Video rate adaptation method based on double-layer edge network and quality of experience driving

Abstract

This application relates to a video bitrate adaptation method based on a two-layer edge network and experience quality driven. The method includes: constructing a... M A two-layer collaborative video caching architecture consisting of micro base stations and several key users includes a set of cache nodes. Within a preset long-term timescale, each cache node in the set caches video files according to a caching strategy. Within a preset short-term timescale, an instantaneous experience quality model is constructed based on the current bitrate, bitrate fluctuations, initial waiting time, and playback integrity. A maximization problem for the instantaneous experience quality model is then constructed based on this model. The maximization problem is solved using a Lagrange duality method based on KKT conditions and subgradient update iterations to obtain the continuously optimal bitrate. This application enables adaptive optimization of video bitrate by comprehensively utilizing caching, computation, and transmission resources within a two-layer collaborative video architecture.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a video bitrate adaptive method based on a two-layer edge network and experience quality driven. Background Technology

[0002] With the explosive growth of mobile internet and the continued surge in user demand for high-definition video content, network infrastructure is facing unprecedented pressure. By the end of 2025, video traffic accounted for 76% of all mobile data traffic, with global monthly mobile network data traffic reaching approximately 188 EB. While the widespread deployment of 5G technology offers higher transmission speeds and lower latency, the explosive growth of video services still places enormous pressure on core networks and backhaul links.

[0003] Since most video content is stored on remote cloud servers, transmitting video via the backhaul link leads to higher transmission latency, reduced video bitrate, and even playback stuttering, severely impacting the user's Quality of Experience (QoE). Edge caching technology, by pre-storing popular content on edge nodes closer to the user, can effectively reduce content retrieval latency and alleviate backhaul link congestion, thereby improving the user's QoE.

[0004] However, edge nodes typically have limited caching and computing resources, and efficiently utilizing these resources is a key challenge. Among related technologies, collaborative caching schemes improve caching efficiency by coordinating the storage of important video segments across multiple base stations, while device-to-device (D2D) communication technology can offload caching and computing tasks to user devices. Meanwhile, Dynamic Adaptive Streaming over HTTP (DASH) can dynamically adjust video quality based on network conditions.

[0005] However, providing multiple bitrate versions of video content at the wireless edge increases storage requirements and potential traffic congestion. To address this, edge video transcoding technology has been introduced. Its principle is to pre-place high-quality versions of video at edge nodes and transcode them to lower bitrate versions as needed. Although existing research has made improvements in areas such as caching strategies, transcoding mechanisms, and bitrate adaptation, current solutions often lack joint optimization of caching, computation, and transmission resources, and QoE models are mostly based on statistics after the entire video is played, making it difficult to support real-time bitrate decisions.

[0006] Therefore, there is an urgent need for a method that can comprehensively utilize caching, computing, and transmission resources under a dual-layer edge caching architecture to achieve real-time QoE-driven adaptive optimization of video bitrate.

[0007] Therefore, it is necessary to provide a new technical solution to improve one or more of the problems existing in the above solutions.

[0008] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0009] The purpose of this application is to provide a video bitrate adaptive method based on a two-layer edge network and experience quality driven, thereby overcoming, at least to some extent, one or more problems caused by the limitations and defects of related technologies.

[0010] A video bitrate adaptive method based on a two-layer edge network and experience quality-driven model is provided according to an embodiment of this application. The method includes: Build including M A two-layer collaborative video caching architecture consisting of micro base stations and several important users, the two-layer collaborative video caching architecture including a set of caching nodes; wherein, each micro base station covers a community, and each important user is a user device within the community with local caching and computing capabilities; Within a preset long time period, each cache node in the cache node set caches video files according to a caching strategy; Each video file is divided into multiple segments according to a preset duration, and each segment corresponds to a bitrate. Within a preset short time scale period, an instantaneous experience quality model is constructed based on the current bitrate, bitrate fluctuation, initial waiting time, and playback integrity. The problem of maximizing the instantaneous experience quality model is constructed based on the instantaneous experience quality model. The maximization problem is solved by using the Lagrange dual method based on KKT conditions and subgradient update iteration to obtain the continuous optimal bitrate; Discretize the continuous optimal bitrate and adjust the user bitrate to obtain the final bitrate decision that satisfies the constraints.

[0011] In the embodiments of this application, the step of caching video files according to a caching strategy by each cache node in the cache node set within a preset long time scale period includes: The popularity of content in the community is modeled using a Zipf distribution with a preset skewness parameter, and the probability of the user device requesting a video file is calculated. Each cache node estimates its local preference for video files by aggregating the request probability of its connected users requesting video files, so that each cache node caches its respective video file according to the local preference.

[0012] In the embodiments of this application, the expression of the instantaneous experience quality model is as follows: (1) In the formula, This represents the instantaneous experience quality model. Indicates the current bitrate. Indicates bitrate fluctuation. Indicates the initial waiting time. Indicates playback completeness. The weight representing the current bitrate. The weights representing bitrate fluctuations The weight representing the initial waiting time. Weights representing playback integrity .

[0013] In the embodiments of this application, the expression for playback integrity is as follows: (2) In the formula, t represents a time slot, and f represents a video file. Indicates user equipment The video file in the buffer of time slot t The amount of data, Indicates the remaining playback time. This indicates the return download speed.

[0014] In embodiments of this application, the step of constructing the maximization problem of the instantaneous experience quality model based on the instantaneous experience quality model includes: The expression for the maximization problem is as follows: (3) In the formula, This means finding a solution that maximizes the bitrate decision variable. Indicates user In the community Does this request a video file? The binary request indicator variable, Indicates time slot, Indicates user, Indicates a video file. Indicates the total number of time slots. Indicates community Total number of users within, Indicates the total number of video files. This indicates the computing power of the micro base station. This means that for any time slot, This indicates a user-video file pair. Indicates in time slot Micro base stations The service provides a collection of user-video file pairs. Indicates being bound by, Indicates the computing power of important users. Indicates an important user. This means for any , Represents a set of important users. Indicates in time slot By important users The service provides a collection of user-video file pairs. Indicates the downlink transmission rate of the UE. Indicates community The user set in Represents a collection of video files. Represents the set of bitrates. Indicates the first constraint. This indicates the second constraint.

[0015] In the embodiments of this application, the step of solving the maximization problem using the Lagrange duality method based on KKT conditions and subgradient update iteration to obtain the continuous optimal bitrate includes: The current bit rate is discretely relaxed into a continuous variable, and two dual variables are introduced using the Lagrange multiplier method, transforming the maximization problem into an unconstrained optimization problem. The unconstrained optimization problem is solved using the KKT conditions to obtain a closed-form solution for the optimal bit rate. The two dual variables are updated using the subgradient descent method until the change in the unconstrained optimization problem converges. The closed-form solution of the optimal bitrate is then updated using the updated two dual variables to obtain the continuous optimal bitrate.

[0016] In the embodiments of this application, the step of discretizing the continuous optimal bitrate and adjusting the user bitrate to obtain the final bitrate decision that satisfies the constraints includes: Discretize the continuous optimal code rates and map them to the nearest feasible code rate level; Compare the loss value corresponding to the most recent bitrate level above with the loss value of the most recent bitrate level below, and select the most recent bitrate level with the larger loss value; If the nearest bitrate level corresponding to the selected large loss value violates the first and second constraints, then all loss values ​​are sorted in ascending order, and the bitrate of the user with the smallest loss value is reduced to the next level, until the bitrate of the next level satisfies the constraints.

[0017] In the embodiments of this application, the expression of the unconstrained optimization problem is as follows: (4) In the formula, This represents an unconstrained optimization problem. , Represents the dual variable. Indicates time slot By cache node The service provides a collection of user-video file pairs. Represents edge nodes Computing capacity, Indicates community The number of cache nodes in the system.

[0018] In the embodiments of this application, the step of solving the unconstrained optimization problem using KKT conditions to obtain a closed-form solution for the optimal bit rate includes: The partial derivative of the unconstrained optimization problem with respect to the current bitrate is calculated using the KKT conditions and set to zero; the expression for the partial derivative of the unconstrained optimization problem with respect to the current bitrate is as follows: (5) In the formula, This represents finding the partial derivative of an unconstrained optimization problem with respect to the current bitrate. Indicates community The collection of cache nodes in; The expression for the partial derivative of playback integrity with respect to the current bitrate is as follows: (6) In the formula, Indicates user equipment The video file in the buffer of time slot t The amount of data, Indicates the remaining playback time. Indicates the return download speed. Indicates playback integrity; when At that time, the closed-form solution for the optimal bit rate is expressed as follows: (7) In the formula, This represents the closed-form solution for the optimal bit rate. Indicates the bitrate of the previous time slot; when Substituting formula (6) into formula (5), we obtain a cubic polynomial equation for the closed-form solution of the optimal code rate. Solving this cubic polynomial equation yields the closed-form solution for the optimal code rate. The expression for the cubic polynomial equation is as follows: (8).

[0019] In the embodiments of this application, when updating the two dual variables using the subgradient descent method, the two dual variables are updated according to the following update expression: (9) In the formula, Indicates step size, Indicates community The collection of cache nodes in the middle, , Represents the dual variable. Represents edge nodes The computational capacity is given by n, where n represents the edge nodes.

[0020] The technical solutions provided by the embodiments of this application may include the following beneficial effects: In one embodiment of this application, a two-layer collaborative video caching architecture consisting of micro base stations and key users is constructed using the above method. Within a preset long-term timescale, a caching strategy based on the most popular local content is implemented for both layers of caching nodes to cache video files at each node. Within a preset short-term timescale, a multi-index instantaneous experience quality model is constructed, comprehensively considering the current bitrate, bitrate fluctuations, initial waiting time, and playback integrity. The Lagrange duality method with KKT conditions and subgradient updates are used to efficiently solve the maximization problem of the instantaneous experience quality model, obtaining the real-time optimal bitrate decision. This application enables adaptive optimization of video bitrate by comprehensively utilizing caching, computing, and transmission resources within a two-layer collaborative video architecture.

[0021] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0023] Figure 1 This schematically illustrates a flowchart of the steps in an exemplary embodiment of the present application for a video bitrate adaptation method based on a two-layer edge network and experience quality driven. Figure 2 This schematic diagram illustrates a two-layer collaborative video caching architecture system model in an exemplary embodiment of this application. Figure 3 The illustration shows the overall QoE comparison curves of various schemes under different preset skewness parameters in the exemplary embodiments of this application; Figure 4 The illustration shows the overall QoE comparison curves of various schemes under different IU computing capabilities in the exemplary embodiments of this application. Detailed Implementation

[0024] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0025] Furthermore, the accompanying drawings are merely illustrative of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0026] This example implementation first provides a video bitrate adaptation method based on a two-layer edge network and experience quality driven. (Reference) Figure 1 As shown, the method may include steps S101 to S106.

[0027] Among them, step S101: constructing including M A two-layer collaborative video caching architecture consisting of micro base stations and several key users, comprising a set of cache nodes; where each micro base station covers a community, and each key user is a user device within the community with local caching and computing capabilities.

[0028] Step S102: Within a preset long time scale period, each cache node in the cache node set caches the video file according to the caching strategy.

[0029] Step S103: Divide each video file into multiple segments according to a preset duration. Each segment corresponds to a bitrate. Within a preset short time scale period, construct an instantaneous experience quality model based on the current bitrate, bitrate fluctuation, initial waiting time, and playback integrity.

[0030] Step S104: Construct a maximization problem of the instantaneous experience quality model based on the instantaneous experience quality model.

[0031] Step S105: The maximization problem is solved by using the Lagrange dual method based on KKT conditions and subgradient update iteration to obtain the continuous optimal bit rate.

[0032] Step S106: Discretize the continuous optimal bitrate and adjust the user bitrate to obtain the final bitrate decision that satisfies the constraints.

[0033] In one embodiment of this application, a two-layer collaborative video caching architecture consisting of micro base stations and key users is constructed using the above method. Within a preset long-term timescale, a caching strategy based on the most popular local content is implemented for both layers of caching nodes to cache video files at each node. Within a preset short-term timescale, a multi-index instantaneous experience quality model is constructed, comprehensively considering the current bitrate, bitrate fluctuations, initial waiting time, and playback integrity. The Lagrange duality method with KKT conditions and subgradient updates are used to efficiently solve the maximization problem of the instantaneous experience quality model, obtaining the real-time optimal bitrate decision. This application enables adaptive optimization of video bitrate by comprehensively utilizing caching, computing, and transmission resources within a two-layer collaborative video architecture.

[0034] Below, we will refer to Figure 1 The steps of the method described above in this example embodiment will be explained in more detail.

[0035] In step S101, constructing includes M A two-layer collaborative video caching architecture consisting of micro base stations and several key users, comprising a set of cache nodes; where each micro base station covers a community, and each key user is a user device within the community with local caching and computing capabilities.

[0036] Specifically, deploying within the internet zone Small-cell base stations (SBS) are used to aggregate... Index. During the observation period Within, users are divided according to the coverage area of ​​each SBS. Non-overlapping communities, using Indicates community The user set in. In the first In each community, the selection ratio was: User equipment (UE) with local caching and computing capabilities is designated as important users (IU). The set of important users (i.e., the set of IU) is denoted as... , Indicates community The number of important users in the system. SBS and IU together constitute a two-layer collaborative video caching architecture. The two-layer collaborative video caching architecture includes a set of cache nodes, which is composed of SBS and IU. ,in, Indicates community Micro base stations.

[0037] Each SBS is located at the center of its coverage area and has caching space. Computational ability and transmission power Each IU has a cache space. Computational ability and transmission power The regional server, acting as a coordination and control server, can centrally orchestrate and maintain the cache directory of all SBSs, enabling regional collaborative caching, and simultaneously distributing real-time video scheduling decisions. The regional server includes SBSs, UEs, and IUs.

[0038] Each video file is divided into segments with a preset duration. Multiple segments. The collection of video files is denoted as... , This indicates the total number of video files, of which, This indicates the number of segments in each video file. Each segment corresponds to a bitrate; that is, the resolution of each segment belongs to the bitrate set. ,in, Indicates the first Bitrate, Indicates the maximum bitrate. The bitrate is... The size of the fragment is denoted as .

[0039] The user's initial position follows a Poisson Point Process (PPP), changing over time. Caching is performed over a preset long-term timescale (e.g., one hour or one day), while task computation and bitrate adaptation are performed over a preset short-term timescale (e.g., one second or other durations). During the observation period... Within the cache, the caching decision (i.e., caching strategy) remains fixed, while the bitrate is adaptively executed in each time slot.

[0040] The communication model of the two-layer collaborative video caching architecture is as follows: Assume all video files in the video library are stored in descending order of popularity. The spectrum occupied by SBS and IU within the same cell is orthogonal, with only intra-cell interference. UE With SBS In the time slot The expression for the downlink transmission rate is as follows: (10) Among them, UE Indicates user equipment , Indicates community micro base stations, Indicates the allocated bandwidth for SBS transmission. , This represents the small-scale fading coefficient. SBS With UE The distance between them Indicates user equipment Non-service micro base stations Small-scale channel gain between This represents a micro base station index that is not a service. This represents the path loss index. Indicates noise power. This indicates the transmission power of the SBS.

[0041] The D2D transmission from the IU to its serving UE adopts a one-to-one mode. with IU In the time slot The transmission rate is: (11) Among them, IU Indicates important users , This indicates the reserved bandwidth for D2D communication. This indicates the transmit power of IU. Indicates user equipment With important users Small-scale channel gain between Indicates user equipment Non-service important users Small-scale channel gain between This represents an important user index that is not part of the service. (Introduced by UE) In the time slot playback buffer size UE Indicates user equipment The expression for the playback buffer size is as follows: (12) in, Indicates the current bitrate. This represents the downlink transmission rate. A binary indicator variable is introduced. To explicitly model D2D relationships. UE In the time slot downlink transmission rate Depends on caching decisions: If UE It itself is a cached video file The IU, then at the local maximum rate If a cached video file exists, retrieve it. And within the D2D communication range of IU If the IU is not currently serving other users, then the UE will be in D2D mode. with IU In the time slot transmission rate Transmission, otherwise handled by the local SBS. With UE with IU In the time slot downlink transmission rate Provide services. Among them, Indicates the duration of a time slot.

[0042] In step S102, within a preset long time scale period, each cache node in the cache node set caches the video file according to the caching strategy.

[0043] Specifically, as described above, caching is performed over a preset long time scale period (such as one hour or one day), meaning that the video file is cached on various cache nodes in the cache node set according to the caching strategy. Task computation and bitrate adaptation are performed over a preset short time scale period (such as one second or other seconds).

[0044] Further, step S102 includes the following: The Zipf distribution with preset skewness parameters is used to model the content popularity in the community and calculate the probability of a user device requesting a video file. Each cache node estimates its local preference for video files by aggregating the request probability of its connected users, so that each cache node caches its own video files according to its local preference.

[0045] Specifically, the expression for the probability of a UE requesting a video file is as follows: (13) In the formula, This indicates the probability that the UE will request a video file. This indicates the preset skewness parameter. Indicates the total number of video files. Indicates video file The popularity exponent term is used as the numerator of the request probability. This represents the summation of the power terms of the popularity of all video files, used as the normalized denominator for the request probability.

[0046] when When the content popularity follows a uniform distribution, when When, it follows the classic Zipf distribution.

[0047] Each cache node Local preferences are estimated by aggregating the request probabilities of its connected users, and cache nodes... Video files are placed according to local preference (i.e., greedy descending preference order) (i.e., each video file is cached separately) until the cache capacity of the cache node is exhausted. The highest bitrate is stored for each selected video file. The expression for local preferences is as follows: (14) In the formula, Indicates local preferences, Represents cache node The set of neighboring users.

[0048] In step S103, each video file is divided into multiple segments according to a preset duration. Each segment corresponds to a bitrate. Within a preset short time scale period, an instantaneous experience quality model is constructed based on the current bitrate, bitrate fluctuation, initial waiting time, and playback integrity.

[0049] Specifically, each video file is divided into segments with a preset duration of [duration]. The system consists of multiple segments, each with a corresponding bitrate. Task computation and bitrate adaptation are performed within a preset short timescale period (e.g., one second or other durations). Within this preset short timescale period, a multi-index instantaneous experience quality model is constructed by integrating the current bitrate, bitrate fluctuation, initial waiting time, and playback integrity. The advantage of using these four indices to construct the instantaneous experience quality model is that it introduces playback integrity as a forward-looking real-time indicator, enabling the entire instantaneous experience quality model to be computed in real-time in each time slot and supporting real-time bitrate optimization. In conventional experience quality models, indicators representing stuttering (e.g., video stuttering time) are post-hoc quantities that can only be obtained after the entire video playback has ended, providing no gradient information for bitrate decisions in the current time slot and thus unsuitable for direct real-time optimization. This application replaces stuttering time with playback integrity, which is a continuously differentiable function of the current bitrate. In each time slot, based on the current buffer state, remaining playback time, and return download rate, the proportion of stutter-free playback can be proactively estimated, making the objective function partially derivative with respect to the current bitrate, thereby supporting efficient solutions based on KKT conditions. Meanwhile, the current bitrate, bitrate fluctuation, and initial waiting time are three indicators that represent real-time image quality, image quality smoothness, and access latency, respectively. Bitrate fluctuation is expressed as the square of the bitrate difference between adjacent time slots, so that the objective function has a closed-form optimal solution when the stuttering constraint is satisfied. The initial waiting time is characterized by five hit locations: self-hit, local micro base station hit, important user hit, regional micro base station hit, and cloud processing, reflecting the access latency differences of nodes at different levels under the two-layer collaborative caching architecture.

[0050] In one embodiment, the expression for the instantaneous experience quality model is as follows: (1) In the formula, This represents the instantaneous experience quality model. Indicates the current bitrate. Indicates bitrate fluctuation. Indicates the initial waiting time. Indicates playback completeness. The weight representing the current bitrate. The weights representing bitrate fluctuations The weight representing the initial waiting time. Weights representing playback integrity .

[0051] Specifically, the main factors affecting the instantaneous experience quality model include video resolution. Video quality volatility Initial waiting time and video buffering time The four weights in formula (1) satisfy the normalization constraint and are all non-negative. Their values ​​are used to balance the relative importance of each indicator in the instantaneous experience quality model. They can be pre-set or calibrated offline according to actual business needs and user preferences. One way is to manually specify them according to the service quality requirements of the video service type. For example, for scenes sensitive to image quality, the weight of the current bitrate can be appropriately increased, and for scenes sensitive to smoothness, the weight of playback integrity can be appropriately increased. Another way is to perform regression fitting on the subjective experience quality score of historical viewing data, with the goal of minimizing the error between the model output and the real subjective score, and calibrate the values ​​of the four weights. After calibration, the four weights remain fixed during the bitrate adaptation process.

[0052] Wherein, video resolution is defined as the average bitrate of the video being viewed, and the expression for video resolution is as follows: (15) In the formula, This indicates the number of segments in each video file. Indicates the time slot length.

[0053] The expression for video quality volatility is as follows: (16) In the formula, This indicates the bitrate at the previous moment. The initial waiting time is determined based on the content hit position, which is the hierarchical position of the requested video file within the caching architecture. Hit positions include five scenarios: self-hit, local SBS hit, IU hit, regional SBS hit, and cloud processing. A self-hit means the user device itself is an important user that has cached the requested video file, allowing local reading without wireless transmission; the value is 0 for a self-hit. A local SBS hit means the requested video file is not cached locally but is cached at a micro base station in the user's community, requiring transcoding and downlink transmission via the micro base station; the initial waiting time for a local SBS hit is... An IU hit means that the video file requested by the user is cached in a nearby IU. Upon an IU hit, the initial wait time is... A regional SBS hit means that the requested file is not cached on this community node, but is cached on other micro base stations under the coordination of the same regional server. It needs to be retrieved through scheduling by the regional server. When a regional SBS hit occurs, the inter-regional server propagation delay needs to be added to the latency required for a local SBS hit. When processing in the cloud, the round-trip latency between the cloud and SBS needs to be added. .in, This indicates the computing power of SBS. Indicates the computing power of IU. This indicates the data size of the highest bitrate segment. This indicates the amount of initial buffer data required to start playback. This indicates the transmission rate between the user equipment and the micro base station. This indicates the transmission rate between user equipment and important users.

[0054] The instantaneous experience quality model is initially defined as a weighted sum of the four indicators mentioned above. The preliminary expression of the instantaneous experience quality model is as follows: (17) in, , Since the four components in the preliminary expression of the instantaneous experience quality model are defined throughout the playback period, they can only be calculated after the video has finished playing. However, the current bitrate... These are real-time variables determined by the current system state. Therefore, this application reformulates the expression of the instantaneous experience quality model into an instantaneous form that can be calculated in real time. Specifically, using... replace ,use replace .

[0055] In one embodiment, the expression for playback integrity is as follows: (2) In the formula, t Indicates time slot, Indicates a video file. Indicates user equipment In the time slot t Video files in the buffer The amount of data, Indicates the remaining playback time. This indicates the return download speed.

[0056] Specifically, in order to replace In real-time format, this application introduces playback integrity. This metric is defined as the proportion of videos requested by a user that can be played without stuttering, assuming the current bitrate and backhaul download rate remain constant. The effect of playback integrity is twofold: First, it is a forward-looking real-time metric that can be calculated in each time slot based solely on the currently observable amount of buffer data, remaining playback time, and backhaul download rate, thus replacing the stuttering time that can only be counted after the video playback ends, allowing the experience quality model to be calculated in real time; Second, it is a continuously differentiable function of the current bitrate, which can provide effective gradient information for bitrate optimization. When the backhaul download rate is not lower than the current bitrate, it takes the value of 1 (expected no stuttering), otherwise it decreases monotonically as the current bitrate increases, thus forming an optimizable trade-off between "increasing the bitrate to improve image quality" and "reducing the bitrate to ensure playback continuity" in the objective function (i.e., the maximization problem). Therefore, the final expression of the instantaneous experience quality model is shown in formula (1).

[0057] In step S104, a maximization problem of the instantaneous experience quality model is constructed based on the instantaneous experience quality model.

[0058] Specifically, within a preset short time scale period, given a cache placement scheme (i.e., within a preset long time scale period, each cache node in the cache node set caches video files according to the caching strategy), the bitrate decision for users in each time slot is optimized to maximize the instantaneous experience quality model.

[0059] Furthermore, the expression for the maximization problem is as follows: (3) In the formula, This means finding a solution that maximizes the bitrate decision variable. Indicates user In the community Does this request a video file? The binary request indicator variable, Indicates time slot, Indicates user, Indicates a video file. Indicates the total number of time slots. Indicates community Total number of users within, Indicates the total number of video files. This indicates the computing power of the micro base station. This means that for any time slot, This indicates a user-video file pair. Indicates in time slot Micro base stations The service provides a collection of user-video file pairs. Indicates being bound by, Indicates the computing power of important users. Indicates an important user. This means for any , Represents a set of important users. Indicates in time slot By important users The service provides a collection of user-video file pairs. Indicates the downlink transmission rate of the UE. Indicates community The user set in Represents a collection of video files. Represents the set of bitrates. Indicates the first constraint. This indicates the second constraint.

[0060] Understandably, the first and second constraints ensure that the computing resources of each cache node do not exceed its maximum capacity. The current bit rate must not exceed the downlink transmission rate, which is a constraint. The current bitrate must belong to the set of bitrates.

[0061] In step S105, the Lagrange duality method based on KKT conditions and the subgradient update iterative solution are used to solve the maximization problem and obtain the continuous optimal bit rate.

[0062] In one embodiment, the step of solving the maximization problem using a Lagrange dual method based on KKT conditions and subgradient update iteration to obtain the continuous optimal bitrate includes: The current bit rate is discretely relaxed into a continuous variable, and two dual variables are introduced using the Lagrange multiplier method, transforming the maximization problem into an unconstrained optimization problem. The KKT conditions are used to solve the unconstrained optimization problem, and a closed-form solution for the optimal bit rate is obtained. The two dual variables are updated using the subgradient descent method until the change in the unconstrained optimization problem converges. The closed-form solution of the optimal bitrate is then updated by updating the two dual variables, thus obtaining the continuous optimal bitrate.

[0063] Specifically, since the bitrate of the transcoded video must belong to a set of resolution levels, the original problem (i.e., the maximization problem) cannot be solved directly. Therefore, firstly, the current bitrate needs to be discretized and relaxed into a continuous variable, that is... The value relaxes to The continuous range. Then, the Lagrange multiplier method is used to introduce dual variables. and dual variables dual variables Corresponding constraints , dual variables Corresponding constraints This transforms the maximization problem into an unconstrained optimization problem. The expression for the Lagrange function corresponding to the Lagrange multiplier method is the expression for the unconstrained optimization problem.

[0064] Furthermore, the expression for the unconstrained optimization problem is as follows: (4) In the formula, This represents an unconstrained optimization problem. , Represents the dual variable. Indicates time slot By cache node The service provides a collection of user-video file pairs. Represents edge nodes Computing capacity, Indicates community The number of cache nodes in the system.

[0065] Specifically, the Karush-Kuhn-Tucker (KKT) conditions are derived for unconstrained optimization problems. Find the partial derivative and set it to zero, as shown in formula (5).

[0066] In one embodiment, the step of solving the unconstrained optimization problem using KKT conditions to obtain a closed-form solution for the optimal bit rate includes: We use the KKT conditions to find the partial derivative of the unconstrained optimization problem with respect to the current bitrate and set it to zero; the expression for finding the partial derivative of the unconstrained optimization problem with respect to the current bitrate is as follows: (5) In the formula, This represents finding the partial derivative of an unconstrained optimization problem with respect to the current bitrate. Indicates community The collection of cache nodes in; The expression for the partial derivative of playback integrity with respect to the current bitrate is as follows: (6) In the formula, Indicates user equipment In the time slot t Video files in the buffer The amount of data, Indicates the remaining playback time. Indicates the return download speed. Indicates playback integrity; when At that time, the closed-form solution for the optimal bit rate is expressed as follows: (7) In the formula, This represents the closed-form solution for the optimal bit rate. Indicates the bitrate of the previous time slot; when Substituting formula (6) into formula (5), we obtain a cubic polynomial equation for the closed-form solution of the optimal code rate. Solving this cubic polynomial equation yields the closed-form solution for the optimal code rate. The expression for the cubic polynomial equation is as follows: (8).

[0067] In one embodiment, when updating two dual variables using subgradient descent, the two dual variables are updated according to the following update expression: (9) In the formula, Indicates step size, Indicates community The collection of cache nodes in the middle, , Represents the dual variable. Represents edge nodes Computing capacity, n This represents an edge node.

[0068] In step S106, the continuous optimal bitrate is discretized and the user bitrate is adjusted to obtain the final bitrate decision that satisfies the constraints.

[0069] In one embodiment, the step of discretizing the continuous optimal bitrate and adjusting the user bitrate to obtain the final bitrate decision that satisfies the constraints includes: Discretize the continuous optimal bitrate and map it to the nearest feasible bitrate level; Compare the loss value corresponding to the most recent bitrate level above with the loss value of the most recent bitrate level below, and select the most recent bitrate level with the larger loss value; If the nearest bitrate level corresponding to the selected large loss value violates the first and second constraints, then all loss values ​​are sorted in ascending order, and the bitrate of the user with the smallest loss value is reduced to the next level, until the bitrate of the next level satisfies the constraints.

[0070] Specifically, in the update and iteration process, in the fixed and Calculate the optimal bitrate, then update the dual variable, and repeat this process until the change in the unconstrained optimization problem converges, i.e.: , This represents the change in an unconstrained optimization problem. This represents the iterative convergence threshold. Subsequently, discretization is used to recover the continuous optimal bitrates and map them to the nearest feasible bitrate level, comparing this to the nearest bitrate level above. The corresponding loss value and the most recent bitrate level below. The corresponding loss value is selected based on the nearest bitrate level with the largest loss value. If the selected nearest bitrate level with the largest loss value violates... and Then, all loss values ​​are sorted in ascending order according to the difference between them. The user with the smallest loss value has their bitrate reduced to the next level, until the bitrate of the next level satisfies the constraint. The expression for the difference between loss values ​​is as follows: (18) In the formula, This represents the difference in loss values. This represents the loss value corresponding to the most recent bitrate level above. This indicates the loss value corresponding to the most recent bitrate level below.

[0071] The present application will be further illustrated by the following embodiments.

[0072] like Figure 2 As shown, three SBSs (i.e., M =3), each SBS is located at the center of its community. Performance evaluation is conducted on a community-by-community basis. Within the community are... Each user has an initial position that follows a Poisson point process (PPP) and moves randomly.

[0073] The content library contains One video file, preset skew parameters Bitrate set Bitrate is measured in Mbps. The number of segments in each video file. Time slot length s, the total number of time slots within each time period s. SBS transmission power dBm, IU transmit power dBm, SBS computing power Mbps, IU computing power Mbps, SBS cache size The number of files, the IU cache size One file.

[0074] Perform steps S102 to S106. Decision is made based on the final bitrate that satisfies the constraints, i.e., based on the final current bitrate that satisfies the constraints. The instantaneous quality of experience (QoE) model value for each user is calculated using formula (1).

[0075] like Figure 3 As shown, different The overall QoE comparison results under the following schemes show that: as The increase in size leads to an overall improvement in QoE for all solutions, due to the larger... This concentrates user requests on a few popular videos, thereby improving the cache hit rate of edge nodes. The proposed solution (i.e., the solution shown in the figure) consistently achieves the highest QoE because it converts the improved cache hit rate into a QoE boost through QoE-driven bitrate adaptation, promoting high-resolution (i.e., bitrate) selection while reducing stuttering probability and quality switching. In contrast, non-adaptive solutions lack QoE-aware bitrate adaptation and cannot fully utilize cache hit rate. The globally most popular caching solution, due to its lower cache hit rate, performs worse in QoE than the proposed solution.

[0076] See Figure 4 Different IU computing power The overall QoE comparison results of the following schemes show that: with As the value increases, the overall QoE of all schemes shows an upward trend. At Mbps, QoE increases in a stepwise manner because each IU serves at most one user per time slot, and the available video resolution is selected from a discrete set. ,only QoE only improves when the resolution threshold is reached. At Mbps, the overall QoE gradually stabilizes. At Mbps, the overall QoE of the proposed solution is approximately 16.1% higher than the globally most popular cache solution and higher than the non-adaptive solution. The (Mbps) plan is approximately 105.2% higher.

[0077] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

Claims

1. A video bitrate adaptive method based on a two-layer edge network and experience quality driven, characterized in that, The method includes: Build including M A two-layer collaborative video caching architecture consisting of micro base stations and several important users, the two-layer collaborative video caching architecture including a set of caching nodes; wherein, each micro base station covers a community, and each important user is a user device within the community with local caching and computing capabilities; Within a preset long time period, each cache node in the cache node set caches video files according to a caching strategy; Each video file is divided into multiple segments according to a preset duration, and each segment corresponds to a bitrate. Within a preset short time scale period, an instantaneous experience quality model is constructed based on the current bitrate, bitrate fluctuation, initial waiting time, and playback integrity. The problem of maximizing the instantaneous experience quality model is constructed based on the instantaneous experience quality model. The maximization problem is solved by using the Lagrange dual method based on KKT conditions and subgradient update iteration to obtain the continuous optimal bitrate; Discretize the continuous optimal bitrate and adjust the user bitrate to obtain the final bitrate decision that satisfies the constraints.

2. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 1, characterized in that, The step of caching video files according to a caching strategy within a preset long-term time scale period includes: The popularity of content in the community is modeled using a Zipf distribution with a preset skewness parameter, and the probability of the user device requesting a video file is calculated. Each cache node estimates its local preference for video files by aggregating the request probability of its connected users requesting video files, so that each cache node caches its respective video file according to the local preference.

3. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 1, characterized in that, The expression for the instantaneous experience quality model is as follows: （1） In the formula, This represents the instantaneous experience quality model. Indicates the current bitrate. Indicates bitrate fluctuation. Indicates the initial waiting time. Indicates playback completeness. The weight representing the current bitrate. The weights representing bitrate fluctuations The weight representing the initial waiting time. Weights representing playback integrity .

4. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 3, characterized in that, The expression for playback integrity is as follows: （2） In the formula, t represents a time slot, and f represents a video file. Indicates user equipment The video file in the buffer of time slot t The amount of data, Indicates the remaining playback time. This indicates the return download speed.

5. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 3, characterized in that, The step of constructing the maximization problem of the instantaneous experience quality model based on the instantaneous experience quality model includes: The expression for the maximization problem is as follows: （3） In the formula, This means finding a solution that maximizes the bitrate decision variable. Indicates user In the community Does this request a video file? The binary request indicator variable, Indicates time slot, Indicates user, Indicates a video file. Indicates the total number of time slots. Indicates community Total number of users within, Indicates the total number of video files. This indicates the computing power of the micro base station. This means that for any time slot, This indicates a user-video file pair. Indicates in time slot Micro base stations The service provides a collection of user-video file pairs. Indicates being bound by, Indicates the computing power of important users. Indicates an important user. This means for any , Represents a set of important users. Indicates in time slot By important users The service provides a collection of user-video file pairs. Indicates the downlink transmission rate of the UE. Indicates community The user set in Represents a collection of video files. Represents the set of bitrates. Indicates the first constraint. This indicates the second constraint.

6. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 5, characterized in that, The step of solving the maximization problem using the KKT-based Lagrange duality method and subgradient update iteration to obtain the continuous optimal bitrate includes: The current bit rate is discretely relaxed into a continuous variable, and two dual variables are introduced using the Lagrange multiplier method, transforming the maximization problem into an unconstrained optimization problem. The unconstrained optimization problem is solved using the KKT conditions to obtain a closed-form solution for the optimal bit rate. The two dual variables are updated using the subgradient descent method until the change in the unconstrained optimization problem converges. The closed-form solution of the optimal bitrate is then updated using the updated two dual variables to obtain the continuous optimal bitrate.

7. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 6, characterized in that, The step of discretizing the continuous optimal bitrate and adjusting the user bitrate to obtain the final bitrate decision that satisfies the constraints includes: Discretize the continuous optimal code rates and map them to the nearest feasible code rate level; Compare the loss value corresponding to the most recent bitrate level above with the loss value of the most recent bitrate level below, and select the most recent bitrate level with the larger loss value; If the nearest bitrate level corresponding to the selected large loss value violates the first and second constraints, then all loss values ​​are sorted in ascending order, and the bitrate of the user with the smallest loss value is reduced to the next level, until the bitrate of the next level satisfies the constraints.

8. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 6, characterized in that, The expression for the unconstrained optimization problem is as follows: （4） In the formula, This represents an unconstrained optimization problem. , Represents the dual variable. Indicates time slot By cache node The service provides a collection of user-video file pairs. Represents edge nodes Computing capacity, Indicates community The number of cache nodes in the system.

9. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 8, characterized in that, The step of solving the unconstrained optimization problem using KKT conditions to obtain a closed-form solution for the optimal bit rate includes: The partial derivative of the unconstrained optimization problem with respect to the current bitrate is calculated using the KKT conditions and set to zero; the expression for the partial derivative of the unconstrained optimization problem with respect to the current bitrate is as follows: （5） In the formula, This represents finding the partial derivative of an unconstrained optimization problem with respect to the current bitrate. Indicates community The collection of cache nodes in; The expression for the partial derivative of playback integrity with respect to the current bitrate is as follows: （6） In the formula, Indicates user equipment In the time slot t Video files in the buffer The amount of data, Indicates the remaining playback time. Indicates the return download speed. Indicates playback integrity; when At that time, the closed-form solution for the optimal bit rate is expressed as follows: （7） In the formula, This represents the closed-form solution for the optimal bit rate. Indicates the bitrate of the previous time slot; when Substituting formula (6) into formula (5), we obtain a cubic polynomial equation for the closed-form solution of the optimal code rate. Solving this cubic polynomial equation yields the closed-form solution for the optimal code rate. The expression for the cubic polynomial equation is as follows: （8）。 10. The video bitrate adaptive method based on a dual-layer edge network and experience quality driven according to claim 6, characterized in that, When updating the two dual variables using subgradient descent, the two dual variables are updated according to the following update expression: （9） In the formula, Indicates step size, Indicates community The collection of cache nodes in the middle, , Represents the dual variable. Represents edge nodes Computing capacity, n This represents an edge node.

Images