Portal side cross existing network joint arrangement method and system for point cloud video call MTP latency risk control

Abstract

This invention provides an entry-side cross-network joint orchestration method and system for MTP latency risk control in point cloud video calls, belonging to the field of holographic video communication technology. This invention introduces candidate logical path abstraction for the existing network black-box bearer environment, enabling the entry-side to complete bearer selection for point cloud video calls without relying on the existing network's link-by-link opening, reducing the cost of cross-network deployment and control modification. It incorporates service admission, path selection, and burst release into the same entry-side decision, improving the problem of fragmented routing, admission, and shaping in traditional solutions. The closed-loop feedback model of MTP latency, MTP margin, and path risk latency shifts the optimization target from average network indicators to the latency default risk that real-time interactive services are more sensitive to, thereby improving the latency guarantee capability of point cloud video calls. It also improves the effective service quality of accepted services and the real-time holographic interactive experience under high-load scenarios.

Description

Technical Field

[0001] This invention relates to the field of holographic video communication technology, specifically to an entry-side cross-network joint orchestration method and system for MTP latency risk control in point cloud video calls. Background Technology

[0002] With the development of technologies such as mobile communication, 3D content acquisition, immersive display, and edge computing, communication services are evolving from traditional voice and 2D video to immersive interactive services. Holographic-type communication (HTC), targeting scenarios such as remote conferencing, remote collaboration, immersive entertainment, and digital twins, is gradually becoming one of the typical service directions of 6G. Related white papers and research all agree that holographic communication, immersive interaction, and 3D reality experiences will drive the evolution of communication networks from traditional "connectivity-based services" to "experience-based services."

[0003] Point cloud video calling is one of the representative business models in the implementation of holographic communication. Point clouds can represent the geometric shape, color texture, and spatial posture of people or objects as discrete 3D point sets, providing richer 3D spatial information compared to traditional 2D video. Point cloud video calling uses dynamic point clouds as an interactive medium, enabling remote users to participate in the communication process in 3D, and is suitable for applications such as immersive meetings, remote teaching, remote collaboration, and interactive entertainment.

[0004] While enhancing the immersive experience, point cloud video calls also place higher demands on network capacity. On the one hand, point cloud video needs to transmit 3D geometry, attribute texture, and temporal change information, resulting in a significantly higher data volume than traditional video services. On the other hand, factors such as user perspective switching, changes in body movements, and variations in scene complexity can cause fluctuations in the instantaneous bitrate. For real-time interactive services, MTP (Motion-to-Photon) latency is a crucial factor affecting the interactive experience. If the MTP latency is too high, users will experience lag in screen response, thus impacting immersion and the naturalness of interaction. Existing research on holographic communication generally indicates that high data volume, low latency, low jitter, and strong interactivity are the main network challenges faced by holographic services.

[0005] In practical network deployments, point cloud video calls typically need to traverse the existing core network. The existing network is characterized by multiple domains, multiple devices, existing routing policies, and management boundaries, making it difficult to rely on fully granular hop-by-hop control for network orchestration. Therefore, how to reasonably adjust the way services enter the core network based on service characteristics, candidate path-level resource feedback, and quality of service feedback, under the limited controllability of the existing network, becomes a crucial issue in point cloud video call service delivery. The key to this issue lies in the fact that the scale of service access affects network load, path selection affects propagation latency and queuing status, and the ingress traffic pattern affects the impact of sudden service surges on the core link. Considering these factors synergistically helps reduce the risk of MTP latency defaults during point cloud video call service operation.

[0006] Existing point cloud video transmission and holographic communication technologies primarily address the issues of data representation, compression encoding, adaptive transmission, and quality of experience assurance for 3D point cloud media. However, they mainly focus on encoder selection, bitrate adjustment, chunked transmission, or delivery of user-perspective content, thus primarily affecting the media transmission link on the endpoint or application side. For cross-network transmission scenarios, the existing network often lacks fine-grained hop-by-hop control by the service orchestration system, only able to expose candidate path-level resource summaries or quality feedback to the entry point. Therefore, simply relying on point cloud compression and adaptive bitrate is insufficient to solve the joint orchestration problem of "whether to accept, which candidate path to take, and at what shaping rate to enter the network" before a service enters the network. It also struggles to directly address MTP latency default risks and establish closed-loop network-side control.

[0007] Existing cross-network resource orchestration technologies primarily address resource organization, service instance creation, path policy coordination, and end-to-end quality assurance when services cross existing networks, cross-domain networks, or shared infrastructure. However, these technologies tend to focus more on macro-level resource abstraction, service chain construction, and end-to-end slice management. When hop-by-hop control is not possible within the existing network, the ingress side cannot achieve feedback based on a limited number of candidate paths, cannot determine admission, path, and shaping rate, and cannot incorporate MTP latency margin, path risk latency, and ingress shaping queue status into a continuous feedback loop.

[0008] Existing traffic engineering and path control methods optimize network operation under given network topology, service requirements, and resource constraints through path selection, load balancing, and resource scheduling. However, they often handle path selection, admission control, or traffic shaping separately, and their optimization objectives are mostly focused on link utilization, average latency, throughput, or slicing benefits. For bursty, low-latency sensitive services like point cloud video calls, rules and strategies such as fixed paths, minimum load paths, or maximum feasible shaping tend to separate admission, path, and shaping actions, making it difficult to coordinate control over MTP latency default risks. Therefore, an ingress-side joint orchestration mechanism is needed to incorporate whether to accept, candidate path binding, and ingress shaping rate configuration into the same decision-making process when a service request arrives, and to continuously adjust the strategy through service quality feedback. Summary of the Invention

[0009] The purpose of this invention is to provide an entry-side cross-network joint orchestration method and system for MTP latency risk control in point cloud video calls, so as to solve at least one of the technical problems existing in the background art.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] In a first aspect, the present invention provides an ingress-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls, comprising:

[0012] When the point cloud video call service arrives at the entry access point, the average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node and exit node characteristics of the current service are extracted to provide service-side information for subsequent state construction and action legality judgment.

[0013] By integrating current service request characteristics, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information, a state vector that can be input for the joint orchestration strategy is constructed.

[0014] Based on the state vector, a pre-trained joint orchestration strategy model is used to output a joint action; the joint action includes at least an admission control variable, a candidate logical path identifier, and an entry shaping rate level.

[0015] Determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level; if the service is rejected, or the acceptance action violates the bandwidth quota constraints, the service will not enter the live network; if the acceptance is valid, the service will be bound to the selected candidate logical path.

[0016] Maintain a flow-level cache queue for each accepted service, and limit the actual output rate of the service into the live network black box according to the configured ingress shaping rate.

[0017] As a further limitation of the first aspect of the present invention, training the joint orchestration strategy based on the PPO update strategy includes: before training, an event-driven joint orchestration environment is constructed based on the existing network abstract model, candidate path set, point cloud service generation model, ingress shaping rate, MTP latency model, and reward function. This environment uses the arrival of service requests as the decision trigger event and advances the queue, path occupancy, and MTP latency status of accepted services using physical time slices. In one training round, the network state, active service set, and remaining bandwidth quota of the path are first reset; then, point cloud video call requests are generated according to the service arrival model; when a request arrives, a state is constructed, and an action is output according to the state. After the action is executed, it is determined whether the action is legal; legally accepted services are added to the active service set and configured with paths and shaping rates, while rejected or illegally accepted services do not enter the existing network.

[0018] As a further limitation of the first aspect of the present invention, after the current decision and before the next business request arrives, the environment continuously updates the business burst input, entry queue, shaped output, actual path occupancy, MTP latency and MTP margin according to physical time slices, and calculates the reward of the current decision, thereby forming training samples for PPO to update the policy network and value network.

[0019] As a further limitation of the first aspect of the present invention, during PPO update, the policy network is used to output the probability distribution of joint actions, and the value network is used to estimate the state value and calculate the advantage function; PPO limits the change in the probability ratio between the old and new policies by pruning the target, so that the policy update will not cause excessive oscillation due to a single round of samples; at the same time, the value function loss improves the accuracy of reward estimation, and the entropy regularization term maintains a certain exploration capability.

[0020] As a further limitation of the first aspect of the present invention, the optimization objective of the joint orchestration strategy is to maximize the cumulative return during the long-term business arrival process, thereby taking into account the acceptance scale, path load, ingress shaping and MTP risk control:

[0021] ;

[0022] In the formula, Indicates a joint orchestration strategy, This represents the long-term discount factor across decision-making rounds. This indicates the number of decisions made within a single round.

[0023] As a further limitation of the first aspect of the invention, joint orchestration is not re-decided for each physical time slice, but is triggered at the moment the service request arrives, assuming the first... The time of arrival of this service request is The system constructs a state at this moment. and output the action. The actions include acceptance, pathing, and reshaping configuration; between the arrival times of two adjacent requests, the system follows... Each physical time slice continuously updates burst input rate, ingress queue and shaped output, path network risk latency, and active service MTP latency.

[0024] Secondly, this invention provides an ingress-side cross-network joint orchestration system for MTP latency risk control in point cloud video calls, comprising:

[0025] The business request feature extraction module is used to extract the average bit rate, duration, MTP latency threshold, business value, burst probability, burst factor, entry node and exit node features of the current business when the point cloud video call business arrives at the entry access point, so as to provide business-side information for subsequent state construction and action legality judgment.

[0026] The state awareness and feature construction module is used to integrate current service request features, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information to construct a state vector that can be input to the joint orchestration strategy.

[0027] The joint orchestration module is used to output joint actions based on the state vector using a pre-trained joint orchestration strategy model; the joint actions include at least the admission control variable, candidate logical path identifier, and entry shaping rate level.

[0028] The action execution and legality check module is used to determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level. If the service is rejected or the acceptance action violates the bandwidth quota constraints, the service will not enter the live network. If the service is legally accepted, the service will be bound to the selected candidate logical path.

[0029] The ingress shaping and flow-level caching module is used to maintain a flow-level cache queue for each accepted service and limit the actual output rate of the service entering the live network black box according to the configured ingress shaping rate.

[0030] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the ingress-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls, as described in the first aspect, is implemented.

[0031] 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 entry-side cross-network joint orchestration method for MTP latency risk control for point cloud video calls as described in the first aspect.

[0032] 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 entry-side cross-network joint orchestration method for MTP latency risk control for point cloud video calls as described in the first aspect.

[0033] Terminology Explanation:

[0034] (1) MTP (Motion-to-Photon): Motion-to-photon latency refers to the total latency between the occurrence of a user's action and the presentation of the corresponding image on the display. In point cloud video calls, MTP latency directly affects the user's perception of interaction response speed, immersion, and naturalness.

[0035] (2) PPO (Proximal Policy Optimization): Proximal policy optimization algorithm is a deep reinforcement learning policy gradient algorithm. In this invention, PPO can be used to train the joint orchestration strategy on the ingress side, so that the policy outputs a joint action of acceptance, path and shaping rate when the business request arrives.

[0036] (3) HTC (Holographic-type Communication): Holographic communication refers to a form of communication that enables remote users to obtain a high sense of presence and immersion through multimodal data transmission and real-time interaction, such as three-dimensional vision, spatial audio, and haptic feedback. Point cloud video calls are one of the more feasible holographic communication service forms at present.

[0037] (4) PCV (Point Cloud Video): Point cloud video refers to video media that uses time-continuous three-dimensional point cloud frames to represent the dynamic changes of people, objects or scenes. Each point usually contains attributes such as three-dimensional coordinates and color, and the amount of data and instantaneous bit rate are significantly higher than those of traditional two-dimensional video.

[0038] (5) QoS (Quality of Service): Service quality is usually used to describe objective transmission indicators such as network bandwidth, latency, jitter, and packet loss.

[0039] (6) QoE (Quality of Experience): The quality of experience refers to the user's subjective perception of the service experience. In point cloud video calls, QoE is related to factors such as media quality, interaction latency, display continuity, and spatial immersion.

[0040] (7) SDN (Software Defined Networking): Software-defined networking improves the flexibility of network path control, policy distribution and resource scheduling by separating the control plane and the forwarding plane.

[0041] (8) NFV (Network Function Virtualization): Network function virtualization deploys traditional dedicated network functions in software on a general computing platform to improve the flexibility of service function deployment and migration.

[0042] (9) SR-TE (Segment Routing Traffic Engineering): Segment routing-based traffic engineering technology that uses segment identifiers to express explicit paths or forwarding intentions to achieve policy-based path control and load scheduling.

[0043] (10) V-PCC (Video-based Point Cloud Compression) and G-PCC (Geometry-based Point Cloud Compression): MPEG point cloud compression related standards. The former mainly projects or organizes point clouds into video form and then compresses them using video encoding tools, while the latter mainly compresses point cloud geometry and attributes.

[0044] The beneficial effects of this invention are as follows: It introduces a candidate logical path abstraction for the existing black-box bearer environment, enabling the ingress side to complete the bearer selection for point cloud video calls without relying on the existing network's link-by-link opening, thus reducing the cost of cross-network deployment and control modification. Furthermore, it proposes a three-element joint action of admission control, candidate path binding, and ingress shaping rate, incorporating service admission, path selection, and burst release into a single ingress-side decision, improving the problem of fragmented routing, admission, and shaping in traditional solutions. In addition, it designs a closed-loop feedback model for MTP latency, MTP margin, and path risk latency, shifting the optimization objective from average network metrics to the latency default risk that real-time interactive services are more sensitive to, thereby improving the latency guarantee capability of point cloud video calls. After introducing strategy optimization methods such as PPO, the system can learn the dynamic trade-offs between service value, path margin, active service risk, and shaping intensity, thereby improving the effective service quality of admitted services and the real-time holographic interactive experience under high-load scenarios.

[0045] 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

[0046] 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.

[0047] Figure 1 This is a functional principle framework diagram of the orchestration system described in an embodiment of the present invention.

[0048] Figure 2 This is a flowchart of the joint orchestration mechanism described in an embodiment of the present invention.

[0049] Figure 3 This is a schematic diagram of the event-driven joint orchestration process described in an embodiment of the present invention. Detailed Implementation

[0050] 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.

[0051] 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.

[0052] 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.

[0053] 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.

[0054] 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.

[0055] 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.

[0056] 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.

[0057] Holographic communication services, such as holographic conferencing, holographic video calls, and immersive remote collaboration, require the network to be able to sense changes in service traffic and proactively adapt to users' differentiated experience needs. Point cloud video calls, as an important form of holographic communication, are characterized by large data volumes, strong instantaneous bursts, and sensitivity to MTP latency. When such services cross existing core network bearers, the nodes and links within the existing network are often difficult to control hop-by-hop finely by external orchestration systems. Therefore, more effective service access and bearer control is needed at the ingress side.

[0058] This invention addresses the problems of bursty service input, limited candidate path resources, ingress shaping queuing, and MTP latency default risk coupled in point cloud video calls across existing networks. It proposes an ingress-side cross-network joint orchestration method and system. This method utilizes available service request characteristics, candidate path-level available bandwidth, path risk latency, active service MTP margin, and service quality feedback on the ingress side to solve the problem of simultaneously determining whether a service is accepted, which candidate path to select, and what ingress shaping rate to configure under existing network black-box conditions.

[0059] The main design features of this invention include: abstracting the existing network as a black-box bearer environment; treating candidate logical paths between the ingress and egress points as selectable bearer objects; constructing service admission, candidate path binding, and ingress shaping rate configuration as a single event-driven joint action; establishing a closed-loop relationship between point cloud burst input, ingress shaping queue, path-level bandwidth reservation, network risk latency, and MTP latency; and employing PPO and other strategy optimization methods to learn the joint orchestration strategy on the ingress side. In summary, this invention solves the MTP latency risk control problem in point cloud video calls across existing networks, reduces reliance on hop-by-hop control within the existing network, and improves the effective service capability of real-time holographic interactive services.

[0060] This invention achieves controllable ingress-side bearer operation under existing black-box network conditions: Existing cross-network bearer technologies typically rely on strong internal network state openness or service-level resource orchestration, making it difficult to establish executable ingress control when fine-grained hop-by-hop control is not possible within the existing network. This invention abstracts the existing network into a black-box bearer environment, treating candidate logical paths between ingress and egress node pairs as selectable bearer objects. This invention completes admission, path, and shaping configuration through path-level summary feedback, aiming to reduce reliance on link-by-link state openness, hop-by-hop forwarding control, and deep cross-domain modifications.

[0061] This invention achieves a three-element joint orchestration for point cloud video call services: Existing point cloud video transmission technologies mainly focus on compression, bitrate, and media quality adaptation, making it difficult to determine whether a service is accepted before entering the network, which path it takes, and at what rate it enters the network. This invention constructs service acceptance control, candidate logical path binding, and ingress shaping rate tiers into a single joint action. The purpose of this invention is to integrate media bursts, network paths, and ingress queuing into a single event-driven decision, avoiding the fragmented decision-making caused by separate routing, separate admission, or separate shaping.

[0062] This invention achieves closed-loop control for MTP latency risk: Existing network optimization methods mostly target average latency, link utilization, throughput, or access benefits, making it difficult to directly constrain the sensitivity of point cloud video calls to MTP thresholds and tail latency. This invention establishes a closed-loop relationship between point cloud burst input, ingress stream-level buffer, shaping output rate, path risk latency, fixed processing latency, and MTP margin. This invention corrects subsequent ingress decisions through runtime MTP latency, default conditions, and service quality feedback, aiming to transform latency risk from post-event statistical indicators into a policy-aware, punitive, and adjustable control object.

[0063] This invention enhances the effective service capacity under high load: existing rules and strategies such as fixed paths, minimum load paths, or maximum feasible shaping are prone to over-acceptance, local congestion, or entry point queuing and transfer when business surges and path resources are scarce. This invention uses PPO and other strategy optimization methods to learn the trade-off relationship between business value, path margin, active business MTP margin, and shaping intensity. The purpose of this invention is not simply to increase the number of accepted services, but to protect the effective completion quality of accepted services when business load increases, reduce the probability of MTP latency default, and improve the real-time holographic interactive experience.

[0064] Example 1

[0065] In this embodiment 1, an ingress-side cross-network joint orchestration system for MTP latency risk control in point cloud video calls is first provided. The system includes a service request and feature extraction module, an orchestration module, an ingress access point, a network black box, an egress access point, a service receiving end, and a quality of service feedback link. The orchestration module is located in the control plane and logically focuses on the ingress orchestration process before the point cloud video call service enters the network. The ingress and egress access points can be regarded as boundary devices between the network black box and the service side. The network black box can still be composed of existing bearer networks such as IPv4 / IPv6, but this embodiment does not require the orchestration module to directly control each internal node or link. Instead, it generates orchestration actions based on the candidate path-level resource status and quality of service feedback available on the ingress side. Figure 1 As shown, the ingress-side cross-network joint orchestration system for MTP latency risk control in point cloud video calls includes the following functional modules:

[0066] (1) Service Request Feature Extraction Module: This module is used to extract features such as average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node, and exit node of the current service when the point cloud video call service arrives at the entry access point. This module provides service-side information for subsequent state construction and action legality judgment.

[0067] (2) State awareness and feature construction module: used to integrate information such as current service request features, candidate path-level resource feedback, path propagation delay, path-level network risk delay, number of active services, active service MTP margin, average shaping delay and service arrival intensity to construct a state vector that can be input for the joint orchestration strategy.

[0068] (3) Joint orchestration module: used to output joint actions based on the state vector. The joint actions include at least the admission control variable, candidate logical path identifier, and entry shaping rate level. This module can use a trained PPO policy network, or it can be replaced with other policy models with the same input-output relationship during implementation.

[0069] (4) Action execution and legality check module: used to determine whether the acceptance action meets the remaining acceptable bandwidth quota constraint at the candidate path level. If the service is rejected or the acceptance action violates the bandwidth quota constraint, the service will not enter the live network; if the acceptance is legal, the service will be bound to the selected candidate logical path and the shaping rate configuration will be sent to the ingress shaping module.

[0070] (5) Ingress Shaping and Stream-Level Caching Module: This module is used to maintain a stream-level caching queue for each accepted service and limit the actual output rate of the service entering the live network black box according to the configured ingress shaping rate. This module can smooth out point cloud traffic bursts, but it can also generate shaping queuing delays. Therefore, its rate configuration needs to be considered in conjunction with path status and MTP risk.

[0071] (6) Service Quality Feedback Module: It is used to collect information such as service MTP latency, MTP default status, path-level network risk latency, average shaping latency and completion validity during service operation, and feed it back to the status awareness module for use when the next service request arrives.

[0072] In this embodiment, the above-described system is used to implement an ingress-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls, including:

[0073] When the point cloud video call service arrives at the entry access point, the average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node and exit node characteristics of the current service are extracted to provide service-side information for subsequent state construction and action legality judgment.

[0074] By integrating current service request characteristics, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information, a state vector that can be input for the joint orchestration strategy is constructed.

[0075] Based on the state vector, a pre-trained joint orchestration strategy model is used to output a joint action; the joint action includes at least an admission control variable, a candidate logical path identifier, and an entry shaping rate level.

[0076] Determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level; if the service is rejected, or the acceptance action violates the bandwidth quota constraints, the service will not enter the live network; if the acceptance is valid, the service will be bound to the selected candidate logical path.

[0077] Maintain a flow-level cache queue for each accepted service, and limit the actual output rate of the service into the live network black box according to the configured ingress shaping rate.

[0078] The system flow in this embodiment can be understood from two levels: "closed-loop orchestration mechanism" and "event-driven decision-making process". The former describes the functional relationship between service requests, network status, state construction, joint orchestration, action execution, and quality of service feedback; the latter describes how joint orchestration actions affect discrete decision rounds and continuous physical time slices when service requests arrive randomly.

[0079] like Figure 2 As shown, after the joint orchestration module outputs a joint orchestration action, the action execution and legality check module breaks down the action into admission control, path selection, and shaping rate configuration, and determines whether the action meets the path-level resource feasibility requirements. After the service enters the live network, the network operation status and service quality feedback flow back to the status awareness module, thus forming a closed-loop control of "status feedback - joint orchestration - action execution - quality feedback".

[0080] like Figure 3 As shown, the joint orchestration in this embodiment is not re-decided for each physical time slice, but is triggered when the service request arrives. Let the first... The time of arrival of this service request is The system constructs a state at this moment. and output the action. The actions include acceptance, routing, and reshaping configuration. Between the arrival times of two adjacent requests, the system follows... Each physical time slice continuously updates the burst input rate, ingress queue and shaping output, path network risk latency, and active service MTP latency. The results of this process generate feedback and influence the state construction when the next service request arrives. Therefore, a single joint orchestration action not only determines whether the current service enters the live network, but also continuously affects subsequent physical time slices through path occupancy, shaping queues, and MTP risks.

[0081] In this embodiment, the Proximity Policy Optimization (PPO) algorithm can be used to train the joint orchestration strategy. Before training, an event-driven joint orchestration environment is constructed based on the existing network abstract model, candidate path set, point cloud service generation model, ingress shaping tier, MTP latency model, and reward function. This environment uses the arrival of service requests as the decision trigger event and advances the queue, path occupancy, and MTP latency status of accepted services using physical time slices. In one training round, the system first resets the network state, active service set, and remaining path bandwidth quota; then, it generates point cloud video call requests according to the service arrival model. When a request arrives, the environment follows the formula... To the style Construction state The policy network outputs an action based on this state. After the environment performs this action, it first follows the formula. Determine if the action is legal; legally accepted services are added to the active service set and configured with paths and shaping rates, while rejected or illegally accepted services are not allowed to enter the live network.

[0082] Following the current decision and before the next business request arrives, the environment continuously updates the business burst input, ingress queue, shaped output, actual path occupancy, MTP latency, and MTP margin according to physical time slices, and based on the formula... The formula is referenced here to calculate the reward for the current decision. This forms the training samples. This provides PPO with updated strategy networks and value networks.

[0083] During PPO updates, the policy network outputs the probability distribution of joint actions, while the value network estimates state values ​​and calculates the advantage function. PPO limits the change in the probability ratio between old and new policies by pruning objectives, preventing excessive oscillations in policy updates due to single-round samples. Simultaneously, the value function loss improves the accuracy of return estimation, and the entropy regularization term maintains a certain level of exploratory capability. Since the action space of this invention includes combinations of acceptance, pathing, and gear shaping, PPO can learn joint decision-making patterns under different business loads and path states within a finite discrete action space.

[0084] Once training is complete, the online execution phase no longer requires retraining the model for each incoming request. The ingress orchestration module only needs to read the current business request features and path-level feedback, construct a state vector, and input it into the trained policy network to directly output the acceptance result, candidate path number, and shaping rate level. The action execution module then performs bandwidth validity checks and distributes path binding and shaping configurations. In this way, online ingress orchestration can be completed quickly when a business request arrives.

[0085] In this embodiment, for candidate paths and business models: for any ingress-exit node pair The set of candidate logical bearer paths exposed by the current network black box to the entry side can be represented as:

[0086] ;

[0087] In the formula, Indicates entry point To the exit The set of candidate paths, Indicates the first Candidate logical paths, This indicates the number of candidate paths. This formula emphasizes that the present invention selects logical bearer paths that are currently exposed or pre-configured on the external network, rather than controlling physical links within the existing network hop-by-hop.

[0088] A point cloud video call service request can be represented as: ;

[0089] In the formula, The long-term average bit rate For the duration of the business, The acceptable MTP latency threshold for the business. For business value, and Describe the intensity and probability of a sudden event, respectively. These represent the service entry and exit points, respectively. This formula is used to uniformly incorporate the resource requirements, latency constraints, and service levels of point cloud video calls into the state structure.

[0090] Whether a new service request is generated within each physical time slice can be represented using the Bernoulli arrival model:

[0091] ;

[0092] in, Indicates time slice A new business request has arrived. This indicates that no new business requests have arrived. This indicates the intensity of service arrival. Using this formula, the present invention can trigger joint orchestration decisions at different frequencies on the ingress side under light, medium, and high load conditions.

[0093] For point cloud burst input models, business In time slice In the event of a sudden event, the input rate is:

[0094] ;

[0095] In the formula, Indicates the instantaneous input rate. This represents the long-term average bitrate. This indicates the burst rate. This formula is used to describe instantaneous high bitrate input caused by changes in point cloud scene complexity, user perspective switching, or changes in encoded frame size.

[0096] business In time slice When no sudden event occurs, its input rate is: ;

[0097] In the formula, This represents the rate coefficient for non-burst states. In non-burst states, the rate is lower than or close to the average bit rate, used to maintain the long-term average rate without deviation from the burst state. .

[0098] To ensure that the long-term expected rate of service input equals the average bit rate It should satisfy:

[0099] ;

[0100] In this equation, the first term on the left represents the expected contribution under burst conditions, and the second term represents the expected contribution under non-burst conditions. This formula ensures that the invention, while modeling burstiness, does not change the long-term average bitrate of the service itself.

[0101] Therefore, the non-sudden state rate coefficient can be obtained:

[0102] ;

[0103] in, The larger or The larger the value, the higher the non-sudden state rate coefficient. The lower the value, the better it offsets the high input rate of sudden events in the long-term average sense.

[0104] Inlet shaping device in time slice The output data volume is:

[0105] ;

[0106] in, To address the queue backlog left over from the previous time slice, Add input data to this time slice. This represents the maximum amount of data that the shaper is allowed to output within this time slice. This formula indicates that the ingress shaper can only release data at the configured rate and cannot push all burst inputs into the live network at once.

[0107] The actual output rate after shaping is:

[0108] ;

[0109] In the formula, This is the actual rate at which services enter the live network black box. This rate directly affects the actual path occupancy and network risk latency, making it a key variable in connection ingress reshaping actions and live network carrying risks.

[0110] The ingress stream-level cache queue has been updated to:

[0111] ;

[0112] If the amount of input data and historical backlog exceed the shaping service capacity of the current time slice, the queue will continue to grow; if the shaping service capacity is sufficient, the queue will be gradually released. This formula reflects that an excessively low shaping rate will increase ingress queuing latency, while an excessively high shaping rate may increase core network path risks.

[0113] The entry shaping delay can be expressed as:

[0114] ;

[0115] In the formula, The unit is Mbit. The unit is Mbps. Dividing the two gives seconds, which are then multiplied by 1000 to convert to milliseconds. To prevent extremely small constants with a denominator of zero, this formula is used to estimate the contribution of inbound queue backlog to MTP delay.

[0116] For the bandwidth reservation and action legality model, candidate paths In time slice The remaining available bandwidth quota can be expressed as:

[0117] ;

[0118] in, Representing a path Hidden internal logic links, Indicates link capacity. This indicates that bandwidth has been reserved on this link. This formula can generate path-level summary results from a simulation environment or the existing network management plane; the ingress side uses... It is not necessary to know the specific location of the bottleneck link.

[0119] The reserved bandwidth on the hidden link can be accumulated according to the shaping rate of the accepted services:

[0120] ;

[0121] In the formula, Time slice A collection of active businesses Indicates business Bound candidate paths, Indicates business The ingress shaping rate. This formula indicates that once a service is accepted, its shaping rate is used as the path-level bandwidth quota for subsequent legitimacy determination.

[0122] For business Joint choreography The legality constraints of the action are:

[0123] ;

[0124] in, This indicates an active rejection of the service, in which case no path bandwidth is consumed; when At that time, only the selected shaping rate No more than the remaining acceptable bandwidth of the path Only after this condition is met will the service be allowed to enter the live network. This formula aligns service acceptance with hard constraints on path-level resources.

[0125] For the MTP latency and path risk model, business In time slice The MTP latency is defined as follows:

[0126] ;

[0127] In the formula For the propagation delay of candidate paths, For path-level network risk latency, For entrance shaping delay, This formula incorporates fixed delays such as end-side processing and protocol processing into the MTP delay estimation, taking into account existing network path factors, ingress shaping factors, and fixed processing factors.

[0128] The actual bandwidth occupied by the internal links of the path can be expressed as:

[0129] ;

[0130] in, It is a hidden link In time slice The actual output usage, For business The actual output rate after ingress shaping. The formula for the actual bandwidth occupied by internal links differs from the formula for the reserved bandwidth on hidden links, which is calculated by adding the shaped rates of the accepted services. The formula for the reserved bandwidth on hidden links focuses on the reserved quota, while the formula for the actual bandwidth occupied by internal links focuses on the actual output during operation.

[0131] The actual occupancy rate of the hidden link is:

[0132] ;

[0133] In the formula, This represents the link capacity. The closer the actual occupancy rate is to 1, the higher the risk of link queuing and congestion. Therefore, this variable is used to construct path-level network risk latency.

[0134] Candidate Path Network risk latency can be expressed as:

[0135] ;

[0136] in, This represents the risk delay coefficient. This formula reflects the non-linear increase in network risk delay as the link occupancy rate within the path increases; in practical implementation, it can be adjusted... Set an upper limit to prevent the denominator from being too small, which could lead to abnormal values.

[0137] business The MTP margin ratio is:

[0138] ;

[0139] in, This indicates that the current MTP latency is still below the threshold. This means that the threshold has just been reached. This indicates that an MTP default has occurred. This formula is used to uniformly map businesses with different MTP thresholds to a dimensionless risk indicator.

[0140] For the state, action, and reward model, the first... When the next service request arrives, the joint orchestration state can be represented as: ;

[0141] In the formula, Indicates the characteristics of the current business request. Indicates the first The path-level state of each candidate path at the decision moment. This indicates the status of the entry point and service quality summary. The formula illustrates that the decision input consists of the business side, the path side, and the feedback side.

[0142] The characteristics of the current business request can be represented as follows:

[0143] ;

[0144] in, These represent the average bit rate, duration, MTP threshold, service value, burst probability, and burst factor of the current service, respectively. This vector is used to characterize the bandwidth requirements, latency sensitivity, and burst risk of the current service.

[0145] No. The status of a candidate path can be represented as:

[0146] ;

[0147] in, Representation and path The number of related active businesses This represents the minimum MTP margin among the services associated with this path. This formula enables the strategy to simultaneously perceive the remaining resources of the path, propagation latency, risk latency, and the most dangerous state of the services already carried.

[0148] Joint choreography can be represented as: In the formula, Indicate whether to accept the current business. Indicates the selected candidate logical path, This represents the inlet shaping rate. This formula is the core action expression that distinguishes this invention from individual admission, individual routing, or individual shaping strategies.

[0149] The ingress shaping rate can be determined by the average bit rate and the discrete bit rate: ;in, The preset shaping ratio set keeps the action space finite while covering downgrade shaping, near-average bitrate shaping, and burst absorption shaping above average bitrate.

[0150] The penalty for unlawful acceptance can be expressed as follows:

[0151] ;

[0152] in, This is the bandwidth violation penalty coefficient. If the acceptance action meets the bandwidth constraint, the numerator is 0; if the shaping rate exceeds the remaining acceptable bandwidth of the path, the larger the excess ratio, the greater the penalty.

[0153] The operating period reward can be represented as:

[0154] ;

[0155] In the formula, the first term is the revenue from continuous active business services, the second term is the penalty for exceeding the MTP threshold, and the third term is the penalty for insufficient MTP security margin. and The severity of default penalties and safety margin penalties are controlled separately. This formula ensures that the strategy not only focuses on whether a business is accepted, but also on whether the accepted business can remain stably in a low-risk MTP state.

[0156] No. The total return for each joint orchestration decision can be expressed as:

[0157] ;

[0158] in, This indicates the number of physical time slices between two consecutive service arrivals. This represents the runtime reward discount factor. This formula aggregates the runtime effects of a single entry orchestration action over several subsequent time slices into the current decision reward.

[0159] The optimization objective of the joint orchestration strategy is: The number of decisions made within a single round. This formula illustrates that the goal of strategy training is to maximize cumulative returns during long-term business arrival, while balancing acceptance scale, path load, entry shaping, and MTP risk control.

[0160] In summary, this embodiment implements an ingress-side cross-network joint orchestration architecture under existing network black-box conditions: the existing network is abstracted as a black-box bearer environment, candidate logical bearer paths between ingress and egress node pairs are used as selectable bearer objects, and online orchestration is performed based on candidate path-level summary feedback. A ternary joint action design is proposed: when a service request arrival event is triggered, the service admission control variable, candidate logical path identifier, and ingress shaping rate level are combined into the same joint orchestration action. A state construction method for point cloud video calls is implemented: the state includes at least the current service's average bitrate, duration, MTP threshold, service value, burst probability, burst factor, and path-level or ingress-side feedback such as candidate path remaining bandwidth quota, propagation delay, path-level network risk delay, number of active services, and MTP margin. A joint modeling method for ingress shaping and MTP delay is proposed: the service MTP delay is calculated based on point cloud service burst input, stream-level buffer queue, shaping output rate, path propagation delay, path-level network risk delay, and fixed processing delay, and this MTP delay is used in policy feedback. An action legality check mechanism was implemented: when the shaping rate corresponding to an acceptance action exceeds the remaining acceptable bandwidth quota of the selected candidate path, the action is judged as illegal, preventing the service from entering the live network, and corresponding penalties are formed during training or evaluation. A reward function construction method for MTP risk control was implemented: the service continuity benefit, MTP over-threshold penalty, low security margin penalty, and illegal action penalty are all incorporated into the decision reward, guiding the strategy to achieve a balance between service access capacity and effective service quality. A reinforcement learning-based strategy training and online execution method was proposed: state, action, reward, and next state samples are collected through simulation or test environments to train a joint orchestration strategy; after deployment, when a new service request arrives, the acceptance, path, and shaping configuration are directly output.

[0161] Example 2

[0162] 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 entry-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls. This method includes:

[0163] When the point cloud video call service arrives at the entry access point, the average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node and exit node characteristics of the current service are extracted to provide service-side information for subsequent state construction and action legality judgment.

[0164] By integrating current service request characteristics, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information, a state vector that can be input for the joint orchestration strategy is constructed.

[0165] Based on the state vector, a pre-trained joint orchestration strategy model is used to output a joint action; the joint action includes at least an admission control variable, a candidate logical path identifier, and an entry shaping rate level.

[0166] Determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level; if the service is rejected, or the acceptance action violates the bandwidth quota constraints, the service will not enter the live network; if the acceptance is valid, the service will be bound to the selected candidate logical path.

[0167] Maintain a flow-level cache queue for each accepted service, and limit the actual output rate of the service into the live network black box according to the configured ingress shaping rate.

[0168] Example 3

[0169] 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 above-described ingress-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls. The method includes:

[0170] When the point cloud video call service arrives at the entry access point, the average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node and exit node characteristics of the current service are extracted to provide service-side information for subsequent state construction and action legality judgment.

[0171] By integrating current service request characteristics, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information, a state vector that can be input for the joint orchestration strategy is constructed.

[0172] Based on the state vector, a pre-trained joint orchestration strategy model is used to output a joint action; the joint action includes at least an admission control variable, a candidate logical path identifier, and an entry shaping rate level.

[0173] Determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level; if the service is rejected, or the acceptance action violates the bandwidth quota constraints, the service will not enter the live network; if the acceptance is valid, the service will be bound to the selected candidate logical path.

[0174] Maintain a flow-level cache queue for each accepted service, and limit the actual output rate of the service into the live network black box according to the configured ingress shaping rate.

[0175] Example 4

[0176] 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 to cause the electronic device to execute instructions for implementing the above-described ingress-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls. The method includes:

[0177] When the point cloud video call service arrives at the entry access point, the average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node and exit node characteristics of the current service are extracted to provide service-side information for subsequent state construction and action legality judgment.

[0178] By integrating current service request characteristics, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information, a state vector that can be input for the joint orchestration strategy is constructed.

[0179] Based on the state vector, a pre-trained joint orchestration strategy model is used to output a joint action; the joint action includes at least an admission control variable, a candidate logical path identifier, and an entry shaping rate level.

[0180] Determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level; if the service is rejected, or the acceptance action violates the bandwidth quota constraints, the service will not enter the live network; if the acceptance is valid, the service will be bound to the selected candidate logical path.

[0181] Maintain a flow-level cache queue for each accepted service, and limit the actual output rate of the service into the live network black box according to the configured ingress shaping rate.

[0182] 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.

[0183] 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.

[0184] 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.

[0185] 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.

[0186] 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. An entry side cross-network joint orchestration method for point cloud video call MTP latency risk control, characterized in that, include: When the point cloud video call service arrives at the entry access point, the average bit rate, duration, MTP latency threshold, service value, burst probability, burst factor, entry node and exit node characteristics of the current service are extracted to provide service-side information for subsequent state construction and action legality judgment. By integrating current service request characteristics, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information, a state vector that can be input for the joint orchestration strategy is constructed. Based on the state vector, a pre-trained joint orchestration strategy model is used to output a joint action; the joint action includes at least an admission control variable, a candidate logical path identifier, and an entry shaping rate level. Determine whether the admission action satisfies the remaining acceptable bandwidth quota constraint at the candidate path level; If a service is rejected, or if the acceptance action violates bandwidth quota constraints, the service will not be allowed into the live network. If the acceptance is valid, the business logic will be bound to the selected candidate logical path. Maintain a flow-level cache queue for each accepted service, and limit the actual output rate of the service into the live network black box according to the configured ingress shaping rate.

2. The entrance side cross-network joint orchestration method for point cloud video call MTP latency risk control according to claim 1, characterized in that, Training the joint orchestration strategy based on the PPO update strategy includes: Before training, an event-driven joint orchestration environment is constructed based on the existing network abstract model, candidate path set, point cloud service generation model, ingress shaping rate, MTP latency model, and reward function. This environment uses the arrival of service requests as the decision trigger event and advances the queue, path occupancy, and MTP latency status of accepted services using physical time slices. In a training round, the network state, active service set, and remaining path bandwidth quota are first reset. Then, point cloud video call requests are generated according to the service arrival model. When a request arrives, a state is constructed, and an action is output based on the state. After executing the action, it is determined whether the action is legal. Legally accepted services are added to the active service set and configured with paths and shaping rates. Rejected or illegally accepted services do not enter the existing network.

3. The entrance side cross-network joint orchestration method for point cloud video call MTP latency risk control according to claim 2, characterized in that, After the current decision and before the next business request arrives, the environment continuously updates the business burst input, entry queue, shaped output, actual path occupancy, MTP latency and MTP margin according to physical time slices, and calculates the reward of the current decision, thereby forming training samples for PPO to update the policy network and value network.

4. The ingress-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls according to claim 3, characterized in that, During PPO updates, the policy network is used to output the probability distribution of joint actions, and the value network is used to estimate state value and calculate the advantage function. PPO limits the change in the probability ratio between the old and new policies by pruning the target, so that the policy update will not cause excessive oscillations due to a single round of samples. At the same time, the value function loss improves the accuracy of reward estimation, and the entropy regularization term maintains a certain exploratory capability.

5. The ingress side cross-network joint orchestration method for point cloud video call MTP latency risk control according to claim 3, characterized in that, The optimization objective of the joint orchestration strategy is to maximize cumulative returns during long-term business arrival, thereby balancing intake scale, path load, ingress shaping, and MTP risk control. ； wherein denotes a joint orchestration policy, denotes a long-term discount factor across decision rounds, denotes the number of decisions that occur within one round.

6. The ingress side cross-network joint orchestration method for point cloud video call MTP latency risk control according to claim 1, characterized in that, Joint orchestration does not involve re-deciding at every physical time slice, but rather is triggered when a business request arrives. Let's say the first... The time of arrival of this service request is The system constructs a state at this moment. and output the action. The actions include acceptance, pathing, and reshaping configuration; between the arrival times of two adjacent requests, the system follows... Each physical time slice continuously updates burst input rate, ingress queue and shaped output, path network risk latency, and active service MTP latency.

7. An ingress side cross-network joint orchestration system for point cloud video call MTP latency risk control, characterized in that, include: The business request feature extraction module is used to extract the average bit rate, duration, MTP latency threshold, business value, burst probability, burst factor, entry node and exit node features of the current business when the point cloud video call business arrives at the entry access point, so as to provide business-side information for subsequent state construction and action legality judgment. The state awareness and feature construction module is used to integrate current service request features, candidate path-level resource feedback, path propagation latency, path-level network risk latency, number of active services, active service MTP margin, average shaping latency, and service arrival strength information to construct a state vector that can be input to the joint orchestration strategy. The joint orchestration module is used to output joint actions based on the state vector using a pre-trained joint orchestration strategy model; the joint actions include at least the admission control variable, candidate logical path identifier, and entry shaping rate level. The action execution and legality check module is used to determine whether the acceptance action meets the remaining acceptable bandwidth quota constraints at the candidate path level. If a service is rejected, or if the acceptance action violates bandwidth quota constraints, the service will not be allowed into the live network. If the acceptance is valid, the business logic will be bound to the selected candidate logical path. The ingress shaping and flow-level caching module is used to maintain a flow-level cache queue for each accepted service and limit the actual output rate of the service entering the live network black box according to the configured ingress shaping rate.

8. A non-transitory computer-readable storage medium, comprising: The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the entry-side cross-network joint orchestration method for MTP latency risk control in point cloud video calls as described in any one of claims 1-6.

9. A computer device, comprising: The system includes a memory and a processor, which 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 ingress-side cross-network joint orchestration method for MTP latency risk control for point cloud video calls as described in any one of claims 1-6.

10. An electronic device, comprising: include: The device includes 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 to cause the electronic device to execute instructions that implement the ingress-side cross-network joint orchestration method for MTP latency risk control for point cloud video calls as described in any one of claims 1-6.

Images