocate backhaul resources for cloud collaboration path requests, and satisfy computing and backhaul resource allocation constraints;

[0012] Resource allocation is sorted or iteratively allocated based on the urgency of the MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, low-return requests are prevented from continuing to enter the edge side. For time slots with long backhaul contention queues, backhaul resources for high-latency-sensitive requests are prioritized to avoid creating new transmission bottlenecks in the cloud-to-collaboration path.

[0013] After the time slot-level path orchestration and resource allocation are completed, the states of three types of queues—the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue—are updated based on the actual computational workload entering the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed. The updated states are then fed back to subsequent time slots and subsequent decision frames, forming a dependency closed-loop control process.

[0014] As a further limitation of the first aspect of the present invention, if the edge processing qualification is equal to 0, the corresponding request does not meet the conditions for complete edge processing and cannot select the edge local closed loop path; if the edge processing qualification is equal to 1, the comprehensive cost of the edge local closed loop path and the cloud collaborative path are compared; when the request meets the edge processing qualification, and the comprehensive cost of the edge path is not higher than the comprehensive cost of the cloud path, and the edge computing resources allow, the edge local closed loop path is selected; otherwise, the cloud collaborative path is selected.

[0015] As a further limitation of the first aspect of the present invention, in order to reflect the real-time MTP latency, demonstrate the default risk, and the impact of queue feedback, the comprehensive cost of the edge path and the comprehensive cost of the cloud path are defined as follows:

[0016]

[0017]

[0018] in, and These represent the MTP latency when requesting to select the edge local closed-loop path and the cloud collaborative path, respectively. and These represent the displayed breach status under the corresponding paths; Indicates the amount of computational work required for the request; Indicates the amount of data that needs to be transmitted or returned.

[0019] As a further limitation of the first aspect of the present invention, at the beginning of each decision frame k, a frame-level cache maintenance action is generated based on the current cache residency status, version validity status, service unit dependency relationship, historical or current request requirements, cache capacity constraints, and cache switching intensity virtual queue, thereby prioritizing the maintenance of service units that can improve dependency closure to meet cache status and edge processing qualifications.

[0020] As a further limitation of the first aspect of the present invention, the Lyapunov drift plus penalty method is adopted to transform the long-term stochastic optimization objective into an online decision problem in the current state. The long-term objective is decomposed into two mutually coupled online subproblems with different decision scales: a frame-level cache maintenance subproblem and a slot-level request processing subproblem. The frame-level cache maintenance subproblem is used to determine which service units the edge node should maintain in the current decision frame. The slot-level request processing subproblem is used to determine whether each request chooses the edge local closed-loop path or the cloud collaborative path and completes the corresponding resource allocation.

[0021] As a further limitation of the first aspect of the invention, a Lyapunov function is constructed to characterize the combined backlog state of the edge computing queue, the backhaul queue, and the cache maintenance virtual queue, and a Lyapunov drift term is defined; at each decision moment, instead of directly solving for the long-term objective, the drift plus penalty term in the current state is minimized based on parameters that balance the optimization of real-time service quality and the stability of the queue.

[0022] Secondly, the present invention provides a dependency-aware edge caching and computing resource collaborative orchestration system for holographic communication, comprising:

[0023] The determination module is used to determine edge processing eligibility for each holographic request in each time slot within the decision frame based on the established edge processing eligibility determination model and MTP latency, queue feedback, and long-term optimization objectives, according to dependency closure determination.

[0024] The determination module is used to select request paths based on edge processing eligibility, determine time slot-level path orchestration, allocate edge computing resources for edge path requests, allocate backhaul resources for cloud collaboration path requests, and satisfy computing and backhaul resource allocation constraints.

[0025] The allocation module is used to sort or iteratively allocate resources based on the urgency of the request MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, it suppresses low-return requests from continuing to enter the edge side; for time slots with long backhaul contention queues, it prioritizes backhaul resources for high-latency-sensitive requests to avoid creating new transmission bottlenecks in the cloud-to-cloud collaboration path.

[0026] The update feedback module is used to update the status of three types of queues—edge critical processing queue, backhaul contention queue, and cache switching intensity virtual queue—based on the actual computational workload of the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed after the time slot-level path orchestration and resource allocation are completed. The updated status is then fed back to subsequent time slots and subsequent decision frames to form a dependency closed-loop control process.

[0027] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the dependent-aware edge caching and computing resource collaborative orchestration method for holographic communication as described in the first aspect.

[0028] Fourthly, the present invention provides a computer device including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the collaborative orchestration method of dependent-aware edge caching and computing resources for holographic communication as described in the first aspect.

[0029] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the method for collaborative orchestration of dependent-aware edge caching and computing resources for holographic communication as described in the first aspect.

[0030] Terminology Explanation: Holographic-type Communication (HTC): Holographic communication is a communication technology for immersive interactive applications. It collects, transmits, processes, and presents three-dimensional visual and multimodal interactive information to provide users with a highly immersive, interactive, and multi-dimensional business experience. Motion-to-Photon Latency (MTP): Motion-to-photon latency refers to the closed-loop time span from the occurrence of a user's action or posture change to the completion of calculation, rendering, and display of the corresponding image on the terminal. It is an important indicator for measuring the real-time experience of interactive services such as holographic communication, virtual reality, and extended reality. Mobile Edge Computing (MEC): Mobile edge computing refers to deploying computing, storage, and network resources close to the user or data source, enabling business requests to be processed locally at the edge, thereby reducing backhaul pressure and improving the service quality of low-latency services. Edge Caching (EC): Edge caching refers to pre-storing or dynamically maintaining the data, models, programs, service components, etc., required for business operations at edge nodes. This allows subsequent requests to directly access relevant resources at the edge, reducing cloud access and backhaul transmission overhead. Service Unit (SU): A service unit is a basic functional component or resource object that supports the holographic business processing. It may include rendering modules, encoding / decoding modules, model inference modules, data processing modules, synchronization control modules, etc. A holographic request typically requires multiple service units to work together. Service Function Chain (SFC): A service function chain is a business processing link formed by combining multiple service functions in a specific order or with dependencies. Request traffic must pass through the corresponding service functions sequentially to complete complete processing. Virtual Network Function (VNF): A virtual network function is a network function module implemented in software. It can be deployed on general-purpose computing, storage, and network resources to replace traditional dedicated hardware network functions. Software-Defined Networking (SDN): SDN decouples the control plane from the data plane of network devices, centralizing control logic in an external controller to achieve centralized network management, programmable control, and dynamic resource scheduling. Network Functions Virtualization (NFV): NFV implements the functions of traditional dedicated hardware network devices in software and runs them on general-purpose commercial servers. It decouples network functions into software instances that can be flexibly migrated, elastically scaled, and deployed on demand across general-purpose computing / storage / network resources.Quality of Service (QoS): QoS refers to a comprehensive indicator that measures the service capabilities of a business through objective network and system metrics such as latency, bandwidth, packet loss rate, jitter, and reliability. Quality of Experience (QoE): QoE refers to the subjective quality of a user's experience with a service, typically influenced by factors such as latency, image quality, smoothness of interaction, stability, and service continuity. Drift-Plus-Penalty (DPP) Method: A commonly used method in Lyapunov online optimization, this method transforms a long-running stochastic optimization problem into an online decision-making problem that depends solely on the current system state by jointly considering queue drift and immediate performance costs within each time slot.

[0031] The beneficial effects of this invention are as follows: This invention addresses the MTP closed-loop latency stability control requirements of holographic communication services, achieving collaborative orchestration among service unit dependencies, edge cache effectiveness, request processing paths, computation / backhaul resource status, and cache maintenance costs. By jointly determining edge processing eligibility through dependency closure and version validity, the accuracy of edge node service capability assessment is improved, avoiding edge processing failures due to missing service units, incomplete dependencies, or version invalidation. By combining edge local closed-loop paths and cloud collaborative paths with MTP latency and display default status, requests can dynamically select more suitable processing paths under different resource states, reducing latency fluctuations and default risks. Through edge critical processing queues, backhaul contention queues, and cache switching intensity virtual queues, computational congestion, backhaul pressure, and cache maintenance intensity are continuously fed back, preventing short-sighted decisions that lead to edge overload, backhaul congestion, or frequent cache switching. Through the HCCO online orchestration process coupling frame-level cache maintenance and slot-level request processing, a comprehensive balance between holographic communication service quality, edge resource utilization efficiency, and long-term cache maintenance costs can be achieved without relying on future business information.

[0032] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a structural diagram of the system framework for a dependency-aware edge caching and computing resource collaborative orchestration system for holographic communication as described in an embodiment of the present invention.

[0035] Figure 2 This is a service unit dependency closure and edge processing qualification determination diagram according to an embodiment of the present invention.

[0036] Figure 3 This is a diagram showing the relationship between MTP latency measurement, queue feedback, and long-term optimization objectives as described in an embodiment of the present invention.

[0037] Figure 4 This is a flowchart of the dependency-aware edge caching and computing resource collaborative orchestration method for holographic communication as described in an embodiment of the present invention. Detailed Implementation

[0038] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0039] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0040] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0041] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0042] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0043] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0044] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0045] Holographic communication is an important service model for applications such as immersive interaction, remote collaboration, and digital twins, characterized by high bandwidth, low latency, strong computing power, and strong interactivity. Unlike traditional video transmission and EC services, holographic communication requires tight integration of user motion perception, edge computing processing, service unit collaborative invocation, and terminal screen synchronization. Simultaneously, user experience is highly dependent on stable MTP latency control. Therefore, holographic communication not only demands high-speed network transmission capabilities but also requires the edge to continuously provide stable and predictable real-time processing capabilities.

[0046] This invention addresses the problem that edge nodes in real-world computing-network converged edge environments are constrained by cache capacity, computing resources, and edge-cloud backhaul links. It combines the Lyapunov drift plus penalty approach to design a dependency-aware edge cache and computing resource collaborative orchestration technology (Holographic Cache - Compute Orchestration, HCCO) for holographic communication, which collaboratively completes path selection, resource allocation, and cache maintenance.

[0047] This invention determines whether an edge node has the service capability to fully process a request by characterizing the service units required for a holographic request, their dependencies, cache residency status, and version validity status. It dynamically orchestrates and calculates request paths and allocates backhaul resources based on the differences between the edge-local closed-loop path and the cloud-collaborative path. A queue feedback mechanism and cache maintenance constraints are introduced, enabling the system to adjust online based on edge computing load, backhaul link pressure, and cache switching intensity. In summary, this invention can improve the edge-local closed-loop processing capability of holographic requests under dynamic business loads, reduce MTP latency and display default risks, and balance edge resource utilization efficiency with long-term cache maintenance costs.

[0048] This invention implements edge processing eligibility determination for holographic requests: by comprehensively considering the service units required for the request and their dependencies, edge cache residency status, and version validity status, it determines whether the edge node has the service capability to fully process the request, avoiding misjudgments caused by determining the edge processing path based on whether a single service is cached. It also implements edge-cloud collaborative path orchestration for MTP closed-loop latency: based on the request's edge processing eligibility, edge computing load, edge-cloud backhaul pressure, and business latency requirements, it dynamically selects between the local edge closed-loop path and the cloud collaborative path, reducing MTP latency fluctuations and display default risks for holographic services. Finally, it implements queue feedback control for holographic service operation status: by continuously characterizing edge computing congestion, backhaul link contention, and cache maintenance intensity, path orchestration, resource allocation, and cache maintenance decisions can be adjusted according to system load changes, avoiding resource congestion or excessive maintenance overhead caused by short-term low-latency selection. It achieves a collaborative optimization that combines frame-level cache maintenance with slot-level request processing: by controlling the loading, refreshing and removing of service units at the frame-level time scale, and controlling request path selection, computing resource allocation and backhaul resource allocation at the slot-level time scale, it achieves a comprehensive balance between the quality of holographic communication services, edge resource utilization efficiency and long-term cache maintenance costs.

[0049] Example 1

[0050] In this embodiment 1, a dependency-aware edge caching and computing resource collaborative orchestration system for holographic communication is first provided, including: a determination module, used to determine edge processing eligibility for each holographic request in each time slot within a decision frame based on an established edge processing eligibility determination model and MTP latency, queue feedback, and long-term optimization objectives, according to dependency closure determination; a determination module, used to select request paths based on edge processing eligibility, determine time slot-level path orchestration, allocate edge computing resources for edge path requests, allocate backhaul resources for cloud collaborative path requests, and satisfy computing and backhaul resource allocation constraints; and an allocation module, used to allocate resources based on the urgency of the request MTP time limit, the degree of queue backlog, and unit resources. Source latency benefits are sorted or iteratively allocated for resource distribution. For time slots with long edge computing queues, low-reward requests are suppressed from entering the edge side. For time slots with long backhaul contention queues, backhaul resources for high-latency-sensitive requests are prioritized to avoid creating new transmission bottlenecks in the cloud-to-collaboration path. An update feedback module is used to update the states of three queue types—edge critical processing queue, backhaul contention queue, and cache switching intensity virtual queue—based on the actual computational workload entering the edge path, the amount of backhaul data entering the cloud path, and the actual service capabilities completed after time slot-level path orchestration and resource allocation. The updated states are then fed back to subsequent time slots and subsequent decision frames to form a dependency closed-loop control process.

[0051] In this embodiment, the aforementioned system is used to implement a dependency-aware edge caching and computing resource collaborative orchestration method for holographic communication. This includes: using a determination module based on an established edge processing eligibility determination model and MTP latency, queue feedback, and long-term optimization objectives, for each holographic request in each time slot within a decision frame, edge processing eligibility is determined according to dependency closure determination; using a determination module to select request paths based on edge processing eligibility, determining time slot-level path orchestration; allocating edge computing resources for edge path requests and backhaul resources for cloud collaborative path requests, while satisfying computing and backhaul resource allocation constraints; using an allocation module to sort or iteratively allocate resources based on the urgency of the request's MTP time limit, queue backlog, and unit resource latency benefit; wherein, for time slots with long edge computing queues, low-yield requests are suppressed from continuing to enter the edge side; for time slots with long backhaul contention queues, priority is given to ensuring backhaul resources for high-latency-sensitive requests, avoiding the formation of new transmission bottlenecks in cloud collaborative paths. After the time slot-level path orchestration and resource allocation are completed using the update feedback module, the states of three types of queues—the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue—are updated based on the actual computational workload entering the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed. The updated states are then fed back to subsequent time slots and subsequent decision frames, forming a dependency closed-loop control process.

[0052] Specifically, if the edge processing qualification is 0, the corresponding request does not meet the conditions for complete edge processing and cannot choose the edge local closed-loop path. If the edge processing qualification is 1, the comprehensive cost of the edge local closed-loop path and the cloud collaborative path is compared. When the request meets the edge processing qualification, and the comprehensive cost of the edge path is not higher than the comprehensive cost of the cloud path, and edge computing resources allow, the edge local closed-loop path is selected; otherwise, the cloud collaborative path is selected. At the beginning of each decision frame k, based on the current cache residency status, version validity status, service unit dependencies, historical or current request requirements, cache capacity constraints, and the cache switching intensity virtual queue, frame-level cache maintenance actions are generated to prioritize the maintenance of service units that can improve dependency closure and satisfy cache status and edge processing qualification. The Lyapunov drift plus penalty method is adopted to transform the long-term stochastic optimization objective into an online decision problem in the current state. The long-term objective is decomposed into two coupled online subproblems with different decision scales: the frame-level cache maintenance subproblem and the slot-level request processing subproblem. The frame-level cache maintenance subproblem is used to determine which service units the edge node should maintain in the current decision frame. The slot-level request processing subproblem is used to determine whether each request chooses the edge local closed-loop path or the cloud collaborative path and completes the corresponding resource allocation.

[0053] Specifically, such as Figure 1As shown, this embodiment first proposes a dependency-aware edge caching and computing resource collaborative orchestration framework for holographic communication. This framework, designed for cloud-edge-device collaborative service scenarios, consists of a holographic terminal, edge nodes, edge-cloud backhaul links, a cloud service center, and a collaborative orchestration controller. The holographic terminal continuously generates holographic service requests during user interaction. Edge nodes are responsible for maintaining partial service unit caches, performing edge-side computation processing, and forwarding requests. The cloud service center stores complete service units and their latest versions, and provides collaborative processing capabilities for requests that the edge cannot fully handle. The collaborative orchestration controller generates cache maintenance, path orchestration, and resource allocation decisions based on request status, service dependencies, cache status, version status, computing resources, backhaul resources, and queue status.

[0054] From a functional perspective, the framework of this system can be divided into the following three layers:

[0055] (1) Holographic Service Layer. This layer generates time slots for the interaction between the holographic communication terminal and the user. Holographic request set within Each request Corresponding to several required service unit sets This involves various processing stages, including encoding, decoding, viewpoint prediction, 3D reconstruction, rendering, and synchronization control. Because service units have pre-dependencies, whether a request can be processed at the edge depends not only on whether the directly required service unit is cached, but also on its dependent closed set. Whether it is valid and available on the edge side.

[0056] (2) Cloud-Edge Resource Layer. This includes edge nodes, edge caches, edge computing resources, edge-cloud backhaul links, and cloud service centers. Edge nodes are located in the first... Intra-frame maintenance service unit cache resident status Version validity status and effective availability status and form an effective set of service units. Edge cache capacity is affected Constraints, edge computing resources are subject to Constraints, edge-cloud backhaul resources are subject to Constraints. For requests that qualify for edge processing, the system may choose a local edge closed-loop path; for requests that do not meet the conditions for complete edge processing or are affected by resource constraints, the system may choose a cloud collaborative path.

[0057] (3) Cooperative Orchestration Control Layer. Responsible for executing the buffer-computation cooperative orchestration algorithm for holographic communication. The controller generates buffer maintenance actions at the frame level. This includes service unit loading, refreshing, and removal. Request processing actions are generated at the time slot level. Including path selection variables Edge computing resource allocation and backhaul resource allocation .in, Indicates a request Select the edge local closed-loop path Indicates a request Select a cloud-based collaborative path. The controller also maintains the edge-critical processing queue. , return contention queue And cache switching strength virtual queue It is used to reflect edge computing congestion, backhaul link pressure, and long-term cache maintenance intensity.

[0058] The holographic service layer provides dynamic requests and service unit requirements; the cloud-edge resource layer provides caching, computing, backhaul, and cloud service capabilities; the collaborative orchestration control layer makes online decisions based on business needs and resource status, and applies the decision results to cache maintenance, path selection, and resource allocation processes. In each decision frame... Initially, the system adjusts the set of effective service units on the edge side based on the current cache status, version status, service unit dependencies, and cache maintenance queue. ; in each time slot contained in the decision frame Within, the system is based on the request set. Edge processing qualification Edge computing queue Competition queue with return It dynamically selects the edge local closed-loop path or the cloud collaborative path, and completes the allocation of computing resources and backhaul resources.

[0059] This framework integrates service unit dependency awareness, cache validity maintenance, MTP path orchestration, compute / return resource allocation, and queue feedback control into a unified system structure, providing a foundation for dependency closure modeling, path latency modeling, and HCCO online optimization processes.

[0060] Specifically, the collaborative orchestration control layer includes a global state awareness module, a dependency awareness decision engine, an eligibility determination and path orchestration module, a resource allocation and maintenance control module, and a monitoring and feedback module.

[0061] The global state awareness module is used to collect and integrate the holographic service request state, service unit dependency relationship, edge cache residency state, service version validity state, edge computing resource state, edge-cloud backhaul link state, and queue state to form a globally observable system state under the current decision frame and current time slot, providing input for subsequent cache maintenance, qualification determination, path selection, and resource allocation.

[0062] The dependency-aware decision engine constructs a closed dependency set based on the service units directly required by the request and their preceding dependencies. It then combines this with the cache residency status and version validity of the service units on the edge side to determine whether each service unit is in a valid and available state. This module also evaluates the contribution of different service units to improving edge processing eligibility, providing a basis for decisions regarding frame-level service unit loading, refreshing, and removal.

[0063] The qualification determination and path orchestration module is used to determine whether a holographic request is eligible for edge processing based on the dependency closure satisfaction status, the set of valid edge service units, and edge access and processing conditions. For requests that are not eligible for edge processing, this module orchestrates them to a cloud collaborative path; for requests that are eligible for edge processing, this module further combines the MTP latency of the edge local closed-loop path and the cloud collaborative path, the risk of default, and the queue feedback status to determine the actual processing path of the request.

[0064] The resource allocation and maintenance control module performs service unit cache maintenance actions at the frame level, including service unit loading, refreshing, and removal; and allocates edge computing resources for requests selecting the edge local closed-loop path and edge-cloud backhaul resources for requests selecting the cloud collaborative path at the time slot level. Under the premise of meeting cache capacity, edge computing resource, and backhaul resource constraints, this module orchestrates resources based on request timeliness, queue backlog, and unit resource latency benefits to improve low-latency request processing capabilities.

[0065] The monitoring and feedback module is used to calculate the actual computational workload entering the edge, the amount of data transmitted back to the cloud collaboration path, the service unit cache maintenance intensity, and the request MTP latency and violation status after request processing and resource allocation are completed. Based on this, it updates the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue. The updated queue status is fed back to subsequent time slots and subsequent decision frames, enabling the control layer to form an online closed-loop control process of status awareness, dependency determination, path orchestration, resource allocation, and queue feedback.

[0066] The overall workflow of the collaborative orchestration control layer is as follows: First, the global state awareness module obtains the current business request, service dependencies, cache version, computing resources, backhaul resources, and queue status; second, the dependency awareness decision engine constructs a closed set of dependencies for the request and identifies valid and available service units on the edge side; third, the qualification determination and path orchestration module determines whether the request is qualified for edge processing and selects between the edge local closed-loop path and the cloud collaborative path; subsequently, the resource allocation and maintenance control module performs cache maintenance at the frame level and allocates computing resources and backhaul resources at the time slot level; finally, the monitoring and feedback module updates the three types of queue status based on the actual processing results and uses the feedback results for subsequent cache maintenance, path selection, and resource allocation decisions.

[0067] like Figure 2 As shown, in the cloud-edge-device collaborative framework of this embodiment (i.e., a dependency-aware edge caching and computing resource collaborative orchestration framework for holographic communication), holographic requests are not completed independently by a single service, but are processed collaboratively by multiple service units. Let the set of service units that can be invoked by holographic services be... , of which Each service unit is denoted as In the time slot Within, the set of holographic requests that arrive is denoted as Any request is denoted as .ask The set of service units that need to be directly called is denoted as .

[0068] The holographic service processing includes encoding, decoding, viewpoint prediction, 3D reconstruction, rendering, and synchronization control. Some service units can only be correctly invoked after other service units have met their requirements. Let... Indicates service unit A set of preceding dependent service units. To determine the request... Can a complete processing chain be formed on the edge nodes, and can requests be defined? The dependency closure set is It satisfies:

[0069] (1)

[0070] in, Not only includes requests The directly required service units also include the prerequisite dependency units that these service units need to satisfy during processing. If any critical service unit is unavailable at the edge, the request cannot directly form a complete edge-local closed-loop processing path.

[0071] Edge node caches have limited capacity and cannot reside in all service units for extended periods. Let the... Intra-frame service unit for decision-making The cache residency status is ,when When already cached on edge nodes, ;otherwise, Meanwhile, considering that model parameters, scene data, rendering modules, and synchronization control modules in holographic services update with the service status, a service unit is set up. The version is valid. When its version is valid, ;otherwise, .

[0072] Define service units by combining cache residency status and version validity status. The effective availability status on the edge node is:

[0073] (2)

[0074] If and only if At that time, the service unit It can be effectively invoked on the edge node by the current holographic request. Therefore, the edge node is in the... The set of valid service units within a decision frame is:

[0075] (3)

[0076] Set up service units The cache space occupied is The maximum cache capacity of edge nodes is Then the edge cache capacity constraint is:

[0077] (4)

[0078] In this embodiment, the relationship between cache residency status, version validity status, and valid availability status is shown in Table 1.

[0079] Table 1

[0080]

[0081] Based on this, define the request. In the Dependency closure within each decision frame satisfies the state as follows: :

[0082] (5)

[0083] Equivalent land can also be expressed as:

[0084] (6)

[0085] in, Indicates a request The required service units and their dependent service units are all available and effective on the edge nodes; This indicates that at least one service unit is missing or its version is outdated, and the request cannot directly enter the edge local closed-loop path.

[0086] Further define the request In the time slot Edge processing qualification :

[0087] (7)

[0088] The edge access and processing conditions may include basic conditions such as the request being within the service range of the edge node, the availability of the wireless access link, and the request type allowing edge processing. Edge processing eligibility is used to constrain subsequent path selection variables. :

[0089] (8)

[0090] in, Indicates a request Select the edge local closed-loop path Indicates a request Choose the cloud-based collaborative path. The system will only allow a request to enter the edge-local closed-loop path if the request qualifies for edge processing; otherwise... If so, the request must be redirected to the cloud collaboration path or wait for subsequent cache maintenance actions to improve the edge service status.

[0091] Through the above modeling, the simple judgment of whether a service is cached is extended to an edge processing qualification judgment under the joint constraints of service unit dependency closure, cache residency, version validity, and edge access conditions. The judgment result determines whether the request has the basis for edge local closed-loop processing, and also serves as an input condition for MTP latency modeling, path orchestration, computing resource allocation, and backhaul resource allocation.

[0092] like Figure 3 As shown in this embodiment, for MTP latency measurement, queue feedback, and long-term optimization goals, for any request... The system selects variables through path selection. Determine how to handle it:

[0093] (9)

[0094] And it satisfies the edge processing eligibility constraints given in formula (8). Among them, the edge local closed-loop path refers to the request completing the required service unit call, calculation processing and result return at the edge node; the cloud collaborative path refers to the request needing to be transferred to the cloud service center for collaborative processing via the edge-cloud backhaul link due to insufficient edge service capabilities or high resource competition pressure. The two types of paths correspond to different MTP latency compositions.

[0095] When a request is made to select an edge-local closed-loop path, its MTP delay is denoted as... This mainly includes wireless access latency, edge queuing latency, edge computing processing latency, and result return latency. When a cloud-based collaborative path is requested, its MTP latency is recorded as... This mainly includes wireless access latency, edge-cloud backhaul transmission latency, cloud processing latency, and result return latency. Request The actual MTP latency is uniformly expressed as:

[0096] (10)

[0097] in, Subject to edge computing resource allocation And the impact of edge critical processing queue state, Resource allocation for backhaul The status of the edge-cloud backhaul link and the processing capacity of the cloud are affected. To meet resource constraints, time slots... The internal computing and backhaul resource allocation satisfies:

[0098] (11)

[0099] (12)

[0100] in, Indicates the edge node in the time slot The upper limit of available computing resources, This indicates the upper limit of edge cloud backhaul resources.

[0101] To characterize the time limit requirements for showcasing holographic services, the following is set up: This is the allowed MTP latency threshold. A presentation breach is considered to have occurred when the actual MTP latency of the request exceeds this threshold. The presentation breach status is defined as follows:

[0102] (13)

[0103] in, Indicates a request If screen updates or interactive feedback fail to be completed within the allowed closed-loop time limit, it may cause screen lag, synchronization failure, or a decline in user experience.

[0104] To avoid path selection focusing solely on the immediate low latency of the current time slot while ignoring subsequent resource congestion, this invention further introduces a queue feedback mechanism. This mechanism includes an edge-critical processing queue. , return contention queue And cache switching strength virtual queue The three types of queues are used to characterize edge computing congestion, edge-cloud backhaul link pressure, and long-term cache maintenance intensity, respectively.

[0105] The edge critical processing queue is used to describe the backlog of computational workload entering the edge node after selecting the local closed-loop path at the edge. Let time slots be defined. The workload of entering the edge processing queue is The service volume provided by the edge node in this time slot is Then the queue is updated as follows:

[0106] (14)

[0107] in, The computational workload can be accumulated from the requests that select edge paths. This can be obtained from the computing service capabilities actually allocated and completed by the edge nodes. When When the value is large, it indicates that the edge nodes are already under high processing pressure, and subsequent path orchestration should not blindly increase edge processing requests.

[0108] The backhaul contention queue is used to describe the backlog of data entering the edge-cloud backhaul link after selecting a cloud-collaboration path. (Time slots are defined.) The amount of data entering the return queue is The service volume provided by the backhaul link in this time slot is Then the queue is updated as follows:

[0109] (15)

[0110] in, It can be obtained by accumulating the amount of data returned from requests that selected the cloud collaboration path. This can be obtained from the actual allocated backhaul link service capacity. When When the value is large, it indicates that the cloud-based collaborative path will introduce a high backhaul waiting latency, and the system needs to reduce the risk of backhaul congestion in path selection and bandwidth allocation.

[0111] The cache switching intensity virtual queue is used to characterize the long-term maintenance pressure caused by loading, refreshing, and removing service units. Let the... The cache maintenance strength or maintenance cost within each decision frame is The upper limit of the long-term average cache maintenance strength constraint is The long-term cache maintenance strength is met;

[0112] (16)

[0113] The virtual queue is then updated as follows:

[0114] (17)

[0115] when A large value indicates that recent cache maintenance actions have approached or exceeded long-term constraint levels. Subsequent frame-level cache maintenance should reduce unnecessary loading, refreshing, and removing actions to avoid excessive maintenance costs or frequent fluctuations in service status. Table 2 shows the feedback effects of the three types of queues.

[0116] Table 2

[0117]

[0118] Based on MTP latency measurement and queue feedback, this embodiment unifies cache maintenance, path orchestration, computing resource allocation, and backhaul resource allocation into a long-term stochastic optimization problem. This aims to minimize long-term average MTP latency and presentation default risk, while controlling long-term cache maintenance costs, under conditions of dynamic request arrival, service version changes, and link state fluctuations. The long-term optimization objective can be expressed as:

[0119] (18)

[0120] in, This indicates a frame-level buffer maintenance action. Indicates a slot-level request processing action; This indicates the display of the penalty weight for breach of contract. This represents the cache maintenance cost weight. The objective is jointly constrained by the cache capacity constraint in formula (4), the edge processing eligibility constraint in formulas (7-9), the computing resource constraint in formula (11), the backhaul resource constraint in formula (12), and the long-term cache maintenance strength constraint in formula (16).

[0121] The aforementioned optimization issues reflect the core control objectives: improving the availability of edge-local closed-loop paths through dependency closure judgments, reducing MTP latency and display default risks through path orchestration and resource allocation, and avoiding long-term imbalances in edge computing, edge-cloud backhaul, and cache maintenance through three types of queue feedback.

[0122] Based on the established edge processing eligibility model and the objectives of MTP latency, queue feedback, and long-term optimization, this embodiment further designs a dependency-aware cache-compute orchestration (HCCO) online orchestration algorithm for holographic communication. For dynamic holographic service requests, it makes online decisions solely based on the currently observable system state, without relying on future request arrivals, service version changes, or link state information.

[0123] HCCO employs a two-timescale structure. One decision frame. It contains multiple time slots, and the time slot set is as follows:

[0124]

[0125] in, This indicates the number of time slots contained within a decision frame. Frame-level scales are primarily used to control cache maintenance actions such as service unit loading, refreshing, and removing; time slot-level scales are primarily used to control request path selection, edge computing resource allocation, and edge-cloud backhaul resource allocation. This structure can avoid frequent service unit cache switching while maintaining the system's rapid response capability to request arrivals and resource status changes.

[0126] In the Each decision frame and its internal time slot In the HCCO observation, the system states include:

[0127] (19)

[0128] in, , , These represent the service unit cache residency state vector, version validity state vector, and valid availability state vector, respectively. This represents the set of currently valid service units for the edge nodes; This represents the set of holographic requests arriving in the current time slot; , , These represent the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue, respectively. and These represent the current upper limit of edge computing resources and the upper limit of edge-cloud backhaul resources, respectively.

[0129] HCCO's decision-making actions are divided into frame-level cache maintenance actions and slot-level request processing actions. Frame-level cache maintenance actions are represented as follows:

[0130] (20)

[0131] in, , , These indicate whether or not the service unit is... Perform load, refresh, and remove operations. Slot-level request processing actions are represented as follows:

[0132] (twenty one)

[0133] in, Choose variables for the path. To assign to the request Edge computing resources To assign to the request Edge cloud backhaul resources.

[0134] Table 3 shows the main states and actions of HCCO.

[0135] Table 3

[0136]

[0137] Since the aforementioned long-term optimization problem involves future service arrivals, service version changes, and link status, direct offline solutions are difficult to implement in real-world systems. Therefore, HCCO employs the Lyapunov drift-penalty method to transform the long-term stochastic optimization objective into an online decision-making problem under the current state.

[0138] Constructing Lyapunov functions:

[0139] (twenty two)

[0140] in, This is used to characterize the combined backlog state of edge computing queues, return queues, and cache maintenance virtual queues. The Lyapunov drift term is defined as:

[0141] (twenty three)

[0142] HCCO does not directly solve for the long-term objective at each decision point, but instead minimizes the drift plus penalty term in the current state:

[0143] in, These are control parameters used to balance optimizing real-time service quality with queue stability. For immediate penalties, it can be represented as:

[0144] (twenty four)

[0145] in, For the request The actual MTP latency, To demonstrate the status of the breach, For the first Buffer maintenance strength within each decision frame To demonstrate the weighting of penalties for breach of contract, The cost weight for cache maintenance.

[0146] Through the above transformation, HCCO decomposes the long-term goal into two coupled online sub-problems with different decision scales. The frame-level cache maintenance sub-problem determines which service units the edge node should maintain within the current decision frame. The slot-level request processing sub-problem determines whether each request should choose an edge-local closed-loop path or a cloud-collaborative path and completes the corresponding resource allocation.

[0147] In each decision frame Initially, HCCO determines the cache residency status based on the current cache state. Version validity status Service unit dependencies, historical or current request requirements, and cache capacity constraints. and cache switching intensity virtual queue Generate frame-level cache maintenance actions Therefore, priority should be given to maintaining states that improve dependency closure satisfaction. and edge processing qualifications The service units include:

[0148] (1) Loading action. For those not cached on the edge but frequently appearing in the dependency closure set. If a service unit in the cache can fulfill the dependency closure conditions of multiple requests after loading, and the cache capacity allows, then the loading action will be performed first. .

[0149] (2) Refresh Action. For service units that are already cached on the edge but whose version is outdated, if the service unit is a critical dependency unit of the current or recent requests, the refresh action will be performed first. This will restore it to a valid and usable state.

[0150] (3) Removal action. For service units that do not participate in dependency closure for a long time, contribute little to improving edge processing eligibility, or occupy a large amount of cache space, a removal action can be performed when capacity is insufficient. .

[0151] Frame-level buffer maintenance actions should satisfy the buffer capacity constraints given by formula (4). When A larger value indicates that recent cache maintenance has been intensive, and the algorithm should reduce unnecessary loading, refreshing, and removal operations. When the cache size is small, the algorithm can appropriately increase the maintenance of critical service units to improve the eligibility for edge processing of requests in subsequent time slots. After cache maintenance is complete, the system is updated. , , and This provides a foundation for effective service capabilities for path orchestration in each time slot within a frame.

[0152] In decision frame Each time slot within HCCO based on the current request set Effective service unit set Edge processing qualification Edge critical processing queue , return contention queue Edge computing resources and return resources Generate time slot-level request processing actions .

[0153] For each request HCCO is first obtained based on the dependency closure criterion. .like If the request does not meet the conditions for complete edge processing, the system must not select the edge local closed-loop path, that is: .like The system then further compares the combined costs of the edge local closed-loop path and the cloud collaborative path. To reflect real-time MTP latency, demonstrate default risk, and the impact of queue feedback, the combined costs of the edge path and the cloud path are defined as follows:

[0154] (25)

[0155] (26)

[0156] in, and These represent the MTP latency when requesting to select the edge local closed-loop path and the cloud collaborative path, respectively. and These represent the displayed breach status under the corresponding paths; Indicates the amount of computational work required for the request; Indicates the amount of data that needs to be transmitted or returned.

[0157] When a request qualifies for edge processing, and the overall cost of the edge path is not higher than the overall cost of the cloud path, and edge computing resources allow, HCCO selects the local edge closed-loop path; otherwise, it selects the cloud collaborative path. The path selection rule in formula (9) can be further expressed as:

[0158] (27)

[0159] After the path selection is completed, HCCO allocates edge computing resources for edge path requests and backhaul resources for cloud collaborative path requests, and satisfies the computing and backhaul resource allocation constraints given by formulas (11) and (12).

[0160] Resource allocation can be sorted or iteratively allocated based on the urgency of the request MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. For time slots with long edge computing queues, the algorithm will suppress low-reward requests from continuing to enter the edge side; for time slots with long backhaul contention queues, the algorithm will prioritize ensuring backhaul resources for high-latency-sensitive requests, avoiding the formation of new transmission bottlenecks in the cloud-to-cloud collaborative path.

[0161] After time-slot-level path orchestration and resource allocation are completed, the system updates the three queue states according to formulas (14), (15), and (17) based on the actual computational workload of the edge path, the amount of data transmitted back to the cloud path, and the actual service capacity completed. The updated states are as follows: , and Feedback will be sent to subsequent time slots and subsequent decision frames, enabling the system to form an online closed-loop control process of dependency closure judgment, cache maintenance, path orchestration, resource allocation, and queue feedback.

[0162] In summary, combining Figure 4 As shown, the HCCO algorithm execution flow proposed in this embodiment mainly includes the following steps:

[0163] Input: Set of service units Dependency Cache capacity Edge computing resources Return resources MTP threshold , maintain constraints Control parameters Displaying default weights Maintenance cost weight .

[0164] Initialization: Cache resident state Version valid status Edge critical processing queue Return to the contention queue Cache switching strength virtual queue .

[0165] For each decision frame :

[0166] 1. Observation , , And service unit dependencies;

[0167] 2. Based on cache capacity, version validity, dependency closure benefits, and Generate frame-level cache maintenance actions ;

[0168] 3. Perform service unit loading, refreshing, and removal;

[0169] 4. Update the available status. and effective service unit set ;

[0170] For each time slot :

[0171] 5. Observation Request Set , , , , ;

[0172] 6. For each request Construct dependent closed sets ;

[0173] 7. According to judge Further obtained ;

[0174] 8. If Then let ;like Then calculate and And determine in combination with resource constraints ;

[0175] 9. Regarding Request to allocate edge computing resources ;right Request to allocate return resources ;

[0176] 10. Calculate and display the status of default ;

[0177] 11. Update and ;

[0178] 12. Based on the maintenance intensity of this frame renew ;

[0179] 13. Proceed to the next decision frame.

[0180] Output: Frame-level buffer maintenance actions Time slot-level path selection and resource allocation actions .

[0181] Through the above process, HCCO can maintain the effective service capabilities of edge nodes at the frame level, respond to dynamic holographic requests and resource changes at the time slot level, and continuously provide feedback on system load and maintenance pressure through three types of queues. The closure mechanism improves the efficiency of converting cache sets into edge processing qualifications, the queue feedback mechanism avoids edge computing congestion or backhaul congestion caused by short-sighted path selection, and the cache switching intensity virtual queue controls the long-term maintenance costs caused by frequent loading, refreshing, and removing of service units. Therefore, this invention can achieve a comprehensive balance between holographic communication MTP latency, display default risk, edge resource utilization, and cache maintenance costs in a dynamic computing network converged edge environment.

[0182] Example 2

[0183] This embodiment 2 provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they implement the aforementioned method for collaborative orchestration of dependent-aware edge caching and computing resources for holographic communication. This method includes:

[0184] Based on the established edge processing eligibility determination model and MTP latency, queue feedback and long-term optimization objectives, edge processing eligibility is obtained for each holographic request in each time slot within the decision frame according to dependency closure determination.

[0185] Select request paths based on edge processing eligibility, determine time slot-level path orchestration; allocate edge computing resources for edge path requests, allocate backhaul resources for cloud collaboration path requests, and satisfy computing and backhaul resource allocation constraints;

[0186] Resource allocation is sorted or iteratively allocated based on the urgency of the MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, low-return requests are prevented from continuing to enter the edge side. For time slots with long backhaul contention queues, backhaul resources for high-latency-sensitive requests are prioritized to avoid creating new transmission bottlenecks in the cloud-to-collaboration path.

[0187] After the time slot-level path orchestration and resource allocation are completed, the states of three types of queues—the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue—are updated based on the actual computational workload entering the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed. The updated states are then fed back to subsequent time slots and subsequent decision frames, forming a dependency closed-loop control process.

[0188] Example 3

[0189] This embodiment 3 provides a computer device, including a memory and a processor. The processor and the memory communicate with each other. The memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute the aforementioned method for collaborative orchestration of dependent-aware edge caching and computing resources for holographic communication. The method includes:

[0190] Based on the established edge processing eligibility determination model and MTP latency, queue feedback and long-term optimization objectives, edge processing eligibility is obtained for each holographic request in each time slot within the decision frame according to dependency closure determination.

[0191] Select request paths based on edge processing eligibility, determine time slot-level path orchestration; allocate edge computing resources for edge path requests, allocate backhaul resources for cloud collaboration path requests, and satisfy computing and backhaul resource allocation constraints;

[0192] Resource allocation is sorted or iteratively allocated based on the urgency of the MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, low-return requests are prevented from continuing to enter the edge side. For time slots with long backhaul contention queues, backhaul resources for high-latency-sensitive requests are prioritized to avoid creating new transmission bottlenecks in the cloud-to-collaboration path.

[0193] After the time slot-level path orchestration and resource allocation are completed, the states of three types of queues—the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue—are updated based on the actual computational workload entering the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed. The updated states are then fed back to subsequent time slots and subsequent decision frames, forming a dependency closed-loop control process.

[0194] Example 4

[0195] This embodiment 4 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes instructions to implement the above-described method for collaborative orchestration of dependent-aware edge caching and computing resources for holographic communication. The method includes:

[0196] Based on the established edge processing eligibility determination model and MTP latency, queue feedback and long-term optimization objectives, edge processing eligibility is obtained for each holographic request in each time slot within the decision frame according to dependency closure determination.

[0197] Select request paths based on edge processing eligibility, determine time slot-level path orchestration; allocate edge computing resources for edge path requests, allocate backhaul resources for cloud collaboration path requests, and satisfy computing and backhaul resource allocation constraints;

[0198] Resource allocation is sorted or iteratively allocated based on the urgency of the MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, low-return requests are prevented from continuing to enter the edge side. For time slots with long backhaul contention queues, backhaul resources for high-latency-sensitive requests are prioritized to avoid creating new transmission bottlenecks in the cloud-to-collaboration path.

[0199] After the time slot-level path orchestration and resource allocation are completed, the states of three types of queues—the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue—are updated based on the actual computational workload entering the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed. The updated states are then fed back to subsequent time slots and subsequent decision frames, forming a dependency closed-loop control process.

[0200] In summary, the dependency-aware edge caching and computing resource collaborative orchestration method for holographic communication described in this embodiment of the invention proposes a cloud-edge-device collaborative orchestration architecture for holographic communication, including a holographic terminal, edge nodes, edge-cloud backhaul links, a cloud service center, and a collaborative orchestration controller. The holographic terminal is responsible for generating holographic requests; the edge nodes are responsible for maintaining service unit caches, performing edge computing processing, and handling request forwarding; the cloud service center is responsible for storing complete service units and their latest versions, and providing collaborative processing capabilities for requests that cannot be fully processed by the edge side; the collaborative orchestration controller generates cache maintenance, path orchestration, and resource allocation decisions based on request status, service dependencies, cache status, version status, computing resources, backhaul resources, and queue status; reflecting the overall system organization of holographic communication and the cloud-edge-device collaborative working logic. This paper innovatively proposes an edge processing eligibility determination mechanism based on service unit dependency closure: by representing a holographic request as a processing procedure completed collaboratively by multiple service units, a dependency closure set is formed based on the service units directly required by the request and their preceding dependencies. Combined with the cache residency status, version validity status, and availability status of edge nodes, the mechanism determines whether the request is eligible to enter the edge local closed-loop path. This avoids incorrect edge processing path selection due to missing dependent service units, incomplete service combinations, or version failures. Furthermore, this paper innovatively proposes an edge local closed-loop path and cloud collaborative path orchestration mechanism oriented towards MTP closed-loop latency: based on edge processing eligibility determination, the request processing path is divided into edge local closed-loop paths and cloud collaborative paths. The MTP latency and display default status of the request are calculated based on the latency components of wireless access, edge computing, edge-cloud backhaul, cloud processing, and result return under both types of paths. Path selection is directly linked to holographic communication interaction experience indicators, shifting the system optimization goal from ordinary task completion latency to the control of MTP closed-loop latency and display default risk. This paper innovatively proposes a queue feedback control mechanism for edge computing, edge-cloud backhaul, and cache maintenance. This mechanism includes an edge critical processing queue, a backhaul contention queue, and a virtual queue for cache switching intensity. These queues are used to characterize edge computing congestion and local processing backlog, edge-cloud backhaul link pressure and cloud-to-cloud collaborative path congestion, and to constrain the long-term maintenance intensity caused by service unit loading, refreshing, and removal. This allows path selection, resource allocation, and cache maintenance to be adjusted based on the long-term operating status of the system, avoiding excessive edgeification, excessive cloud collaboration, and frequent cache switching. Furthermore, this paper innovatively proposes a long-term stochastic joint optimization model for holographic communication. The model optimizes long-term average MTP latency, display default risk, and cache maintenance costs, while being constrained by cache capacity, edge processing eligibility, edge computing resources, edge-cloud backhaul resources, and long-term cache maintenance intensity. This model achieves unified modeling of holographic service quality, edge resource utilization efficiency, and cache maintenance costs, providing an objective function and constraint basis for online orchestration algorithms.We innovatively propose a dependency-aware cache-computation cooperative online orchestration algorithm (HCCO): it adopts a two-timescale structure that is coupled with frame-level cache maintenance and slot-level request processing. After the request is processed, the relevant queues are updated and the queue status is fed back to subsequent decisions, thereby transforming the long-term stochastic optimization problem into an online closed-loop control process that depends on the current state.

[0201] In practical applications, different data structures or computational methods can be used to implement service unit dependency closure and edge processing eligibility determination mechanisms. Service units and their dependencies can be represented as directed dependency graphs, adjacency matrices, dependency tables, or pre-computed closed sets. Topological traversal, matrix operations, or table lookup methods can be used to determine whether the requested service units and their preceding dependencies are all in an edge-available state, thereby completing dependency closure determination, version validity verification, and edge processing eligibility determination. For the HCCO two-timescale online orchestration process, different online decision-making methods can be used. Frame-level cache maintenance can use dependency benefit sorting, rolling window prediction, or heuristic greedy selection to determine service unit loading, refreshing, and removal. Slot-level path orchestration and resource allocation can use queue-weighted cost comparison, model predictive control, or reinforcement learning strategies to generate edge / cloud path selection and computation, and transmit resource allocation results back, thus also achieving the goal of holographic communication cache-computation collaborative orchestration.

[0202] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0203] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0204] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0205] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0206] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication, characterized in that, include: Based on the established edge processing eligibility determination model and MTP latency, queue feedback and long-term optimization objectives, edge processing eligibility is obtained for each holographic request in each time slot within the decision frame according to dependency closure determination. Select the request path based on edge processing eligibility and determine the slot-level path orchestration; Allocate edge computing resources for edge path requests and backhaul resources for cloud collaboration path requests, while satisfying the constraints on computing and backhaul resource allocation; Resource allocation is sorted or iteratively allocated based on the urgency of the MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, low-return requests are prevented from continuing to enter the edge side. For time slots with long backhaul contention queues, backhaul resources for high-latency-sensitive requests are prioritized to avoid creating new transmission bottlenecks in the cloud-to-collaboration path. After the time slot-level path orchestration and resource allocation are completed, the states of three types of queues—the edge critical processing queue, the backhaul contention queue, and the cache switching intensity virtual queue—are updated based on the actual computational workload entering the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed. The updated states are then fed back to subsequent time slots and subsequent decision frames, forming a dependency closed-loop control process.

2. The method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication according to claim 1, characterized in that, If the edge processing qualification is 0, the corresponding request does not meet the conditions for complete edge processing and cannot choose the edge local closed-loop path; if the edge processing qualification is 1, the comprehensive cost of the edge local closed-loop path and the cloud collaborative path are compared; when the request meets the edge processing qualification, and the comprehensive cost of the edge path is not higher than the comprehensive cost of the cloud path, and the edge computing resources allow, the edge local closed-loop path is selected; otherwise, the cloud collaborative path is selected.

3. The method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication according to claim 2, characterized in that, To reflect real-time MTP latency, demonstrate default risk, and assess the impact of queue feedback, the comprehensive cost of the edge path and the comprehensive cost of the cloud path are defined as follows: in, and These represent the MTP latency when requesting to select the edge local closed-loop path and the cloud collaborative path, respectively. and These represent the displayed breach status under the corresponding paths; Indicates the amount of computational work required for the request; Indicates the amount of data that needs to be transmitted or returned.

4. The method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication according to claim 1, characterized in that, At the beginning of each decision frame k, a frame-level cache maintenance action is generated based on the current cache residency status, version validity status, service unit dependencies, historical or current request requirements, cache capacity constraints, and the cache switching intensity virtual queue. This prioritizes maintaining service units that can improve dependency closure, satisfy cache status, and qualify for edge processing.

5. The method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication according to claim 1, characterized in that, The Lyapunov drift plus penalty method is adopted to transform the long-term stochastic optimization objective into an online decision problem in the current state. The long-term objective is decomposed into two coupled online subproblems with different decision scales: the frame-level cache maintenance subproblem and the slot-level request processing subproblem. The frame-level cache maintenance subproblem is used to determine which service units the edge node should maintain in the current decision frame. The slot-level request processing subproblem is used to determine whether each request chooses the edge local closed-loop path or the cloud collaborative path and completes the corresponding resource allocation.

6. The method for collaborative orchestration of dependency-aware edge caching and computing resources for holographic communication according to claim 1, characterized in that: Construct a Lyapunov function to characterize the combined backlog state of the edge computing queue, the backhaul queue, and the cache maintenance virtual queue, and define the Lyapunov drift term; at each decision time, instead of directly solving for the long-term objective, minimize the drift plus penalty term in the current state based on parameters that balance the optimization of real-time service quality and queue stability.

7. A dependency-aware edge caching and computing resource collaborative orchestration system for holographic communication, characterized in that, include: The determination module is used to determine edge processing eligibility for each holographic request in each time slot within the decision frame based on the established edge processing eligibility determination model and MTP latency, queue feedback, and long-term optimization objectives, according to dependency closure determination. The determination module is used to select the request path based on edge processing eligibility and determine the slot-level path orchestration. Allocate edge computing resources for edge path requests and backhaul resources for cloud collaboration path requests, while satisfying the constraints on computing and backhaul resource allocation; The allocation module is used to sort or iteratively allocate resources based on the urgency of the request MTP time limit, the degree of queue backlog, and the latency benefit per unit of resource. Specifically, for time slots with long edge computing queues, it suppresses low-return requests from continuing to enter the edge side; for time slots with long backhaul contention queues, it prioritizes backhaul resources for high-latency-sensitive requests to avoid creating new transmission bottlenecks in the cloud-to-cloud collaboration path. The update feedback module is used to update the status of three types of queues—edge critical processing queue, backhaul contention queue, and cache switching intensity virtual queue—based on the actual computational workload of the edge path, the amount of data transmitted back to the cloud path, and the actual service capabilities completed after the time slot-level path orchestration and resource allocation are completed. The updated status is then fed back to subsequent time slots and subsequent decision frames to form a dependency closed-loop control process.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the collaborative orchestration method for dependent-aware edge caching and computing resources for holographic communication as described in any one of claims 1-6.

9. A computer device, characterized in that, The method includes a memory and a processor, the processor and the memory communicating with each other, the memory storing program instructions executable by the processor, and the processor calling the program instructions to execute the collaborative orchestration method of dependent-aware edge caching and computing resources for holographic communication as described in any one of claims 1-6.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions that implement the method for collaborative orchestration of dependent-aware edge caching and computing resources for holographic communication as described in any one of claims 1-6.

Images