Method and device for minimizing network access delay of multi-UAV scalable video cache

Abstract

The present application relates to a method and device for minimizing the network access delay of a multi-drone scalable video cache, which provides a hierarchical cache solution for scalable video coding in a multi-drone network, adopts SVC technology to hierarchically encode video files, and establishes an optimization model to characterize the joint optimization problem of user association, drone location deployment and hierarchical cache placement strategy, thereby minimizing the total video access delay of all users in the network. In addition, in order to solve the established combined non-convex optimization problem, the problem to be optimized is decomposed into a user association optimization sub-problem model based on penalized continuous convex approximation (P‑SCA) and a drone location deployment and hierarchical cache placement optimization sub-problem model based on penalized convex difference (P‑DC), and the two sub-problem models are alternately optimized according to an iterative optimization algorithm until the iterative optimization algorithm converges. The algorithm has low complexity and effectiveness, so that significant gains are achieved in video access delay performance.

Description

Background Art

[0002] In recent years, with the continuous expansion of 5G network coverage, wireless data traffic, mainly video services, has also been growing exponentially. Faced with huge video traffic data, it is extremely challenging to meet strict quality of service (QoS) requirements only through remote cloud video services.

[0003] Edge caching based on unmanned aerial vehicles (UAVs) has been increasingly used. UAVs can be quickly deployed to hot spots and provide cached content to users. In addition, since UAVs can fly at higher altitudes, they can establish line-of-sight (LoS) channels with users with better transmission performance. Based on Scalable Video Coding (SVC) technology, each video is encoded into a base layer and multiple enhancement layers to gradually improve the video quality. However, SVC increases the difficulty of designing the caching strategy for UAVs. In addition, due to the limited cache capacity of UAVs, the layered caching strategy needs to balance the requirements of video quality and video diversity. In addition, the performance of the cache placement strategy will change with the changes in the UAV location deployment strategy and the user association strategy, which are also important factors affecting video access latency.

[0004] Therefore, it is challenging to jointly optimize the hierarchical cache deployment and drone location deployment and user association for scalable video services, and this problem has not been addressed in related technologies. Summary of the invention

[0005] The present application provides a method and device for minimizing the network access delay of a multi-drone scalable video cache, so as to minimize the total video access delay of users, and provide different video services according to the user's demand for videos of different quality levels. The technical solution of the present application is as follows:

[0006] According to a first aspect of an embodiment of the present application, a method for minimizing network access delay of a multi-UAV scalable video cache is provided, the method comprising:

[0007] Step 1: Based on a multi-UAV assisted cellular network supporting caching and multiple assumptions, a UAV caching model is constructed, wherein the cellular network includes a cellular base station and M UAVs equipped with caching units and K ground users, each of which has a cache capacity of s k Can only be used with one drone m Associated, each drone u m Can associate multiple users k ;

[0008] Step 2: Use Scalable Video Coding (SVC) technology to encode the video file in layers, and determine the user s by calculating the average delay of the download delay required for the air-to-ground link and the download delay required for the ground-to-ground link. k Decode video v n The total delay model of video access when there are l layers;

[0009] Step 3: Based on the user association strategy {a km}、UAV location deployment strategy {q m} and the tiered cache placement strategy {x mnl}Joint optimization, establish the optimization problem model P1, minimize the total video access delay of all ground users;

[0010] The step 3 specifically includes: decomposing the optimization problem model P1 into a user association optimization sub-problem model based on penalized continuous convex approximation (P-SCA) and a drone location deployment and hierarchical cache placement optimization sub-problem model based on penalized convex difference (P-DC), and alternately optimizing the two sub-problem models according to an iterative optimization algorithm until the iterative optimization algorithm converges.

[0011] Optionally, the multiple assumptions set include at least:

[0012] Altitude assumption, where the drone is assumed to stay at a fixed altitude H, which can be set as the minimum altitude at which the drone can safely fly within its service area;

[0013] The user uses the assumption that each user can only be associated with one drone, and each drone can be associated with multiple users. Define a binary association variable between a drone and a user as a km ∈{0,1}, if the ground user s k With drone u m If a km =1, otherwise a km =0;

[0014] Video layered coding assumptions, where it is assumed that in the video content library There are N SVC-encoded videos in the video file, that is, each video file is encoded into L layers to provide users with different quality services of the video;

[0015] User preference assumption, where the probability of a user requesting a video file can be expressed as The skewness parameter θ is used to characterize the concentration of user requests. The larger the value of θ is, the fewer video files are needed to satisfy most user requests.

[0016] Video layer distribution assumption, where it is assumed that the user's requests for L video layers satisfy a random uniform distribution, that is, g nl Indicates the user's response to video v n The preference of the l-th layer request of video v n The request probability of the lth layer is

[0017] The video distribution request is known assumption, in which the video request distribution is assumed to be known by analyzing the user request pattern over a period of time or inferring the request distribution information through machine learning methods.

[0018] Optionally, the optimization problem model P1 can be expressed as:

[0019]

[0020] And the constraints are:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] Among them, d min represents the minimum safe distance between any two drones; q m For drone u m The horizontal position of km For users k With drone u m Associated variable, x mnl is the cache decision variable, λ nl For video v n The request probability of the lth layer, o nl For video v n The file size of the first layer of v n The average delay of the lth layer, For users k To decode the video v n The total video access delay for l layers.

[0027] Optionally, the P-SCA-based user association optimization sub-model specifically includes:

[0028] Deployment strategy at the drone location {qm} and the tiered cache placement strategy {x mnl Given a given situation, construct a user association strategy {a km}'s optimized sub-model;

[0029] Fixed total delay The value of

[0030] Convert the user association optimization sub-model P1 into the second user association optimization sub-model P2:

[0031]

[0032] Among them, the constraints are formula (5) and formula (6);

[0033] Transform the constraint (5) into constraints (8) and (9)

[0034]

[0035]

[0036] Optionally, a penalty term is added to the second user association optimization sub-model P2 to km} is penalized for the non-binary integer solution to obtain the third optimized sub-model;

[0037] Apply the first-order Taylor expansion to the penalty term The SCA method is used to solve the approximate fourth optimized sub-model P4 of the third optimized sub-model P3.

[0038] Optionally, the optimization problem model P1 is decomposed into a user association optimization sub-problem model based on penalized continuous convex approximation (P-SCA) and a drone location deployment and hierarchical cache placement optimization sub-problem model based on penalized convex difference (P-DC), and the two sub-problem models are alternately optimized according to an iterative optimization algorithm until the iterative optimization algorithm converges, specifically including:

[0039] Using the obtained user association strategy {a km}Optimize the drone location deployment strategy m} and the tiered cache placement strategy {x mnl}.

[0040] Optionally, the user association strategy obtained by using km}Optimize the drone location deployment strategy m} and the tiered cache placement strategy {x mnl}, including:

[0041] The slack variable Y is introduced to replace the fourth optimization sub-model P4 to obtain the fifth optimization sub-model P5:

[0042] Add a penalty term ξ for all {x mnl} is penalized to obtain the sixth optimized sub-model P6;

[0043] Introducing a second slack variable Convert the sixth optimized sub-model into the seventh optimized sub-model P7;

[0044] The seventh optimized sub-model P7 is converted into an eighth optimized sub-model P8, wherein the target value of the eighth optimized sub-model P8 is the upper bound of the target value of the seventh optimized sub-model P7.

[0045] Optionally, the fourth optimized sub-model P4 and the eighth optimized sub-model P8 are alternately optimized by a block coordinate descent method (BCD), specifically including:

[0046] Step a: Initialize the drone location deployment User Association and tiered cache placement Set r←0;

[0047] Step b: Fixing and Given an initial value Use CVX to get the optimal solution of the fourth optimization submodel P4

[0048] Step c: Fix Given an initial value and Using CVX to obtain the optimal solution of the eighth optimization submodel P8

[0049] Step d: Set r←r+1 and repeat until the target value decreases by a ratio lower than the solution accuracy threshold ∈, where ∈ represents the accuracy of the optimization solution.

[0050] Optionally, the step 2 uses Scalable Video Coding (SVC) technology to perform hierarchical encoding on the video file, and determines the user s by calculating the average delay of the download delay required for the air-to-ground link and the download delay required for the ground-to-ground link. k Decode video v n When there are l layers, the total delay model of video access includes:

[0051] When a ground user downloads video layering data from a UAV via an air-to-ground link, assuming that the user downloads the video from the associated UAV based on an orthogonal channel without interference, user sk From UAV m The average delay for downloading 1 bit of data is:

[0052]

[0053] in β0 is the channel power gain at a reference distance of 1 meter, It's a drone m and users k The distance between mk For users k From the associated drone u m The achievable rate when downloading video content, P U represents the transmission power of the UAV, σ 2 represents the noise power of the user, and B represents the bandwidth allocated to each user;

[0054] When a ground user downloads video layer data from a base station via a ground-to-ground link, user s k The delay of downloading 1 bit of data from the base station is:

[0055]

[0056] in R bk For users k The achievable rate when downloading the required content from the base station, β bk From the base station to user s k The channel gain, where α ≥ 2 is the path loss component, κ < 1 represents the additional impact of scattering and non-line-of-sight links due to building obstruction, and P B represents the transmission power of the base station, represents base station and user s k The distance between.

[0057] In this case, when the user s k When downloading the required content from the base station, the achievable rate can be expressed as Where P B Indicates the transmit power of the base station.

[0058] According to a second aspect of an embodiment of the present application, a device for minimizing network access delay of a multi-UAV scalable video cache is provided, comprising:

[0059] processor;

[0060] a memory for storing instructions executable by the processor;

[0061] Wherein, the processor is configured to execute the instructions to implement the multi-UAV scalable video cache network access delay minimization method described in any one of the methods mentioned in the first aspect above.

[0062] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0063] The present application relates to a method and device for minimizing the access delay of a multi-drone scalable video cache network, providing a hierarchical cache solution for scalable video coding in a multi-drone network, adopting SVC technology to hierarchically encode video files, and establishing an optimization model to characterize the joint optimization problem of user association, drone location deployment and hierarchical cache placement strategy, so as to minimize the total video access delay of all users in the network. In addition, in order to solve the established combined non-convex optimization problem, the problem to be optimized is decomposed into a user association optimization sub-problem model based on penalized continuous convex approximation (P-SCA) and a drone location deployment and hierarchical cache placement optimization sub-problem model based on penalized convex difference (P-DC), and the two sub-problem models are alternately optimized according to an iterative optimization algorithm until the iterative optimization algorithm converges and ensures the low complexity and effectiveness of the algorithm, so as to achieve significant gains in video access delay performance.

[0064] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0066] Figure 1 is a schematic diagram of an architecture for constructing a downlink system supporting multi-UAV communication according to an exemplary embodiment;

[0067] Figure 2 It is a flowchart of a method for minimizing network access delay of multi-UAV scalable video cache according to an exemplary embodiment;

[0068] Figure 3 is a schematic diagram of the convergence process of Algorithm 1 according to an exemplary embodiment;

[0069] Figure 4 Optimized drone deployment locations and drone-user associations are exemplified;

[0070] Figure 5 The effect of different cache capacities of drones on the total delay of video access in different schemes is shown by way of example;

[0071] Figure 6 The relationship between the total delay of video access and the number of video files N is shown by way of example;

[0072] Figure 7 A schematic diagram exemplarily illustrates the effect of the skew parameter θ of the present application on the total delay of video access. DETAILED DESCRIPTION

[0073] In order to enable ordinary persons in the art to better understand the technical solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0074] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.

[0075] The application scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field can know that with the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0076] In this paper, the N×M dimensional complex and real matrix sets are respectively used and It means that the k-th diagonal element of the square matrix S is [S] {k,k} Indicates that the transpose and conjugate transpose of the matrix M are represented by M T and M H express.

[0077] like Figure 2 As shown, the present application provides a method for minimizing the network access delay of multi-UAV scalable video cache, which specifically includes:

[0078] Step 1: Based on a multi-UAV assisted cellular network supporting caching and multiple assumed conditions, a UAV caching model is constructed, wherein the cellular network includes a cellular base station and M UAVs equipped with caching units, and K ground users, wherein each user can only be associated with one UAV, and each UAV can be associated with multiple users.

[0079] like Figure 1As shown, the present application provides a multi-drone assisted cellular network supporting caching, which includes a cellular base station and M drones equipped with caching units. The drone is connected to the base station via a wireless link, and the base station is connected to the original video content server via a backhaul link.

[0080] Drones are deployed in the network to provide cached video content to K ground users. The sets of drones and ground users are represented as and Ground users k The horizontal coordinate of The horizontal coordinates of the base station are expressed as The base station height is denoted as H b . UAV m The horizontal position of Indicates that the cache capacity is set to C m >0, unit is bit.

[0081] This application makes the following assumptions for the convenience of calculation:

[0082] First, the altitude assumption condition assumes that the UAV stays at a fixed altitude H, which can be set as the minimum altitude at which the UAV can fly safely within its service area.

[0083] Second, the user uses the assumption that each user can only be associated with one drone, while each drone can be associated with multiple users. Define a binary association variable between a drone and a user as a km ∈{0,1}, specifically, if the ground user s k With drone u m If a km =1, otherwise a km = 0. Then we get a constraint condition:

[0084] Third, the assumption of video layered coding is that in the video content library There are N SVC-encoded videos in the video file, that is, each video file is encoded into L layers to provide users with different quality services of the video. Specifically, the first layer of the video is the base layer, and the subsequent layers are enhancement layers. Define video v n The file size of the first layer is o nl ,1≤n≤N,1≤l≤L, usually the larger the l is, the smaller the data volume of the video layer is, that is, o n1 ≥…≥o nL .

[0085] Fourth, user preference assumption, assuming that the user's preferences for different video contents (that is, the request probability of a video file) can be expressed as The skewness parameter θ is used to characterize the concentration of user requests. The larger the value of θ, the fewer video files are needed to meet most user requests. In addition to video popularity, user video quality preferences must also be considered. After SVC encoding, users can only decode the video data of layer l when they receive data from layer 1 to layer l.

[0086] Fifth, the video layer distribution assumption, assuming that the user's request for L video layers satisfies a random uniform distribution, that is, Among them, g nl Indicates the user's response to video v n The preference of the l-th layer request of video v n The request probability of the lth layer is

[0087] Sixth, the assumption that video request distribution is known. Assuming that the video request distribution is known, in fact, relevant information can be obtained by analyzing the user request pattern over a period of time, or the request distribution information can be inferred through machine learning methods.

[0088] During off-peak request periods, the drone obtains video layer data from the base station through the backhaul link. Define the cache decision variable x mnl ∈{0,1}, when the video file v n The lth layer is cached on drone u m In the middle, x mnl =1, otherwise x mnl = 0. Due to the size limit of the drone cache capacity, the following constraints are obtained: Among them C m Indicates drone u m The maximum size that can be cached.

[0089] In an exemplary embodiment of the present application, a user can obtain the required video layer from the associated drone via an air-to-ground link, thereby generating a lower access latency. If the drone does not cache the video layer data required by the user, the user will download the required content from the base station via a ground-to-ground link because the base station caches the entire video content library.

[0090] Step 2: Use Scalable Video Coding (SVC) technology to encode the video file in layers, and determine the user s by calculating the average delay of the download delay required for the air-to-ground link and the download delay required for the ground-to-ground link. k Decode video v n The total delay model of video access when there are l layers.

[0091] In some embodiments, the ground user downloads the video layering data from the drone by computing an air-to-ground link (i.e., an air-to-ground link). Since the drone flies at a higher altitude, a LoS channel can be established between it and the ground user. m To users k The channel gain can be expressed as Where β0 is the channel power gain at a reference distance of 1 meter, It's a drone m and users k The distance between.

[0092] Assume that a user downloads a video from a drone associated with him / her over an orthogonal channel without interference. k From the associated drone u m When downloading video content, the achievable rate can be expressed as Where P U represents the transmission power of the UAV, σ 2 represents the noise power of the user, and B represents the bandwidth allocated to each user. In this case, user s k From UAV m The average delay (in seconds) for downloading 1 bit of data is:

[0093]

[0094] in

[0095] In some other embodiments, consider that a ground user downloads the video layer data from the base station via a ground-to-ground link (ie, a ground-to-ground link). k The channel gain is expressed as where α ≥ 2 is the path loss component and κ < 1 represents the additional impact of scattering and non-line-of-sight links due to building obstructions. k When downloading the required content from the base station, the achievable rate can be expressed as Where P B Indicates the transmit power of the base station.

[0096] So user s k The delay (in seconds) of downloading 1 bit of data from the base station can be expressed as:

[0097]

[0098] in

[0099] If user s k and drone um is associated, and u m Cached users k Required video v n The lth layer (i.e. x mnl =1), then the download delay is Otherwise (i.e. x mnl =0), the download delay is Thus we get v n The average delay of the lth layer is

[0100] It is worth noting that if you want to decode the video v n The data content of the first layer of the video needs to be transmitted to the user from the first layer to the first layer, that is, a total of bits. Therefore, if user s k To decode the video v n The total video access delay can be expressed as

[0101] Step 3: Based on the user association strategy {a km}、UAV location deployment strategy {q m} and the tiered cache placement strategy {x mnl}Joint optimization, establish an optimization problem model to minimize the total video access delay of all ground users.

[0102] The optimization problem model P1 provided in this application can be expressed as:

[0103]

[0104] And the constraints are:

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] Among them, d min represents the minimum safe distance between any two drones; q m For drone u m The horizontal position of km For users k With drone u m Associated value, λnl For video v n The request probability of the lth layer, x mnl is the cache decision variable, λ nl For video v n The request probability of the lth layer.

[0111] In some embodiments, the method is based on the user association strategy {a km}、UAV location deployment strategy {q m} and the tiered cache placement strategy {x mnl Joint optimization, establish the optimization problem model P1, minimize the total video access delay of all ground users, specifically including:

[0112] The optimization problem model P1 is decomposed into a user association optimization subproblem model based on penalized continuous convex approximation (P-SCA) and a drone location deployment and hierarchical cache placement optimization subproblem model based on penalized convex difference (P-DC), and the two subproblem models are alternately optimized according to an iterative optimization algorithm until the iterative optimization algorithm converges.

[0113] The objective function of the optimization problem model P1 in this application is a non-smooth function, which contains binary variables, complex rate expressions and mutually coupled optimization variables. In addition, constraint (7) is a non-convex constraint. Therefore, the optimization problem model P1 is a mixed integer non-convex optimization problem that is difficult to solve directly.

[0114] Therefore, the optimization problem model P1 is decomposed into two sub-problem models, including a user association optimization sub-model based on P-SCA and a drone location deployment and hierarchical cache placement optimization sub-model based on P-DC. Then, an overall iterative optimization algorithm is proposed to alternately optimize the two sub-problems until the algorithm converges.

[0115] 1. User association optimization sub-model based on P-SCA

[0116] Deployment strategy at the drone location {q m} and the tiered cache placement strategy {x mnl Given a given situation, construct a user association strategy {a km}'s optimized sub-model.

[0117] In this case, the total delay The value of is fixed. The user association optimization sub-model P1 can be converted into the second user association optimization sub-model P2:

[0118]

[0119] st(5),(6).

[0120] And transform the binary constraint (5) into the intersection of the following two constraints:

[0121]

[0122]

[0123] Obviously, any feasible solution that satisfies equations (8) and (9) must also satisfy constraint (5), and vice versa. This is because 0≤a km ≤1, so so Combined with constraint (9), we can get That is to say In summary, we can get

[0124] Furthermore, in order to obtain {a km} binary integer solution, by adding a penalty term to the objective function {a km} is penalized for the non-binary integer solution, and the third optimization sub-model is obtained:

[0125]

[0126] The constraints are formulas (6) and (8).

[0127] Wherein, η represents a penalty factor. When the penalty factor η is large enough, that is, η>>1, the third optimization sub-model P3 is equivalent to the second user association optimization sub-model P2.

[0128] However, the third optimization submodel P3 is still a non-convex optimization problem because it contains a non-convex objective function.

[0129] This application uses the first-order Taylor expansion to The SCA method is used to solve the approximation problem of the third optimization submodel P3.

[0130] Specifically, when performing the rth iteration, at a given local point Perform a first-order Taylor expansion at Therefore, the third optimized sub-model P3 can be approximated as the fourth optimized sub-model P4:

[0131]

[0132] The constraints are formula (6) and formula (8).

[0133] It can be verified that the fourth optimization submodel P4 is a convex optimization problem. This type of problem can be effectively solved directly by the standard solution of convex optimization problems or existing convex optimization solvers such as CVX. In addition, since the fourth optimization submodel P4 uses The lower bound expression replaces Therefore, any feasible solution of the fourth optimization submodel P4 must also be a feasible solution of the third optimization submodel P3, and thus the target value of the fourth optimization submodel P4 is the upper bound of the target value of the third optimization submodel P3.

[0134] 2. Optimization model based on drone location deployment and hierarchical cache placement

[0135] This application uses the obtained user association strategy {a km} to optimize the drone location deployment strategy {q m} and the tiered cache placement strategy {x mnl In order to simplify the problem, a slack variable Υ is introduced to replace the fourth optimization sub-model P4, and the fifth optimization sub-model P5 is obtained:

[0136]

[0137] And the constraints are:

[0138]

[0139]

[0140] and constraints (3) and (4).

[0141] This application adopts the method of adding penalty items to deal with constraint condition (4). Specifically,

[0142] First, transform constraint (4) into the intersection of the following two constraints:

[0143]

[0144]

[0145] Similarly, a feasible solution that satisfies both constraints (12) and (13) will also satisfy constraint (4), and vice versa.

[0146] Then, a penalty term ξ is added for all {x mnl The non-binary integer solution of} is penalized to obtain the sixth optimization sub-model:

[0147]

[0148] And the constraints are (3) and (10)-(12).

[0149] It can be proved that when ξ>>1, the sixth optimization submodel of the problem is equivalent to the fifth optimization submodel.

[0150] In this application, a second slack variable is introduced Convert the sixth optimized sub-model into the seventh optimized sub-model P7:

[0151]

[0152] The constraints are:

[0153]

[0154]

[0155] and constraints (3), (11), and (12).

[0156] Obviously, in the optimal solution of the seventh optimization submodel P7, the left side of constraint (15) must be equal to the right side. Specifically, we can use proof by contradiction to prove that in the optimal solution of the seventh optimization submodel P7, constraint (15) satisfies the strict inequality, that is, At this point, you can always reduce The value of Increase until both sides of formula (15) are equal, all other constraints in the seventh optimization sub-model P7 are still satisfied and the target value does not change.

[0157] This shows that although formula (15) satisfies the optimal solution of the equality constraint, it contradicts the assumption, so the assumption is invalid. Finally, it can be concluded that the seventh optimization submodel P7 is equivalent to the sixth optimization submodel P6. However, since constraints (11), (14) and (15) are all non-convex constraints and the objective function is also a non-convex function, the seventh optimization submodel P7 is still a non-convex optimization problem.

[0158] In order to deal with this problem, the present application uses the SCA method to obtain the approximate problem of the seventh optimization submodel P7, and adopts the DC difference method to deal with the coupled variables Specifically, Next, the convex and Processing, at the rth iteration, at the given local point and Perform a first-order Taylor expansion on it and obtain the corresponding lower bound expressions:

[0159]

[0160]

[0161] Substituting the lower bound expression on the right side in formula (16) and (17) into the original expression, we can get The upper bound expression of is:

[0162]

[0163] In order to deal with constraint (11), this application transforms the convex term ‖q i -q j ‖ 2 Linearization, when performing the rth iteration, and The following inequality is obtained at two local points:

[0164]

[0165] It should also be noted that R mk For ‖q m -w k ‖ 2 The term is convex by calculating the logarithmic function at a given local point in the rth iteration Applying the first-order Taylor expansion, we get R mk The lower bound expression of

[0166]

[0167] in

[0168] By respectively i -q j ‖ 2 Item, R mk Item and The terms are replaced by the corresponding lower bound expressions given in equations (19), (20) and (16), and the Replace the upper bound expression with The approximate problem of the seventh optimization sub-model P7, that is, the eighth optimization sub-model P8, can be obtained:

[0169]

[0170] The constraints of the eighth optimization sub-model are:

[0171]

[0172]

[0173]

[0174] And formulas (3) and (12).

[0175] The eighth optimization submodel P8 is a standard form of convex optimization problem, which can be solved using existing convex optimization solvers, such as CVX. In addition, because the present application uses upper and lower bound approximate expressions, the constraints of the eighth optimization submodel P8 are stricter than those of the seventh optimization submodel P7.

[0176] Therefore, the target value of the eighth optimization sub-model P8 serves as the upper bound of the target value of the seventh optimization sub-model P7, and its feasible solution is also the feasible solution of the seventh optimization sub-model P7.

[0177] Based on the above discussion, the optimization problem model P1 can be solved by alternately optimizing the fourth optimization sub-model P4 and the eighth optimization sub-model P8 through the block coordinate descent method (BCD).

[0178] Specifically, in Algorithm 1 provided in this application, a decreasing target value can be obtained in each iteration through the BCD algorithm, and converges to the (Karush-Kuhn-Tucker) KKT solution after several iterations. In addition, two standard convex optimization subproblems will be solved in each iteration. Among them, the complexity of Algorithm 1 is O((KMNL) 3.5 log(1 / ∈), ∈ represents the accuracy of the optimization solution. The specific steps of Algorithm 1 are:

[0179] Step a: Initialize the drone location deployment User Association and tiered cache placement Set r←0 to loop;

[0180] Step b: Fixing and Given an initial value Use CVX to get the optimal solution of the fourth optimization submodel P4

[0181] Step c: Fix Given an initial value and Using CVX to obtain the optimal solution of the eighth optimization submodel P8

[0182] Step d: Set r←r+1 and repeat until the target value decreases by a ratio lower than the solution accuracy threshold ∈.

[0183] In some exemplary embodiments, the present application evaluates the performance of a multi-UAV scalable video cache network access delay minimization method provided in the present application through some specific experimental processes and experimental results.

[0184] like Figure 3 As shown in the figure, the convergence process of Algorithm 1 is shown. It can be clearly seen that Algorithm 1 converges within 10 iterations, which reflects the convergence and low complexity of Algorithm 1. Figure 4 As shown in the figure, the optimized drone deployment location and drone-user association are shown. As can be seen from the figure, the number of users associated with each drone may be different, and the location deployment of drones and the association of users affect each other. Therefore, these two strategies should be jointly designed and optimized with the hierarchical cache placement strategy.

[0185] For example, a multi-UAV-assisted cellular network for scalable video caching service is constructed, which consists of a ground base station, M = 2 UAVs and K = 10 ground users, which are randomly distributed in an area of ​​1.6 × 1.6 square kilometers, such as Figure 4 As shown. The base station is located at L b =[-800,-800] T , height H b =20m, in Figure 4 The lower left corner is represented by a rectangle. The video content library contains N SVC-encoded video files, each of which is encoded into L = 5 layers. The user's preference for different video files follows a Zipf distribution with a skewness parameter of θ, and the requests for the L layers in each video follow a uniform distribution. For simplicity, it is assumed that all drones equipped with cache units are deployed at an altitude of H = 100 meters. Unless otherwise stated, this application assumes that N = 20, θ = 0.8, If the drone does not cache the video layer data required by the user, the user will download the corresponding video layer from the base station through the ground-to-ground link. For the channel parameters, set σ 2 =-110dBm, β0 = -50dB, P U =0.1W, P B =0.1W, α=2.2, κ=0.1.

[0186] In order to better reflect the superiority of the proposed algorithm performance, the following three benchmark solutions are compared with the present application solution:

[0187] 1) Random caching scheme. The drone randomly caches different layers of video data;

[0188] 2) K-Means clustering scheme. First, K-means clustering is performed on users to obtain the drone location deployment strategy and user association strategy. Then, based on the clustering results, the optimization problem model P1 is solved to obtain the hierarchical video cache placement result.

[0189] 3) Base layer priority scheme: Each drone caches the base layer of all video files first, and then caches the subsequent enhancement layers. This scheme can improve the diversity of videos and provide more users with videos with basic quality assurance.

[0190] Figure 5 Reflects the different cache capacities C of drones m Impact on the total video access latency of different schemes. It can be seen that increasing the cache capacity of the drone can reduce the total video access latency. The larger the cache capacity of the drone, the more obvious the performance gain. This is because as the cache capacity of the drone increases, more requests can be satisfied locally without having to connect to the remote base station video server through the backhaul link.

[0191] Figure 6 This shows the relationship between the total video access delay and the number of video files N. It can be seen that the total video access delay decreases as N increases. This is because the cache size of the drone is fixed. If the size of the video file requested by the user exceeds the maximum cache capacity of the drone, the user needs to obtain the corresponding video content from the base station, which will cause additional delay.

[0192] In addition, Figure 7 As shown, this is a schematic diagram of the effect of the skewness parameter θ on the total latency of video access. As θ increases, the popularity distribution of videos, that is, the distribution of users' preferences for video files, shows a gradually steepening trend. When θ is large enough, most of the user's needs are concentrated on only a few videos. Once the drone caches these videos, most of the user's requests can be met, and the total latency of video access will also be reduced. This shows that the effectiveness of the caching strategy increases with the steepness of the video popularity distribution.

[0193] Compared with the K-Means clustering scheme, this application reflects the gains brought by the joint optimization of drone location deployment and user association; compared with the random cache scheme and the base layer priority scheme, this application reflects the additional gains of the layered cache placement strategy.

[0194] The present application provides a method for minimizing the network access delay of a multi-UAV scalable video cache. In order to meet users' different quality requirements for video content, SVC technology is used to perform layered encoding of video files, and a layered caching scheme is used to cache different layered data of videos on multiple UAVs. An optimization model is established to characterize the joint optimization problem of user association, UAV location deployment, and layered cache placement strategy to minimize the total video access delay of all users in the network. In addition, in order to solve the established combined non-convex optimization problem, the problem to be optimized is split into two sub-problem models, and the P-SCA method and the P-DC method are respectively used to perform joint iterative optimization on these two sub-problems. It only takes a few iterations to converge to a result that meets the KKT condition, ensuring the low complexity and effectiveness of the algorithm.

[0195] Those skilled in the art will appreciate that various aspects of the present application may be implemented as a system, method or program product. Therefore, various aspects of the present application may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to as "circuit", "module" or "system" herein.

[0196] In some possible implementations, the electronic device according to the present application may include at least one processor and at least one memory. The memory stores program code, and when the program code is executed by the processor, the processor executes the operation data management method according to various exemplary implementations of the present application described above in this specification. For example, the processor may execute the steps in the operation data management method.

[0197] Furthermore, the device for minimizing the access delay of a multi-UAV scalable video cache network according to this embodiment of the present application can execute the steps of the method for minimizing the access delay of a multi-UAV scalable video cache network mentioned in the above embodiment.

[0198] In an exemplary embodiment, various aspects of a method and apparatus for minimizing network access delay of a multi-drone scalable video cache provided by the present application can also be implemented in the form of a program product, which includes program code. When the program product is run on a computer device, the program code is used to enable the computer device to execute the steps of the method for maximizing the quality of experience in a multi-antenna drone video transmission system according to various exemplary embodiments of the present application described above in this specification.

[0199] It should be noted that, although several units or subunits of the device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided into multiple units to be embodied.

[0200] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0201] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0202] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable image scaling device to produce a machine, so that the instructions executed by the processor of the computer or other programmable image scaling device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0203] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable image scaling device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, the instruction device being implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0204] These computer program instructions may also be loaded onto a computer or other programmable image scaling device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0205] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0206] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for minimizing network access delay of multi-UAV scalable video cache, characterized in that: The method comprises: Step 1: Based on a multi-UAV assisted cellular network supporting caching and multiple assumptions, a UAV caching model is constructed, wherein the cellular network includes a cellular base station and M UAVs equipped with caching units and K ground users, each of which has a cache capacity of s k Can only be used with one drone m Associated, each drone u m Can associate multiple users k ; Step 2: Use Scalable Video Coding (SVC) technology to encode the video file in layers, and determine the user s by calculating the average delay of the download delay required for the air-to-ground link and the download delay required for the ground-to-ground link. k Decode video v n The total delay model of video access when there are l layers; Step 3: Based on the user association strategy {a km }、UAV location deployment strategy {q m } and hierarchical cache placement strategy {H mnl }Joint optimization, establish the optimization problem model P1, minimize the total video access delay of all ground users; The step 3 specifically includes: decomposing the optimization problem model P1 into a user association optimization sub-problem model based on penalized continuous convex approximation (P-SCA) and a drone location deployment and hierarchical cache placement optimization sub-problem model based on penalized convex difference (P-DC), and alternately optimizing the two sub-problem models according to an iterative optimization algorithm until the iterative optimization algorithm converges; The multiple assumptions set include at least: Altitude assumption, where the drone is assumed to stay at a fixed altitude H, which can be set as the minimum altitude at which the drone can safely fly within its service area; The user uses the assumption that each user can only be associated with one drone, and each drone can be associated with multiple users. Define a binary association variable between a drone and a user as a km ∈{0,1}, if the ground user s k With drone u m If a km =1, otherwise a km =0; Video layered coding assumptions, where it is assumed that in the video content library There are N SVC-encoded videos in the video file, that is, each video file is encoded into L layers to provide users with different quality services of the video; User preference assumption, where the probability of a user requesting a video file can be expressed as The skewness parameter θ is used to characterize the concentration of user requests. The larger the value of θ is, the fewer video files are needed to satisfy most user requests. Video layer distribution assumption, where it is assumed that the user's requests for L video layers satisfy a random uniform distribution, that is, g nl Indicates the user's response to video v n The preference of the l-th layer request of video v n The request probability of the lth layer is The video request distribution is known, where the video request distribution is assumed to be known by analyzing the user request pattern over a period of time or inferring the request distribution information through machine learning methods; The optimization problem model P1 is expressed as: And the constraints are: Among them, C m For drone u m The cache capacity, d min represents the minimum safe distance between any two drones; q m For drone u m The horizontal position of km For users k With drone u m Associated variable, x mnl is the cache decision variable, λ nl For video v n The request probability of the lth layer, o nl For video v n The file size of the first layer of Request v for the user n The average delay of the lth layer, For users k To decode the video v n The total video access delay for l layers.

2. The method according to claim 1, characterized in that The P-SCA-based user association optimization sub-model specifically includes: Deployment strategy at the drone location {q m } and the tiered cache placement strategy {x mnl Given a given situation, construct a user association strategy {a km }'s optimized sub-model; Fixed total delay The value of Convert the user association optimization sub-model P1 into the second user association optimization sub-model P2: Among them, the constraints are formula (5) and formula (6); Convert binary constraint (5) into constraints (8) and (9) 3. The method according to claim 2, characterized in that In the second user association optimization sub-model P2, a penalty term is added to the objective function {a km } is penalized for the non-binary integer solution to obtain the third optimized sub-model; Apply the first-order Taylor expansion to the penalty term The SCA method is used to solve the approximation of the third optimized sub-model P3 as the fourth optimized sub-model P4.

4. The method according to claim 3, characterized in that The optimization problem model P1 is decomposed into a user association optimization sub-problem model based on penalized continuous convex approximation (P-SCA) and a drone location deployment and hierarchical cache placement optimization sub-problem model based on penalized convex difference (P-DC), and the two sub-problem models are alternately optimized according to an iterative optimization algorithm until the iterative optimization algorithm converges, specifically including: Using the obtained user association strategy {a km }Optimize the drone location deployment strategy m } and the tiered cache placement strategy {x mnl }.

5. The method according to claim 4, characterized in that The user association strategy obtained by using km }Optimize the drone location deployment strategy m } and the tiered cache placement strategy {x mnl }, including: The slack variable Y is introduced to replace the fourth optimization sub-model P4 to obtain the fifth optimization sub-model P5: Add a penalty term ξ for all {x mnl } is penalized to obtain the sixth optimized sub-model P6; Introducing a second slack variable Convert the sixth optimized sub-model into the seventh optimized sub-model P7; The seventh optimized sub-model P7 is converted into an eighth optimized sub-model P8, wherein the target value of the eighth optimized sub-model P8 is the upper bound of the target value of the seventh optimized sub-model P7.

6. The method according to claim 5, characterized in that The fourth optimized sub-model P4 and the eighth optimized sub-model P8 are alternately optimized by a block coordinate descent method (BCD), specifically including: Step a: Initialize the drone location deployment User Association and tiered cache placement Set r←0; Step b: Fixing and Given an initial value Use CVX to get the optimal solution of the fourth optimization submodel P4 Step c: Fix Given an initial value and Using CVX to obtain the optimal solution of the eighth optimization submodel P8 Step d: Set r←r+1 and repeat until the target value decreases by a ratio lower than the solution accuracy threshold ∈, where ∈ represents the accuracy of the optimization solution.

7. The method according to claim 4, characterized in that The step 2 uses Scalable Video Coding (SVC) technology to perform hierarchical encoding on the video file, and determines the user s by calculating the average delay of the download delay required for the air-to-ground link and the download delay required for the ground-to-ground link. k Decode video v n When there are l layers, the total delay model of video access includes: When a ground user downloads video layering data from a UAV by computing an air-to-ground link, it is assumed that the user downloads the video from the associated UAV based on an orthogonal channel without interference. k From UAV m The average delay for downloading 1 bit of data is: in β0 is the channel power gain at a reference distance of 1 meter, It's a drone m and users k The distance between mk For users k From the associated drone u m The achievable rate when downloading video content, P U represents the transmission power of the UAV, σ 2 represents the noise power of the user, and B represents the bandwidth allocated to each user; When a ground user downloads video layer data from a base station via a ground-to-ground link, user s k The delay of downloading 1 bit of data from the base station is: in R bk For users k The achievable rate when downloading the required content from the base station, β bk From the base station to user s k The channel gain, where α ≥ 2 is the path loss component, κ < 1 represents the additional impact of scattering and non-line-of-sight links due to building obstruction, and P B represents the transmission power of the base station, represents base station and user s k The distance between In this case, when the user s k When downloading the required content from the base station, the achievable rate can be expressed as Where P B Indicates the transmit power of the base station.

8. A device for minimizing network access delay of multi-UAV scalable video cache, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the multi-UAV scalable video cache network access delay minimization method as described in any one of claims 1 to 7.

Images