A joint allocation method of communication and computing resources to minimize offloading cost in VR systems

Abstract

The present invention relates to a joint allocation method of communication and computing resources for minimizing the offloading cost of a VR system based on NOMA-MEC in CCHN, which can be used in the offloading decision α of VR devices. m and cellular resource block selection β m,n When the downlink transmission power p is known, the continuous convex approximation algorithm is used to jointly optimize the downlink transmission power p m , downlink transmission time #imgabs0# MEC server calculation frequency #imgabs1# and VR device calculation frequency #imgabs2# The present invention can minimize the total CCHN computing offloading cost of VR devices in the NOMA‑MEC system by reasonably allocating communication and computing resources while meeting the constraints of VR device latency and energy consumption, and has low complexity.

Description

Technical Field

[0001] The present invention relates to the technical field of resource allocation of a VR system based on NOMA-MEC in CCHN, and in particular to a joint allocation method of communication and computing resources with minimized unloading cost of a VR system based on NOMA-MEC in CCHN, including power and time of downlink transmission, and computing frequency of a MEC server and a VR device. Background Art

[0002] VR is a high-end human-computer interface interaction technology. With the development of society, 5G technology has spawned a new industry chain of diversified VR applications, such as 360° VR videos and online VR games. In order to meet the needs of massive VR services to complete ultra-large amounts of data-intensive computing and transmission in a very short time, MEC has been proposed as one of the key driving forces. By offloading computing tasks to edge servers for processing, the pressure on VR devices can be greatly alleviated. However, the scarcity of spectrum resources is still an inevitable bottleneck, which seriously affects the performance of MEC networks. In order to pursue higher spectrum efficiency, NOMA and CCHN are introduced. NOMA allows multiple VR devices to share the same time-frequency resources, and CCHN allows VR devices without cognitive radio (CR) capabilities to obtain CR services using any CR technology. The CR router in CCHN can be used as a MEC server closer to the VR device, providing VR devices with faster access and lower jitter services. In addition, the global greenhouse gas emission problem makes it crucial to study green communications, which can not only extend the life of VR devices, but also reduce the maintenance costs of MEC servers.

[0003] Therefore, considering the MEC server computing cost, MEC server transmission cost and CRB spectrum occupation cost, it is very important to study the allocation of communication and computing resources under the constraints of VR device latency and energy consumption, and to minimize the total computing offloading cost of the NOMA-MEC-based VR system in CCHN. Summary of the invention

[0004] In view of the problems existing in the prior art, the purpose of the present invention is to provide a resource allocation method for a VR system based on NOMA-MEC in CCHN, which minimizes the total CCHN computing offloading cost of VR devices in the NOMA-MEC system by reasonably jointly allocating communication and computing resources while satisfying the constraints of VR device latency and energy consumption.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A method for jointly allocating communication and computing resources in CCHN with minimization of the cost of VR system offloading based on NOMA-MEC, the method comprising the following steps:

[0007] Step 1: Place a group of VR devices in the network and a set of available Uninstallation decision α for a given VR device m and CRB selects β m,n ; Given the initial step size θ(0) and the step size update factor δ, the initial iteration index k off =1 and k loc =1, maximum number of iterations Γ;

[0008] Step 2: Generate and uninstall VR device set Local VR device set Uninstall CRB Set Local CRB Set The uninstalled VR device set and the local VR device set are further represented as and

[0009] Step 3: Uninstall the VR device set and uninstall CRB set Next, the objective function and constraints of the offloading communication and computing resource allocation problem P1 are obtained;

[0010] Step 4: Perform constraint equivalence transformation of offloading communication and computing resource allocation problem P1;

[0011] Step 5: Randomly generate an initial feasible solution set that satisfies the constraints of the offloading communication and computing resource allocation problem P1, namely

[0012] Step 6: offload the convex approximation of the communication and computing resource allocation problem P1 to obtain the convex problem P2;

[0013] Step 7: Solve the convex problem P2 to obtain the solution And update the optimal solution

[0014] Step 8: Update the feasible solution set of problem P2 constraints Where θ(k off ) is the step size, and its update method is θ(k off )=θ(k off -1)(1-δθ(k off -1);

[0015] Step 9, k off =k off +1, number of iterations k off >Γ, go to step 10, otherwise go to step 6;

[0016] Step 10: Set up local VR device and local CRB set Next, we obtain the objective function and constraints of the local communication and computing resource allocation problem P3;

[0017] Step 11: perform constraint equivalence transformation of the local communication and computing resource allocation problem P3;

[0018] Step 12: Randomly generate an initial feasible solution set that satisfies the local communication and computing resource allocation problem P3 constraints, namely

[0019] Step 13: Convex approximation of the local communication and computing resource allocation problem P3 to obtain the convex problem P4;

[0020] Step 14: Solve the convex problem P4 and obtain the solution And update the optimal solution

[0021] Step 15: Update the feasible solution set of problem P4 constraints Where θ(k loc ) is the step size, and its update method is θ(k loc )=θ(k loc -1)(1-δθ(k loc -1);

[0022] Step 16, k loc =k loc +1, number of iterations k loc >Γ, go to step 17, otherwise go to step 13;

[0023] Step 17: Output the uninstall decision α on a given VR device m and CRB selects β m,n Solution set of offloading communication and computing resource allocation problem under and local communication and computing resource allocation problem solutions Corresponding unloading cost and local costs And system cost

[0024] In step 3, the objective function and constraints of the offloading communication and computing resource allocation problem P1 are as follows:

[0025] (1) Objective function: Unloading cost C off Calculate the MEC server computing cost for VR devices MEC server transmission cost and CRB spectrum occupation costs

[0026]

[0027] Among them, η MEC >0 is the fee that the MEC server needs to pay for each unit of energy consumed. The fee that VR devices need to pay for each unit time CRB n is occupied, κ MEC >0 is the effective capacitance coefficient of the MEC server. m is the 2D FOVs data size of the task, ρ m The number of CPU cycles required to process one bit; is the computing frequency assigned to VR device m by the MEC server, p m is the downlink transmission power allocated to VR device m, is the downlink transmission time using CRB n;

[0028] (2) Constraint C1: The computing resources of the MEC server are limited, and the sum of the computing frequencies allocated to VR devices cannot exceed the maximum computing frequency of the MEC.

[0029]

[0030] (3) Constraint C2: Indicates the downlink transmission time using CRB n. In order to ensure the transmission of data required by VR device m,

[0031]

[0032] Among them, O m is the data size of 3D FOVs, R m is the achievable data rate of VR device m;

[0033] (4) Constraint C3: The VR device is allowed to share the CRB of the adjacent cell only if the interference caused to the base station in the adjacent cell does not exceed a given maximum value.

[0034]

[0035] in, is the channel gain between the qth BS and VR device m associated with CRB n, is the maximum allowed interference of the qth BS associated with CRB n;

[0036] (5) Constraint C4: The total latency of VR device m is expressed as the MEC server computing latency The downlink transmission delay t n tr The sum of the required Before completion,

[0037]

[0038] (6) Constraint C5: The total energy consumption of VR device m is calculated as the receiving energy consumption, which cannot exceed the maximum energy consumption that the VR device can bear.

[0039]

[0040] in, It is to accept energy consumption.

[0041] The step 4 is specifically as follows:

[0042] (1) Introducing auxiliary variables Convert the C2 constraint of problem P1 into the following three equivalent constraints:

[0043]

[0044]

[0045]

[0046] (2) Introducing auxiliary variables Convert the C4 constraint of problem P1 into the following two equivalent constraints:

[0047]

[0048]

[0049] In step 6, the convex approximation process is as follows: in the solution set of the previous iteration superior:

[0050] (1)P1 objective function C off Convex approximation of: C off Non-convex term The convex approximation of

[0051]

[0052] in So, in the kth off In the +1 iteration, the P1 objective function C off The convex approximation of

[0053]

[0054] (2) The three equivalent constraints in step 4 (1) are in Ψ off (k off ) are convex approximations on

[0055]

[0056]

[0057]

[0058] (3) The second constraint after equivalent (2) in step 4 is in Ψ off (k off ) is a convex approximation to

[0059]

[0060] (4) According to step 6 (1), (2), (3), generate problem P1 in the kth off In +1 iteration, we get Convex problem P2 on .

[0061] In step 10, the objective function and constraints of the local communication and computing resource allocation problem P3 are as follows:

[0062] (1) Objective function: local cost C loc Calculated as the transmission cost of MEC server for VR devices and CRB spectrum occupation costs

[0063]

[0064] (2) Constraint C6: Indicates the downlink transmission time using CRB n. In order to ensure the transmission of data required by VR device m,

[0065]

[0066] (3) Constraint C7: VR devices are allowed to share the CRB of adjacent cells only if the interference to the BS in the adjacent cells does not exceed a given maximum value.

[0067]

[0068] (4) Constraint C8: The total latency of VR device m is expressed as the local computing latency Downlink transmission delay The sum of the required Before completion,

[0069]

[0070] in, It is the local computing frequency of the VR device;

[0071] (5) Constraint C9: The total energy consumption of VR device m is calculated as the receiving energy consumption and local computing energy consumption The sum of the energy consumption cannot exceed the maximum energy consumption that the VR device can bear.

[0072]

[0073] where κ LOC >0 is the effective capacitance coefficient of VR equipment.

[0074] In step 11, the constraint equivalent transformation of the local communication and computing resource allocation problem P3 is specifically as follows:

[0075] (1) Introducing auxiliary variables Transform constraint C6 of problem P3 into the following three equivalent constraints:

[0076]

[0077]

[0078]

[0079] (2) Introducing auxiliary variables Transform constraint C8 of problem P3 into the following two equivalent constraints:

[0080]

[0081]

[0082] In step 13, the convex approximation process is as follows: in the solution set of the previous iteration superior:

[0083] (1) P3 objective function C loc The convex approximation of

[0084]

[0085] (2) The three equivalent constraints in step 11(1) are in Ψ loc (k loc ) are convex approximations on

[0086]

[0087]

[0088]

[0089] (3) The second constraint after the equivalent in step 11(2) is in Ψ loc (k loc ) is a convex approximation to

[0090]

[0091] (4) According to (1)(2)(3) of step 13, generate problem P3 in the kth loc In the +1 iteration, Convex problem P4 on .

[0092] In step 7 and step 14, the convex problem P2 and the convex problem P4 are solved by using the Matlab CVX toolbox.

[0093] After adopting the above solution, the uninstall decision α of the VR device can be m and cellular resource block selection β m,n When the downlink transmission power p is known, the continuous convex approximation algorithm is used to jointly optimize the downlink transmission power p m , downlink transmission time MEC server calculation frequency Calculate frequency with VR devices The present invention can minimize the total CCHN computing offloading cost of VR devices in the NOMA-MEC system by reasonably allocating communication and computing resources under the constraints of VR device latency and energy consumption, and has low complexity. Compared with the ECRA algorithm, the FTPA algorithm, the No Time Optimization algorithm (NoTO) and the RFS algorithm, the present invention can significantly save system costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 Convergence curve of relative error versus maximum number of iterations Γ obtained by comparing the present invention (SCA) with the exhaustive search (OES) algorithm in the problem of offloading communication and computing resource allocation;

[0095] Figure 2 Convergence curve of relative error versus maximum number of iterations Γ obtained by comparing the present invention (SCA) with the exhaustive search (OES) algorithm on the local communication and computing resource allocation problem;

[0096] Figure 3 To compare the performance of the proposed method (SCA) with the exhaustive search (OES) algorithm on 30 random examples for the problem of offloading communication and computational resource allocation;

[0097] Figure 4 To compare the performance of the present invention (SCA) with the exhaustive search (OES) algorithm on 30 random examples for the local communication and computing resource allocation problem;

[0098] Figure 5To compare the performance of the proposed method (SCA) with ECRA, FTPA, NoTO and RFS algorithms under different numbers of VR devices in the problem of offloading communication and computing resource allocation;

[0099] Figure 6 To compare the performance of the present invention (SCA) with ECRA, FTPA, NoTO and RFS algorithms under different numbers of VR devices in terms of local communication and computing resource allocation;

[0100] Figure 7 To compare the performance of the present invention (SCA) with ECRA, FTPA, NoTO and RFS algorithms under different input data sizes in the problem of offloading communication and computing resource allocation;

[0101] Figure 8 To compare the performance of the present invention (SCA) with ECRA, FTPA, NoTO and RFS algorithms under different input data sizes in the local communication and computing resource allocation problem;

[0102] Fig. 9 To compare the performance of the present invention (SCA) with ECRA, FTPA, NoTO and RFS algorithms under different maximum allowed delays in the problem of offloading communication and computing resource allocation;

[0103] Fig.10 In order to compare the performance of the present invention (SCA) with ECRA, FTPA, NoTO and RFS algorithms under different maximum allowed delays in the local communication and computing resource allocation problem. DETAILED DESCRIPTION

[0104] The present invention discloses a method for optimizing the total computing unloading cost of a VR system based on NOMA-MEC in CCHN. The application scenario is a downlink, which includes a CR router, a group of VR devices and a set of available cellular resource blocks (CRBs) The CR router is equipped with a computing module as a MEC server close to the VR device. Each VR device can only use one CRB, and VR devices using the same CRB are divided into the same NOMA group, and then share a downlink CRB to transmit data from the MEC server to the VR device end. The VR device decodes the data from the MEC server by applying continuous interference cancellation technology.

[0105] For ease of processing, a quasi-static network is considered, that is, the VR device remains stationary during the entire process of task calculation and transmission. Data transmission from the MEC server to the VR device is completed by sharing the spectrum of adjacent cellular units. In order to minimize mutual interference between cellular users (CUs), adjacent cellular units occupy different spectrums, and the spectrum is further divided into CRBs. Considering that the data transmission process from the MEC server to the VR device will cause additional interference to CUs, the spectrum of CCHN needs to be purchased from other service providers, and the task calculation and downlink transmission process on the MEC server will bring certain energy consumption. The total computing offloading cost of the system mainly includes three parts:

[0106] ① The MEC server performs computing tasks for VR devices, which consumes electricity and causes MEC server computing costs.

[0107]

[0108] ②The MEC server allocates downlink transmission power to VR devices to transmit data, resulting in transmission energy consumption and MEC server transmission costs.

[0109]

[0110] ③ The CRB spectrum occupation cost that VR devices need to pay for sharing CRB,

[0111]

[0112] Therefore, the computational offloading cost of the entire system can be calculated as,

[0113]

[0114] Among them, α m and β m,n They represent the offloading decision and CRB selection of VR devices respectively. When α m =1 means that the computing task is processed on the MEC server. m =0 means the computing task is processed on the VR device, β m,n =1 indicates that VR device m uses CRBn to receive data. MEC >0 is the fee that the MEC server needs to pay for each unit of energy consumed. The fee that VR devices need to pay for each unit time CRB n is occupied, κ MEC >0 is the effective capacitance coefficient of the MEC server. m is the 2D FOVs data size of the task, ρ m The number of CPU cycles required to process one bit. is the computing frequency assigned to VR device m by the MEC server, p m is the downlink transmission power allocated to VR device m, is the downlink transmission time using CRB n.

[0115] When VR device uninstalls decision α m and CRB selects β m,n When given, the uninstalled VR device set, local VR device set, uninstalled CRB set, and local CRB set in the system can be expressed as and According to the NOMA grouping on different CRBs, the offloaded VR device set and the local VR device set are further represented as and Now, the problem of minimizing the total CCHN computing offloading cost of VR devices in the NOMA-MEC system can be divided into two problems for discussion: the offloading communication and computing resource allocation problem and the local communication and computing resource allocation problem.

[0116] Problem 1: Offloading communication and computing resource allocation problem: When α m =1, the tasks of the VR device are calculated on the MEC server. Offloading cost C off Calculate the MEC server computing cost for VR devices MEC server transmission cost and CRB spectrum occupation costs

[0117]

[0118] The offloading communication and computation resource allocation problem P1 can be described as

[0119]

[0120] Among them, constraint C1 means that the sum of the computing frequencies allocated by the MEC server to the VR device cannot exceed the maximum computing frequency of the MEC server Constraint C2 indicates that the downlink transmission time The 3D FOVs data O required by VR device m needs to be guaranteed m The transmission of R m is the achievable data rate of VR device m, calculated as

[0121]

[0122] h m and They represent the channel coefficient between the MEC server and VR device m and the maximum interference power and noise level of VR device m on CRB n respectively. Constraint C3 indicates that VR device m is allowed to share CRB n of the adjacent cell only if Q n The interference caused by each BS does not exceed a given maximum value. Bundle C4 represents the total delay of VR device m, that is, the sum of MEC server calculation delay and downlink transmission delay does not exceed the given deadline Constraint C5 represents the total energy consumption of VR device m, that is, the received energy consumption does not exceed the energy consumption that the VR device can bear

[0123] Problem 2: Local communication and computing resource allocation problem: Local communication and computing resource allocation problem: When α m = 0, the task of the VR device is calculated locally on the VR device. Local cost C loc Calculated as the transmission cost of MEC server for VR devices and CRB spectrum occupation costs

[0124]

[0125] The local communication and computation resource allocation problem P3 can be described as

[0126]

[0127] Among them, constraint C6 means that during the downlink transmission time The 2DFOVs data required by VR device m needs to be guaranteed m The transmission of R m is the achievable data rate of VR device m. Constraint C7 indicates that VR device m is allowed to share the CRB n of the adjacent cell only if the Q n The interference caused by each BS does not exceed a given maximum value. Constraint C8 represents the total delay of VR device m, that is, the sum of the local computing delay and the downlink transmission delay of the VR device does not exceed the given deadline Constraint C9 represents the total energy consumption of VR device m, that is, the sum of the receiving energy consumption and the local computing energy consumption does not exceed the energy consumption that the VR device can bear. where κ LOC >0 is the effective capacitance coefficient of VR equipment.

[0128] Obviously, the objective function, constraints C2 and C4 of problem P1, and the objective function, constraints C6 and C8 of problem P3 are non-convex. Therefore, problems P1 and P3 are non-convex problems and are usually difficult to solve. In order to make problems P1 and P3 easier to handle, we perform an equivalent transformation on the non-convex constraints. Then, using the SCA algorithm, the non-convex functions and constraints of problems P1 and P3 are transformed into convex functions and constraints on the given point, i.e., the solution set of the previous iteration, and then transformed into a convex problem solution. Finally, the optimization of the downlink transmission power p is achieved. m , downlink transmission time MEC server calculation frequency Calculate frequency with VR devices

[0129] The communication and computing resource joint allocation method of the present invention specifically comprises the following steps:

[0130] Step 1: Place a group of VR devices in the network and a set of available Uninstallation decision α for a given VR device m and CRB selects β mn ; Given the initial step size θ(0) and the step size update factor δ, the initial iteration index k off =1 and k loc =1, maximum number of iterations Γ.

[0131] Step 2: Generate and uninstall VR device set Local VR device set Uninstall CRB Set Local CRB Set The uninstalled VR device set and the local VR device set are further represented as and

[0132] Step 3: Uninstall the VR device set and uninstall CRB set Next, we obtain the offloading communication and computing resource allocation problem P1.

[0133] Step 4: Offload the communication and computational resource allocation problem P1 by equivalent constraint transformation:

[0134] (1) Introducing auxiliary variables Convert the C2 constraint of problem P1 into the following three equivalent constraints:

[0135]

[0136]

[0137]

[0138] (2) Introducing auxiliary variables The C4 constraint of problem P1 is transformed into the following two equivalent constraints:

[0139]

[0140]

[0141] Step 5: Randomly generate an initial feasible solution set that satisfies the constraints of the offloading communication and computing resource allocation problem P1, namely

[0142] Step 6: Convex approximation of offloading communication and computing resource allocation problem P1: The solution set of the previous iteration superior:

[0143] (1)P1 objective function C off Convex approximation of: C off Non-convex term The convex approximation of

[0144]

[0145] in So, in the kth off In the +1 iteration, the P1 objective function C off The convex approximation of

[0146]

[0147] (2) The three equivalent constraints in step 4 (1) are in Ψ off (k off ) are convex approximations on

[0148]

[0149]

[0150]

[0151] (3) The second constraint after equivalent (2) in step 4 is in Ψ off (k off ) is a convex approximation to

[0152]

[0153] (4) According to step 6 (1), (2), (3), generate problem P1 in the kth off In +1 iteration, we get Convex problem P2 on .

[0154] Step 7: Solve the convex problem P2 using the Matlab CVX toolbox to obtain the solution And update the optimal solution

[0155] Step 8: Update the feasible solution set of problem P2 constraints Where θ(k off ) is the step size, and its update method is θ(k off )=θ(k off -1)(1-δθ(k off -1)).

[0156] Step 9, k off =k off +1, number of iterations k off >Γ, go to step 10; otherwise, go to step 6.

[0157] Step 10: Set up local VR device and local CRB set Next, we obtain the local communication and computing resource allocation problem P3.

[0158] Step 11: Constraint equivalence transformation of local communication and computing resource allocation problem P3:

[0159] (1) Introducing auxiliary variables Transform constraint C6 of problem P3 into the following three equivalent constraints:

[0160]

[0161]

[0162]

[0163] (2) Introducing auxiliary variables Transform constraint C8 of problem P3 into the following two equivalent constraints:

[0164]

[0165]

[0166] Step 12: Randomly generate an initial feasible solution set that satisfies the local communication and computing resource allocation problem P3 constraints, namely

[0167] Step 13: Convex approximation of the local communication and computing resource allocation problem P3: In the previous iteration solution set superior:

[0168] (1) P3 objective function C loc The convex approximation of

[0169]

[0170] (2) The three equivalent constraints in step 11(1) are in Ψ loc (k loc ) are convex approximations on

[0171]

[0172]

[0173]

[0174] (3) The second constraint after the equivalent in step 11(2) is in Ψ loc (k loc ) is a convex approximation to

[0175]

[0176] (4) According to steps 13(1)(2)(3), generate problem P3 in the kth loc In the +1 iteration, Convex problem P4 on .

[0177] Step 14: Solve the convex problem P4 using the Matlab CVX toolbox to obtain the solution And update the optimal solution

[0178] Step 15: Update the feasible solution set of problem P4 constraints Where θ(k loc ) is the step size, and its update method is θ(k loc )=θ(k loc -1)(1-δθ(k loc -1)).

[0179] Step 16, k loc =k loc +1, number of iterations k loc >Γ, go to step 17; otherwise, go to step 13.

[0180] Step 17: Output the uninstall decision α on a given VR device m and CRB selects β m,n Solution set of offloading communication and computing resource allocation problem under and local communication and computing resource allocation problem solutions Corresponding unloading cost and local costs And system cost

[0181] In order to evaluate the performance of the present invention, the following simulation is performed. The service radius of the cellular unit centered on the MEC service is set to 500m, and the VR devices are randomly placed within this range. In the 6 cells adjacent to the cellular unit under consideration, the base station is located at the center of the cell. The channel gain is calculated according to the logarithmic distance path loss model, and the path loss exponent is 4. The simulation parameters are set as follows: the maximum transmission power of the VR device is 23dBm, the number of CRBs is 20, the channel bandwidth B is 500MHz, the 2D FOVs data size range is [1,6]Mbits, the 3D FOVs data size range is [2-12]Mbits, and the calculation load ρ m The maximum allowable delay of VR equipment is 20ms, the noise power spectrum density is -174dBm / Hz, and the VR equipment receives power 0.01W, the maximum computing power of the MEC server At 1000 GHz, the effective capacitance switching coefficient κ of MEC server and VR device is MEC and κ LOC 10 respectively -26 and 10 -27 , the maximum allowable energy consumption of VR devices The cost of energy consumed by the MEC server is η per joule. MEC For 10 -5 , the unit time cost of CRB n The value range is [10 -4 ,10 -2 ].

[0182] Figure 1 and Figure 2 The trend of the relative error of the system cost value obtained by the SCA algorithm and the optimal exhaustive (OES) algorithm with the number of iterations for the offload communication and computing resource allocation problem and the local communication and computing resource allocation problem under different numbers of VR devices is described respectively. It can be seen that the cost obtained by the proposed SCA algorithm decreases rapidly with the increase of the number of iterations. When the maximum number of iterations is 5, the relative error can basically be controlled below 5%, indicating that the convergence speed of the SCA algorithm is fast. In the simulation of the present invention, the maximum number of iterations of the SCA algorithm is set to Γ=5.

[0183] Since the time and space complexity of the OES algorithm increases exponentially with the number of VR devices, we only compare the SCA algorithm and the OES algorithm in a small network scenario. Figure 3 and Figure 4As shown in Figure 1, we compared the SCA algorithm and the OES algorithm in 30 random examples, where Case 1-Case 10, Case 11-Case 20, and Case 21-Case 30 were simulated and run in the scenarios where the number of VR devices M = 3, 4, and 5, respectively. It can be seen that the running time of the OES algorithm increases rapidly with the increase in the number of VR devices, but the SCA algorithm can achieve performance similar to that of the OES algorithm in a very short time.

[0184] In order to measure the performance of the SCA algorithm in a large network, we proposed four benchmark schemes for comparison, namely, the average computational resource allocation algorithm (ECRA), the fractional transmit power allocation algorithm (FTPA), the no time optimization algorithm (NoTO), and the random optimization algorithm (RFS). The default value of the number of VR devices is set to 20, the computing CPU frequency of the MEC server and VR devices in the ECRA algorithm is fixed to 3GHz, the downlink transmission power in the FTPA algorithm is allocated according to the channel gain sorting, and the downlink transmission time in the NoTO algorithm is fixed to 4ms.

[0185] like Figure 5 , Figure 6 and Figure 7 , Figure 8 As shown in Figure 1, the system cost increases with the number of VR devices and the amount of input data. This is because when the number of VR devices involved in transmission and calculation or the amount of input data increases, the system needs to provide more communication and computing resources to meet the transmission and computing requirements of all VR devices, so the total cost of the system increases accordingly. Fig. 9 , Fig.10 As shown in the figure, the cost of the system decreases with the increase of the maximum allowable delay of the VR device. Because when the delay requirement of the task is more stringent, the system needs to pay more communication and computing resources to ensure that the transmission and calculation are completed within the maximum allowable delay of the task, so the system cost increases. In addition, compared with the RFS, NoTO, ECRA and FTPA algorithms, the proposed SCA algorithm reduces the system cost by 83.04%, 58.47%, 29.79% and 12.52% on average.

[0186] The above description is merely an embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any minor modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A joint allocation method of communication and computing resources with minimization of VR system offloading cost based on NOMA-MEC in CCHN, characterized by: The method comprises the following steps: Step 1: Place a group of VR devices in the network and a set of available Uninstallation decision α for a given VR device m and CRB selects β m,n ; Given the initial step size θ(0) and the step size update factor δ, the initial iteration index k off =1 and k loc =1, maximum number of iterations Γ; Step 2: Generate and uninstall VR device set Local VR device set Uninstall CRB Set Local CRB Set The uninstalled VR device set and the local VR device set are further represented as and Step 3: Uninstall the VR device set and uninstall CRB set Next, the objective function and constraints of the offloading communication and computing resource allocation problem P1 are obtained; Step 4: Perform constraint equivalence transformation of offloading communication and computing resource allocation problem P1; Step 5: Randomly generate an initial feasible solution set that satisfies the constraints of the offloading communication and computing resource allocation problem P1, namely Step 6: offload the convex approximation of the communication and computing resource allocation problem P1 to obtain the convex problem P2; Step 7: Solve the convex problem P2 and get the solution And update the optimal solution Step 8: Update the feasible solution set of problem P2 constraints Where θ(k off ) is the step size, and its update method is θ(k off )=θ(k off -1)(1-δθ(k off -1); Step 9, k off =k off +1, number of iterations k off >Γ, go to step 10, otherwise go to step 6; Step 10: Set up local VR device and local CRB set Next, we obtain the objective function and constraints of the local communication and computing resource allocation problem P3; Step 11: perform constraint equivalence transformation of the local communication and computing resource allocation problem P3; Step 12: Randomly generate an initial feasible solution set that satisfies the local communication and computing resource allocation problem P3 constraints, namely Step 13: Convex approximation of the local communication and computing resource allocation problem P3 to obtain the convex problem P4; Step 14: Solve the convex problem P4 and obtain the solution And update the optimal solution Step 15: Update the feasible solution set of problem P4 constraints Where θ(k loc ) is the step size, and its update method is θ(k loc )=θ(k loc -1)(1-δθ(k loc -1); Step 16, k loc =k loc +1, number of iterations k loc >Γ, go to step 17, otherwise go to step 13; Step 17: Output the uninstall decision α on a given VR device m and CRB selects β m,n Solution set of offloading communication and computing resource allocation problem under and local communication and computing resource allocation problem solutions Corresponding unloading cost and local costs And system cost In step 3, the objective function and constraints of the offloading communication and computing resource allocation problem P1 are as follows: (1) Objective function: Unloading cost C off Calculate the MEC server computing cost for VR devices MEC server transmission cost and CRB spectrum occupation costs Among them, η MEC >0 is the fee that the MEC server needs to pay for each unit of energy consumed. The cost of VR equipment occupying CRBn per unit time, κ MEC >0 is the effective capacitance coefficient of the MEC server; I m is the 2DFOVs data size of the task, ρ m The number of CPU cycles required to process one bit; is the computing frequency assigned to VR device m by the MEC server, p m is the downlink transmission power allocated to VR device m, is the downlink transmission time using CRB n; (2) Constraint C1: The computing resources of the MEC server are limited, and the sum of the computing frequencies allocated to VR devices cannot exceed the maximum computing frequency of the MEC. (3) Constraint C2: Indicates the downlink transmission time using CRBn. In order to ensure the transmission of data required by VR device m, there is Among them, O m is the data size of 3DFOVs, R m is the achievable data rate of VR device m; (4) Constraint C3: The VR device is allowed to share the CRB of the adjacent cell only if the interference caused to the base station in the adjacent cell does not exceed a given maximum value. in, is the channel gain between the qth BS and VR device m associated with CRBn, is the maximum allowed interference of the qth BS associated with CRBn; (5) Constraint C4: The total latency of VR device m is expressed as the MEC server computing latency Downlink transmission delay The sum of the required Before completion, (6) Constraint C5: The total energy consumption of VR device m is calculated as the receiving energy consumption, which cannot exceed the maximum energy consumption that the VR device can bear. in, is to accept energy consumption; The step 4 is specifically as follows: (1) Introducing auxiliary variables Convert the C2 constraint of problem P1 into the following three equivalent constraints: (2) Introducing auxiliary variables Convert the C4 constraint of problem P1 into the following two equivalent constraints: In step 6, the convex approximation process is as follows: in the solution set of the previous iteration superior: (1)P1 objective function C off Convex approximation of: C off Non-convex term The convex approximation of in So, in the kth off In the +1 iteration, the P1 objective function C off The convex approximation of (2) The three equivalent constraints in step 4 (1) are in Ψ off (k off ) are convex approximations on (3) The second constraint after equivalent (2) in step 4 is in Ψ off (k off ) is a convex approximation to (4) According to step 6 (1) (2) (3), generate problem P1 in the kth off In +1 iteration, we get Convex problem P2 on ; In step 10, the objective function and constraints of the local communication and computing resource allocation problem P3 are as follows: (1) Objective function: local cost C loc Calculated as the transmission cost of MEC server for VR devices and CRB spectrum occupation costs (2) Constraint C6: Indicates the downlink transmission time using CRBn. In order to ensure the transmission of data required by VR device m, there is (3) Constraint C7: VR devices are allowed to share the CRB of adjacent cells only if the interference to the BS in the adjacent cells does not exceed a given maximum value. (4) Constraint C8: The total latency of VR device m is expressed as the local computation latency Downlink transmission delay The sum of the required Before completion, in, It is the local computing frequency of the VR device; (5) Constraint C9: The total energy consumption of VR device m is calculated as the receiving energy consumption and local computing energy consumption The sum of the energy consumption cannot exceed the maximum energy consumption that the VR device can bear. where κ LOC >0 is the effective capacitance coefficient of VR equipment; The step 11 is specifically as follows: (1) Introducing auxiliary variables Transform constraint C6 of problem P3 into the following three equivalent constraints: (2) Introducing auxiliary variables Transform constraint C8 of problem P3 into the following two equivalent constraints: In step 13, the convex approximation process is as follows: in the solution set of the previous iteration superior: (1) P3 objective function C loc The convex approximation of (2) The three equivalent constraints in step 11(1) are in Ψ loc (k loc ) are convex approximations on (3) The second constraint after the equivalent in step 11(2) is in Ψ loc (k loc ) is a convex approximation to (4) According to (1)(2)(3) of step 13, generate problem P3 in the kth loc In +1 iteration, we get Convex problem P4 on .

2. A method for jointly allocating communication and computing resources with minimized VR system offloading cost based on NOMA-MEC in CCHN according to claim 1, characterized in that: In step 7 and step 14, the convex problem P2 and the convex problem P4 are solved by using the MatlabCVX toolbox.

Images