Ris cooperation-based task offloading scheduling method and mobile edge computing system

Abstract

The present application relates to a kind of based on RIS cooperation's task unloading scheduling method and mobile edge computing system.The method described in the present application, including: obtaining the cascade channel information and noise power between user and access point, and combining a optimal energy consumption objective function to carry out task unloading initialization, generate initial task unloading allocation iteration node and optimal task unloading scheme;Using lower bound branch algorithm based on user allocation to the current iteration node is branched, and the optimal task unloading scheme is updated, and according to the optimal task unloading scheme updated, the current iteration node is pruned;Then, judge the iteration node of pruning whether to meet iteration condition: if yes, then according to optimal task unloading scheme to user computing task is unloaded allocation;If not, then continue iteration.The RIS cooperation's task unloading scheduling method described in the present application has the advantage that the user's endurance time is greatly prolonged.

Description

Technical Field

[0001] This invention relates to the field of resource scheduling in mobile edge computing, and in particular to a task offloading scheduling method, apparatus and mobile edge computing system based on RIS collaborative scheduling. Background Technology

[0002] With the rapid development of mobile internet and Internet of Things (IoT) devices, applications such as high-definition video, semantic recognition, virtual reality, and smart homes are increasingly demanding real-time performance, high throughput, and low power consumption. However, user terminals have limited battery power, computing resources, and processing speed, which can easily lead to problems such as network latency, bandwidth bottlenecks, and concentrated computing loads, making it difficult to meet the needs of these applications. Therefore, Mobile Edge Computing (MEC) is now being adopted, deploying computing servers at the network edge and connecting them through access points (APs) set up by base stations. This allows user devices to offload some or all of their computing tasks to the MEC via the base station's access points. After completing the computing tasks, the MEC sends the results back to the user device, enabling the user device to offload computing tasks at a closer physical distance, thereby reducing transmission latency, reducing core network load, and improving computing efficiency.

[0003] However, MEC is susceptible to environmental factors such as building obstruction, severe weather, and wireless signal fading, leading to a decrease in the quality of the MEC's ​​wireless link. This, in turn, causes a decrease in data transmission speed, task offloading failure, or increased system power consumption. Therefore, existing technologies employ Reconfigurable Intelligent Surfaces (RIS) to optimize the wireless communication quality of MEC. The RIS consists of a large number of tunable reflective elements, each of which dynamically adjusts the reflection phase through passive beamforming to optimize the propagation path of the wireless signal. In particular, by intelligently controlling the wireless channel at the physical layer, the RIS can enhance signal strength, reduce multipath fading, and reduce interference, significantly improving the communication performance of the MEC system. For details on the connection between user equipment and base stations or access points (APs) via the RIS, please refer to [reference needed]. Figure 1 , Figure 1 A simplified diagram illustrating the use of RIS signal transmission for user equipment.

[0004] Among them, RIS is often used to assist wireless communication between user equipment and access point (AP), especially in the case of non-line-of-sight (NLoS) channels. By dynamically adjusting the reflection phase to optimize channel gain, it provides additional propagation paths and significantly improves the reliability and transmission rate of the system.

[0005] However, due to the limited computing resources of MEC access points (APs) and the limited number of user equipment that can be connected to each AP, existing task allocation strategies are based solely on static channel conditions or task priorities. This leads to low-computation tasks potentially over-consuming AP resources, while high-computation tasks are forced to execute locally due to insufficient resources, resulting in a significant increase in user equipment power consumption and severely reducing user equipment endurance. Therefore, existing technologies suffer from insufficient coordination between task allocation, resource scheduling, and RIS control in MEC systems, leading to high user equipment power consumption and low MEC resource utilization. Summary of the Invention

[0006] Therefore, the purpose of this invention is to provide a user task offloading and scheduling method based on RIS collaboration.

[0007] A user task offloading and scheduling method based on RIS collaboration includes the following steps:

[0008] S1: Obtain the cascaded channel information and noise power between the user and the access point;

[0009] S2: Perform task offloading initialization based on cascaded channel information, noise power, and an optimal energy consumption objective function, generating an initial task offloading allocation iteration node and an optimal task offloading scheme; wherein, the initial task offloading allocation iteration node includes an initialized user allocation matrix, an initialized derivative scheme state, and an initialized minimum energy consumption estimate.

[0010] S3: Use a user-assigned lower bound branching algorithm to branch the current task unloading assignment iteration node and update the optimal task unloading scheme to obtain the updated task unloading assignment iteration node and the optimal task unloading scheme.

[0011] S4: Based on the updated target energy consumption value of the optimal task unloading scheme, prune the task unloading allocation iteration nodes to obtain refined task unloading allocation iteration nodes.

[0012] S5: Determine whether the refined task unloading and allocation iteration node satisfies the one-branch bounding iteration condition: if yes, then unload and allocate the user's computing tasks according to the current optimal task unloading scheme to complete the unloading and scheduling of user tasks; if no, then continue to execute step S3.

[0013] The RIS-based user task offloading scheduling method described in this invention, compared to existing technologies, achieves joint optimization of the user allocation matrix and RIS phase adjustment coefficients through multi-level optimization iteration of task offloading allocation iteration nodes, combined with the lower bound branch algorithm and the RIS phase optimization algorithm. This effectively solves the problems of uneven task allocation, insufficient resource scheduling, and low RIS control coordination rate in traditional MEC systems. Simultaneously, it significantly reduces user device energy consumption, improves the utilization rate of MEC system computing resources, and enhances the task offloading efficiency of user devices. Therefore, under limited computing resources, this invention, through multi-level optimization iteration, approximates the global optimal solution of the RIS phase adjustment coefficients and user allocation, enabling RIS-assisted MEC systems to maintain high efficiency and robustness in dynamically changing network environments.

[0014] Furthermore, the optimal energy consumption objective function E S (X OPT ,θ OPT The specific representation of ) is as follows:

[0015]

[0016] In the formula, v OPT X represents the target energy consumption value corresponding to the optimal task unloading scheme; OPT This represents the current optimal user allocation matrix, where x is an element. ij This indicates that the i-th user selects the j-th access point; θ OPT This represents the phase adjustment coefficient vector of the current optimal RIS coordinateable unit;

[0017] For locally computed energy consumption models, when x i0 =1 indicates that the i-th user chooses to perform calculations locally; k0 is the energy consumption correlation coefficient; α i To calculate the complexity correlation coefficient; f i The local calculation frequency for the i-th user; len i M represents the task length for the i-th user; M represents the total number of users.

[0018] For the remote computing model, β i P is the uplink cost factor. T,i R represents the transmit power of the device for the i-th user; N represents the total number of access points; R ij (θ OPT ) indicates that the i-th user and the j-th access point are at θ OPT The actual data transmission rate is calculated as follows:

[0019]

[0020] In the formula, B W,j This represents the bandwidth allocated to the current user by the j-th access point; Let σi represent the channel gain vector between the i-th user and the j-th access point, and transpose it using Hermite; σ0 represents the noise power.

[0021] The optimal energy consumption objective function includes several constraints, including user selection constraints, maximum access point allocation constraints, user allocation element constraints, and RIS phase constraints.

[0022] The specific representation of the user-selected constraints is as follows:

[0023]

[0024] The maximum allocation constraint for access points is specifically expressed as follows:

[0025]

[0026] In the formula, L represents the total number of user connections at the j-th access point; nk (j) represents the maximum number of connections for the j-th access point;

[0027] The specific constraints on the user-assigned elements are as follows:

[0028] x ij ∈{0,1},

[0029] The RIS phase constraint condition is specifically expressed as follows:

[0030] |θ l |=1,

[0031] In the formula, θ l This represents the phase adjustment coefficient, i.e., the phase angle, of the l-th coordinable unit of the RIS; |θ l |=1 represents the phase coefficient as a unit modulus complex number; L represents the total number of coordinateable units of the RIS, i.e., the total number of reflective units.

[0032] This invention simplifies the remote computing model for the user's task unloading process by ignoring the energy consumption of data backhaul and MEC server computing during the task unloading process. This is because the amount of data backhauled is usually small, and its energy consumption is negligible compared to the energy consumption of task unloading and computing, so it can be ignored. Although the MEC server's computing energy consumption exists, since the access point (AP) is usually directly connected to the power supply, the AP's energy consumption is provided by an external power supply and is not part of the user's system energy consumption, so it can also be ignored.

[0033] Meanwhile, this invention is based on joint optimization of the objective function, that is, minimizing the target energy consumption of the system by adjusting the user allocation matrix and the RIS phase adjustment coefficient, so as to reduce the energy consumption of user equipment and improve the battery life of user equipment. In the joint optimization process, the focus is on reducing signal attenuation, interference and multipath effects by adjusting the phase of the RIS reflection unit; while the adjustment of the user allocation matrix focuses on the computational requirements of the task and the computational resources of the access point, so as to avoid low-computation tasks occupying too much access point resources. Accordingly, through the design and joint optimization of the above objective function, the task offloading efficiency and resource utilization of the system can be significantly improved, thereby reducing the computational burden of user equipment and effectively improving the battery life of user equipment.

[0034] Furthermore, the initial minimum energy consumption estimate lb0 is calculated using a minimum energy consumption estimation algorithm, specifically as follows:

[0035]

[0036] In the formula, lb0 represents the initial minimum energy consumption estimate; This represents the current user allocation matrix; Est(·) represents the minimum energy consumption estimation algorithm, the specific steps of which are as follows:

[0037] S201. Based on the size of the user allocation matrix of the current task unloading allocation iteration node, generate a simulated user allocation index array and an initial energy consumption estimate.

[0038] S202. Generate a simulated user selection submatrix based on the current simulated user allocation index array;

[0039] S203. Concatenate the simulated user selection submatrix with the current user allocation matrix and determine whether the maximum allocation constraint of an access point is satisfied: if yes, calculate the energy consumption estimate of the concatenated user allocation matrix according to the user allocation energy consumption model and execute step S204; if no, execute step S205.

[0040] S204. Determine whether the energy consumption estimate of the spliced ​​user allocation matrix is ​​less than the current energy consumption estimate: if yes, update the current energy consumption estimate and execute step S205; if no, directly execute step S205.

[0041] S205. Determine whether the current iteration number meets an energy consumption estimation condition: if yes, then use the current energy consumption estimate as the minimum energy consumption estimate for the current task unloading allocation iteration node; if no, then update the current simulated user allocation index array and execute step S202.

[0042] This invention uses a minimum energy consumption estimation algorithm to calculate the minimum energy consumption estimate corresponding to the iteration node, i.e., the lower bound of the node, thereby ensuring that the RIS-based collaborative user task offloading scheduling method described in this invention can gradually approach the global optimal solution. At the same time, through the minimum energy consumption estimation algorithm, the system can effectively evaluate different task offloading schemes and resource allocations in each iteration, ensuring that updates are made according to the scheme with the minimum energy consumption, and preventing it from getting trapped in a local optimum.

[0043] Furthermore, the generated representation of the simulated user selection submatrix is ​​as follows:

[0044] x ij ∈{0,1}

[0045] In the formula, X c To simulate user selection of submatrices, The matrix represents a size of (M-n0)×(N+1), where n0 is the current user allocation matrix. The number of rows; x ij Represents each element x in the matrix ij The rule for the i-th user to choose to access the j-th access point is as follows:

[0046] j = select i ,

[0047] In the formula, This indicates iterating through all users; select i An index is assigned to the simulated user corresponding to the i-th user, and its specific calculation is as follows:

[0048]

[0049] In the formula, select i Assign the i-th element of the index array to the simulated user; Used to represent traversing all elements in the simulated user-assigned index array;

[0050] The expression for the maximum allocation constraint of the access point is as follows:

[0051]

[0052] In the formula, Used to indicate traversing all access points; The energy consumption estimate based on the concatenated allocation matrix X is represented as follows:

[0053]

[0054] Let this be the upper bound of the data transmission rate between the i-th user and the j-th access point, and its specific calculation is as follows:

[0055]

[0056] The specific implementation of updating the current simulated user allocation index array in step S205 is as follows:

[0057] select1 = select1 + 1

[0058] In the formula, select1 represents the first element in the current simulated user allocation index array.

[0059] This invention effectively expands the explorable space of task offloading schemes by introducing a simulated user selection submatrix, thereby improving the accuracy of the minimum energy consumption estimate calculated in step S204. This ensures that the minimum energy consumption estimation algorithm can more accurately evaluate the energy consumption of task offloading schemes and ensures that each iteration approaches the optimal solution, avoiding premature abandonment of the current branch due to local optima, thus guaranteeing the optimization process of the RIS collaborative user task offloading scheduling method.

[0060] Furthermore, the maximum allocation constraint of access points introduced in step S203 further enhances the limitation that the simulated user selection submatrix is ​​more in line with the actual resource allocation. That is, when excluding unnecessary exploration space, it makes it conform to the actual access situation of the current access point, ensuring the balance of task unloading and preventing unreasonable task unloading decisions.

[0061] Furthermore, the user-assigned lower bound branch algorithm includes the following steps:

[0062] S301. Reset the branch of the user allocation matrix for the task unloading allocation iteration node to obtain the iteration node of the current branch;

[0063] S302. Determine whether the iteration node of the current branch is a leaf node: If not, use the minimum energy consumption estimation algorithm to calculate, update the minimum energy consumption estimation value of the iteration node of the current branch according to the calculation result, and execute step S305; if yes, execute step S303.

[0064] S303. The RIS phase optimization algorithm is used to update the phase of the iteration node of the current branch to obtain the optimal task unloading scheme and the corresponding target energy consumption value of the current branch.

[0065] S304. Determine whether the target energy consumption value of the current branch is less than the target energy consumption value of the current optimal task unloading scheme: if yes, update the current optimal task unloading scheme and execute step S305; if no, directly execute step S305.

[0066] S305. Determine whether the current iteration number of the lower bound branch algorithm based on user allocation is less than an iteration condition: if yes, then execute step S301; if no, then complete the iteration of the lower bound branch algorithm.

[0067] This invention expands the explorable space of the user-assigned lower bound branching algorithm by resetting the branch, ensuring that new branch nodes are generated in each iteration, and provides effective energy efficiency assessment by calculating the minimum energy consumption estimates of the parent node and leaf node, thereby ensuring that each branch scheme moves toward the optimal solution and that the algorithm can update the optimal task unloading scheme in a timely manner.

[0068] Furthermore, this invention uses the RIS phase optimization algorithm to calculate the phase angle of the corresponding RIS reflection unit based on the current user allocation matrix, thereby optimizing the signal transmission path, reducing signal attenuation, interference and multipath effects, further improving the transmission efficiency of the system, and ensuring that the final signal quality reaches the optimal level.

[0069] Furthermore, the branch reset in step S301 is specifically represented as follows:

[0070]

[0071] In the formula, Let represent the user allocation matrix of the (N+1)i′+kth task unloading and allocation iteration node, where i′ represents the iteration number of the current task unloading and allocation iteration node; and k is the iteration number of the lower bound branch algorithm based on user allocation.

[0072] Indicates the i-th min The user allocation matrix of the transposed task unloading allocation iteration node, where i min The index of the iterative node that has the minimum energy consumption estimate;

[0073] The identity matrix represents the transpose of the matrix, with a size of (N+1), where the element at the k-th position is 1 and all other elements are 0.

[0074] The step S302 updates the minimum energy consumption estimate of the iterative node of the current branch based on the calculation results. The specific calculation is as follows:

[0075]

[0076] In the formula, lb (N+1)i′+k This represents the minimum energy consumption estimate of the iteration node (N+1)i′+k of the current branch; Est(·) is the minimum energy consumption estimation algorithm.

[0077] Accordingly, the branch reset described in this invention concatenates the user allocation matrix of the current iteration branch with an identity matrix to ensure that during each branch iteration, the minimum energy consumption estimation algorithm and the RIS phase optimization algorithm jointly drive the lower bound branch algorithm to find the globally optimal energy efficiency solution in a wider solution space, thereby ensuring that the task offloading scheme can be effectively optimized and minimizing the energy consumption of user equipment.

[0078] Furthermore, the specific calculation of the phase update in step S303 is expressed as follows:

[0079]

[0080] In the formula, θ (N+1)i′+k Opt represents the phase adjustment coefficient vector of the RIS reconcilable unit corresponding to the optimal task offloading scheme of the current branch; RIS The specific steps of the RIS phase update algorithm are as follows:

[0081] S30301. Calculate the current phase update parameters based on the current RIS phase adjustment coefficient vector and the corresponding cascaded channel information; wherein, the current phase update parameters include data transmission rate, signal-to-noise ratio gain, and channel gain correction factor;

[0082] The data transmission rate is specifically expressed as follows:

[0083]

[0084] In the formula, R ij (θ (u) ) represents the data transmission rate corresponding to the phase adjustment coefficient vector of the u-th iteration; θ (u) Let θ be the phase adjustment coefficient vector for the u-th iteration. When u = 0, θ (0) Each element has a modulus of 1; where u represents the number of iterations of the RIS phase update algorithm;

[0085] The specific expression for the signal-to-noise ratio gain is as follows:

[0086]

[0087] In the formula, This represents the signal-to-noise ratio gain in the u-th iteration; A matrix representing the cascaded channel information between all users and all access points, transposed using Hermite;

[0088] The channel gain correction factor is specifically expressed as follows:

[0089]

[0090] In the formula, This represents the channel gain correction factor for the u-th iteration; The expected term representing the signal gain; This is the noise power term;

[0091] S30302. Based on the current phase update parameters, use the coupled influence optimization algorithm to perform phase optimization on all elements of the current RIS phase adjustment coefficient vector to obtain the optimized RIS phase adjustment coefficient vector.

[0092] S30303. Determine whether the current phase adjustment coefficient vector of the RIS has converged: if not, continue to execute step S30301; if yes, output the current phase adjustment coefficient vector of the RIS and complete the iteration of the user-assigned lower bound branch algorithm.

[0093] The convergence is determined by the target energy consumption value corresponding to the optimal energy consumption objective function, as specifically expressed below:

[0094]

[0095] In the formula, δ represents the target energy consumption value corresponding to the phase adjustment coefficient vector in the (u+1)th iteration; δ is the convergence threshold value of the RIS phase update algorithm.

[0096] Accordingly, the present invention uses the user allocation matrix after the aforementioned branch reset as the input of the RIS phase update algorithm to find the phase angle of the RIS reflection unit that best matches the current user allocation matrix, thereby ensuring that the calculated phase angle can significantly improve the signal transmission quality corresponding to the current user allocation matrix, and thus reduce the total power consumption of the system.

[0097] Furthermore, the coupling effect optimization algorithm optimizes the q-th element of the current RIS phase adjustment coefficient vector. The phase optimization is expressed as follows:

[0098]

[0099] In the formula, This represents the q-th element of the phase adjustment coefficient vector in the (t+1)-th iteration of the coupling effect optimization algorithm, with an initial value of φ. (0) Let θ be the current phase adjustment coefficient vector of RIS. (u) ; Let be the optimization direction vector of the q-th phase in the t-th iteration of the coupling effect optimization algorithm, specifically represented as follows:

[0100]

[0101] In the formula, Gain effect used to represent the signal-to-noise ratio; This represents the coupling effect term, specifically the coupling effect factor Γ between the p-th and q-th elements of the phase adjustment coefficient vector. pq , its Γ pq Specifically, it is expressed as follows:

[0102]

[0103] In the formula, Represents the square of the channel gain correction factor; Let represent the outer product of the channel gain matrices; where the superscript (*) denotes complex conjugate; the iterative condition of the coupling effect optimization algorithm is:

[0104] Determine ||φ (t+1) -φ (t) || Is it less than a threshold value ε? If so, then adjust the phase adjustment coefficient vector δ of the RIS from the previous iteration. (t) As the phase adjustment coefficient vector of the optimized RIS, i.e., θ (u+1) =φ (t) If not, then complete phase optimization; otherwise, adjust φ. (t+1) As the phase adjustment coefficient vector of RIS in the current coupled influence optimization algorithm, the coupled influence optimization algorithm is used to optimize the phase of all elements; wherein, the threshold value ε∈[0,1].

[0105] In the iterative process of the RIS phase update algorithm, this invention combines a coupling effect optimization algorithm to consider the mutual influence between individual reflection units on the currently adjusted reflection units. That is, the phase change of each reflection unit not only affects the signal transmission of its access point, but also has a chain reaction on other access points and users. This effectively avoids the multipath effect and signal interference that are ignored when optimizing the phase of a single reflection unit, and ensures that the global optimization conforms to the actual physical situation.

[0106] Specifically, this invention employs the channel gain correction factor and the gain effect of signal-to-noise ratio (SNR) as core calculation parameters for phase update, which are used to adjust the phase of each reflection unit in real time to reduce signal attenuation, interference, and multipath effects, thereby significantly improving the signal quality and transmission rate of the system. In addition, the introduction of coupling effect terms further improves the coordination between the current reflection unit and other reflection units when adjusting the current reflection unit, ensuring that the phase adjustment not only optimizes the signal of the current user, but also takes into account the overall performance of the entire system, achieving the optimal balance between system performance and energy efficiency.

[0107] A user task offloading scheduling device based on RIS coordination includes a user cascaded channel information acquisition unit, a task offloading scheme initialization unit, a task offloading scheme branching unit, a task offloading scheme pruning unit, and a task offloading scheme iteration judgment unit.

[0108] The user's cascaded channel information acquisition unit is used to acquire the cascaded channel information and noise power between the user and the access point;

[0109] The task offloading scheme initialization unit is used to perform task offloading initialization based on cascaded channel information, noise power, and an optimal energy consumption objective function, and to generate an initial task offloading allocation iteration node and an optimal task offloading scheme; wherein, the initial task offloading allocation iteration node includes an initialized user allocation matrix, an initialized derivative scheme state, and an initialized minimum energy consumption estimate.

[0110] The task unloading scheme branch unit is used to branch the current task unloading allocation iteration node using a user-assigned lower bound branch algorithm and update the optimal task unloading scheme to obtain the updated task unloading allocation iteration node and the optimal task unloading scheme.

[0111] The task unloading scheme pruning unit is used to prune the task unloading allocation iteration nodes according to the updated target energy consumption value of the optimal task unloading scheme, so as to obtain refined task unloading allocation iteration nodes.

[0112] The task unloading scheme iteration judgment unit is used to determine whether the refined task unloading allocation iteration node satisfies the one-branch delimitation iteration condition: if yes, then the user's computing task is unloaded and allocated according to the current optimal task unloading scheme to complete the unloading scheduling of the user's task; if no, then the task unloading scheme branch unit is called again.

[0113] A mobile edge computing system based on a reconfigurable smart metasurface includes a reconfigurable metasurface, i.e., a RIS, a base station, and a RIS-based task offloading and scheduling device that is communicatively connected to the reconfigurable smart metasurface and the base station.

[0114] The reconfigurable metasurface is provided with multiple tunable units for receiving computing task offloading requests from user equipment and adjusting the reflection phase through the tunable units to reflect and transmit signal information to the base station; wherein, the user equipment is used to generate computing tasks and send the computing tasks and corresponding computing requirements to the reconfigurable metasurface.

[0115] The base station is equipped with an access point to the edge computing server, which is used to calculate the cascaded channel information and noise power between the current user and the access point according to the computing task and the corresponding computing requirements, and input them to the RIS-based collaborative task offloading scheduling device.

[0116] The RIS-based task offloading and scheduling device is used to execute the RIS-based task offloading and scheduling method described above.

[0117] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0118] Figure 1 A simplified diagram illustrating the use of RIS signal transmission by user equipment;

[0119] Figure 2 A simplified schematic diagram of signal transmission in a mobile edge computing system based on a reconfigurable smart metasurface;

[0120] Figure 3 This is a simplified structural diagram of the RIS-based task offloading and scheduling device described in this invention.

[0121] Figure 4 This is a simplified flowchart illustrating the RIS-based task offloading and scheduling method described in this invention.

[0122] Figure 5 This is a schematic diagram comparing the joint RIS phase optimization of the present invention with other phase selection methods;

[0123] Figure 6 This is a simplified pseudocode flowchart of the minimum energy consumption estimation algorithm described in this invention.

[0124] Figure 7 This is a simplified pseudocode flowchart of the RIS phase update algorithm described in this invention.

[0125] Figure 8 This is a simplified pseudocode flowchart of the coupling effect optimization algorithm described in this invention. Detailed Implementation

[0126] To address the insufficient coordination of task offloading, resource scheduling, and RIS control in existing MEC systems, this invention acquires cascaded channel information and noise power between users and access points (APs), and combines this with an energy consumption model and an optimal energy consumption objective function for task offloading initialization to generate an initial task offloading scheme and target energy consumption value. A user-assignment-based lower bound branching algorithm is then used to branch the current task offloading scheme, generating several unassigned task offloading schemes. Next, the target energy consumption value is updated based on these unassigned schemes, and the remaining unassigned schemes are pruned based on the updated target energy consumption value to obtain a refined task offloading scheme. Finally, it is determined whether the refined task offloading scheme satisfies the one-branch bounding iteration condition: if yes, the current user is offloaded and assigned according to the current task offloading scheme; otherwise, the user-assignment-based lower bound branching algorithm is used again to branch the current refined task offloading scheme, continuing the iteration.

[0127] Accordingly, this invention optimizes the task offloading allocation scheme based on the branch and bound algorithm. By dynamically adjusting the task offloading and RIS configuration, it ensures that the system can achieve optimal energy consumption under different network conditions, thereby minimizing the total energy consumption of user task offloading, significantly extending the user's battery life, improving the efficiency of communication resource utilization between the MEC system and RIS control, and significantly enhancing the resource scheduling capability of the RIS-assisted MEC system.

[0128] Based on the above design, this invention proposes a user task offloading and scheduling method based on RIS collaboration, and based on this method, proposes a user task offloading and scheduling device based on RIS collaboration.

[0129] Please see Figure 2 , Figure 2 This is a simplified schematic diagram of signal transmission in a mobile edge computing system based on a reconfigurable smart metasurface.

[0130] A mobile edge computing system based on a reconfigurable intelligent metasurface includes a user equipment 100, a reconfigurable metasurface (RIS) 101, a base station 102, and a task offloading and scheduling device based on RIS collaboration as described in this invention, which is connected to the reconfigurable metasurface 101 and the base station 102.

[0131] The user equipment 100 is used to generate computing tasks, such as data processing, video streaming, IoT sensor data, etc., and send the computing tasks and corresponding computing requirements to the reconfigurable metasurface 101.

[0132] The reconfigurable metasurface 101 is provided with multiple tunable units for receiving computing task offloading requests from user equipment and adjusting the reflection phase through the tunable units to reflect and transmit signal information to the base station 102.

[0133] The base station 102 is provided with an access point to the edge computing server, which is used to calculate the cascaded channel information and noise power between the current user and the access point (AP) according to the computing task and the corresponding computing requirements, and input them into the RIS-based collaborative task offloading scheduling device of the present invention.

[0134] Please also refer to Figure 3 and Figure 4 , Figure 3 This is a simplified structural diagram of the RIS-based task offloading and scheduling device described in this invention. Figure 4 This is a simplified flowchart illustrating the RIS-based task offloading and scheduling method described in this invention.

[0135] The RIS-based task offloading scheduling device includes a user cascaded channel information acquisition unit 1, a task offloading scheme initialization unit 2, a task offloading scheme branching unit 3, a task offloading scheme pruning unit 4, and a task offloading scheme iteration judgment unit 5.

[0136] The user's cascaded channel information acquisition unit 1 is used to perform step S1: acquire the cascaded channel information and noise power between the user and the access point.

[0137] Specifically, the cascaded channel information between the user and the access point is represented as follows:

[0138]

[0139] In the formula, h ij h represents the concatenated channel information between the i-th user and the j-th edge computing access point (AP), where i∈[1,M], h∈[1,N], and M and N represent the total number of users and the total number of access points (APs), respectively; a,il This represents the channel information from the i-th user to the l-th tunable unit of the RIS; h b,lj This represents the channel information from the l-th tunable unit of the RIS to the j-th edge computing access point of the base station; θ l This represents the phase adjustment coefficient, or phase angle, of the l-th coordinable unit of the RIS, which satisfies |θ i |=1, That is, the phase coefficient is a unit modulus complex number, which represents the phase adjustment of the signal by the RIS; L represents the total number of coordinateable units of the RIS, that is, the total number of reflection units;

[0140] For ease of processing and calculation, this invention defines the cascaded channel information as a vector representation, then h ij Further defined as:

[0141]

[0142]

[0143] In the formula, This represents the channel gain vector between the i-th user and the j-th edge computing access point (AP), i.e., the concatenated channel information, and its specific size is... That is, a complex vector of size 1×L, used to represent the sum of the contributions of the L tunable units of the RIS to the signal, and its H superscript indicates the Hermite transpose; This represents the phase adjustment coefficient vector of all RIS coordinateable units, that is, the phase adjustment amount (phase angle) of each reflective unit of the RIS, and its T superscript is used to indicate ordinary transpose; Symbols represent definitions.

[0144] The task offloading scheme initialization unit 2 is used to execute step S2: perform task offloading initialization based on cascaded channel information, noise power, and an optimal energy consumption objective function, and generate initial task offloading allocation iteration nodes and optimal task offloading scheme.

[0145] Specifically, the initial task offloading assignment iteration node includes an initialized user assignment matrix, an initialized derivative scheme state, and an initialized minimum energy consumption estimate.

[0146] The specific representation of the initialized user allocation matrix is ​​as follows:

[0147]

[0148] In the formula, This represents the user allocation matrix for the 0th task unloading and allocation iteration node, with its specific values ​​being an empty set. This indicates initialization; the specific representation of the derived scheme state of the initialization is as follows:

[0149] status0 = 1

[0150] In the formula, status0 represents the derivation scheme status of the 0th task unloading allocation iteration node, with a value of 1 indicating that it is derivable, and a value of 0 indicating that it is not derivable; the initial minimum energy consumption estimate, i.e., the lower bound of the node, is calculated using a minimum energy consumption estimation algorithm, specifically as follows:

[0151]

[0152] In the formula, lb0 represents the initial minimum energy consumption estimate; This represents the current user allocation matrix, whose initial value corresponds to... For cascaded channel information; σ0 is noise power; Est(·) represents the minimum energy consumption estimation algorithm, the specific steps of which are as follows:

[0153] Please also refer to Figure 6 , Figure 6 This is a simplified pseudocode flowchart of the minimum energy consumption estimation algorithm described in this invention.

[0154] S201. Based on the size of the user allocation matrix of the current task unloading allocation iteration node, generate a simulated user allocation index array and an initial energy consumption estimate.

[0155] Specifically, the simulated user allocation index array select is represented as follows:

[0156]

[0157] In the formula, {0} represents an array of all zeros, used to simulate the situation where users have not yet been assigned tasks, where M is the total number of users, and n0 is the number of rows in the current user assignment matrix;

[0158] Wherein, the initial energy consumption estimate t min The default value is 1000. The initial energy consumption estimate is set to a large value to ensure that the subsequent iteration process can be carried out effectively, thereby preventing the loop process from being disturbed by too small a value.

[0159] S202. Generate a simulated user selection submatrix based on the current simulated user allocation index array.

[0160] Specifically, the generated representation of the simulated user selection submatrix is ​​as follows:

[0161] x ij ∈{0,1}

[0162] In the formula, X c To simulate user selection of submatrices, The matrix represents a size of (M-n0)×(N+1), where (M-n0) is the number of rows, (N+1) is the number of columns, and N represents the total number of access points; x ij Represents each element x in the matrix ij It can only take the value 0 or 1, used to indicate whether the i-th user selects the j-th access point. If x ij =1, which means that the i-th user chooses the j-th access point, and the rule for choosing the j-th access point is as follows:

[0163] j = select i ,

[0164] In the formula, i = [1, M] represents all rows of the simulated user selection submatrix, i.e., traversing all users; select i An index is assigned to the simulated user corresponding to the i-th user, and its specific calculation is as follows:

[0165]

[0166] In the formula, select i Assign the i-th element of the index array to the simulated user to represent the access point number selected by the user; This is used to represent traversing all elements in the simulated user-assigned index array.

[0167] S203. Concatenate the simulated user selection submatrix with the current user allocation matrix and determine whether the maximum allocation constraint of an access point is satisfied. If yes, calculate the energy consumption estimate of the concatenated user allocation matrix according to the user allocation energy consumption model and execute step S204. If no, execute step S205.

[0168] Specifically, the concatenated user allocation matrix is ​​represented as follows:

[0169]

[0170] in, Assign a matrix to the current user, the matrix of size M×(N+1); X c The user selection submatrix is ​​of size (M-n0)×(N+1); X is the concatenated user allocation matrix of size M×(N+1).

[0171] The specific judgment rule for the maximum allocation constraint of the access point is expressed as follows:

[0172]

[0173] In the formula, L represents the total number of user connections at the j-th access point; nk (j) represents the maximum number of connections for the j-th access point; Used to indicate traversing all access points.

[0174] When the total number of user connections at all access points meets the maximum allocation constraint for the access points, the energy consumption estimate of the concatenated allocation matrix is ​​calculated, and its specific calculation is expressed as follows:

[0175]

[0176] In the formula, This is the energy consumption estimate based on the concatenated allocation matrix, i.e., the user allocation energy consumption model. Its input is the concatenated allocation matrix, and its output is the energy consumption estimate.

[0177] For locally computed energy consumption models, when x i0 =1 indicates that the i-th user chooses to perform calculations locally; k0 is the energy consumption correlation coefficient; α i To calculate the complexity correlation coefficient; L i The local calculation frequency for the i-th user; len i This represents the task length for the i-th user, in bits.

[0178] For the energy consumption model of remote computing, β i This is the uplink cost coefficient, and its value is β. i ≥1 indicates that there is uplink overhead; P t,i Let be the device transmit power of the i-th user; Let Shannon's formula be the upper bound of the data transmission rate between the i-th user and the j-th access point. Its specific calculation is as follows:

[0179]

[0180] In the formula, B W,j This represents the bandwidth allocated to the current user by the j-th access point; Let σi represent the channel gain vector between the i-th user and the j-th access point; σ0 is the noise power.

[0181] S204. Determine whether the energy consumption estimate of the spliced ​​user allocation matrix is ​​less than the current energy consumption estimate: if yes, update the current energy consumption estimate and proceed to the next step; if no, proceed directly to the next step.

[0182] Specifically, the judgment expression for updating the current energy consumption estimate is as follows:

[0183]

[0184] S205. Determine whether the current iteration number meets an energy consumption estimation condition: if yes, then use the current energy consumption estimate as the minimum energy consumption estimate for the current task unloading allocation iteration node; if no, then update the current simulated user allocation index array and execute step S202.

[0185] Specifically, the specific judgment expression for the energy consumption estimation condition is as follows:

[0186]

[0187] In the formula, count is the number of iterations of the current minimum energy consumption estimation algorithm, and its initial value is 0;

[0188] If the energy consumption estimation condition is met, the current energy consumption estimate is output, and the iteration ends.

[0189] If the energy consumption estimation condition is not met, the current simulated user allocation index array is updated using the following formula, and step S202 is continued for iteration:

[0190] select1 = select1 + 1

[0191] In the formula, select1 represents the first element in the current simulated user allocation index array, which is used to simulate the first user selecting the next access point, driving the update of the simulated user selection submatrix, thereby exploring more simulated user selection submatrixes.

[0192] Simultaneously, a feasible solution to the optimal energy consumption objective function is adopted as the initial optimal task offloading scheme {X}. OPT ,θ OPT}, and calculate the corresponding target energy consumption value, as shown below:

[0193]

[0194] In the formula, v OPT This represents the target energy consumption value corresponding to the optimal task offloading scheme, expressed as an objective function E for optimal energy consumption. S (X OPT ,θ OPT ) is obtained through calculation;

[0195] X OPT This represents the current optimal user allocation matrix, whose initial values ​​can be obtained through random allocation or through a heuristic algorithm (such as a channel quality-based greedy algorithm), and its elements x ij This indicates that the i-th user selects the j-th access point;

[0196] θ OPT This represents the phase adjustment coefficient vector of the current optimal RIS coordinateable unit, whose initial value can be obtained by random assignment or through a heuristic algorithm.

[0197] R ij (θ OPT ) represents the connection between the i-th user and the j-th access point at θ OPT The actual data transmission rate is calculated as follows:

[0198]

[0199] Wherein, the optimal energy consumption objective function E S (X OPT ,θ OPT Several constraints are set; the constraints include user selection constraints, maximum access point allocation constraints, user allocation element constraints, and RIS phase constraints.

[0200] The user selection constraints are used to ensure that each user can only choose local computing or select a certain access point for task offloading, and are specifically expressed as follows:

[0201]

[0202] In the formula, x ij Assign elements in the matrix to users to indicate whether the i-th user has selected the j-th access point. If j=0, it is considered that the user has selected local computing.

[0203] The maximum allocation constraint for access points is used to ensure that the total number of user connections for each access point does not exceed the maximum number of connections for that access point, and it is specifically expressed as follows:

[0204]

[0205] In the formula, L nk (j) represents the maximum number of connections for the j-th access point;

[0206] The user-assigned element constraint is used to ensure that the elements of the user-assigned matrix are binary values, i.e., yes or no, to prevent overflow, and is specifically expressed as follows:

[0207] x ij ∈{0,1},

[0208] The RIS phase constraint condition is used to ensure that the absolute value of the phase adjustment coefficient of each coordinateable unit in the RIS is 1, so that it conforms to the physical properties of RIS phase adjustment, and its specific expression is as follows:

[0209] |θ l |=1,

[0210] The task unloading scheme branch unit 3 is used to execute step S3: using a user-assigned lower bound branch algorithm to branch the current task unloading allocation iteration node and update the optimal task unloading scheme to obtain the updated task unloading allocation iteration node and the optimal task unloading scheme.

[0211] Specifically, the user-assigned lower bound branch algorithm includes the following steps:

[0212] S301. Reset the branch of the user allocation matrix for the task unloading allocation iteration node to obtain the iteration node of the current branch.

[0213] Specifically, the branch reset is represented as follows:

[0214]

[0215] In the formula, Let i' represent the user allocation matrix of the (N+1)i′+kth task unloading allocation iteration node, where i′ represents the iteration number of the current task unloading allocation iteration node, that is, the number of loops in the task unloading scheme branch unit 3 repeatedly executed through step S5; k is the iteration number of the lower bound branch algorithm based on user allocation.

[0216] Indicates the i-th min The user allocation matrix of the transposed task unloading allocation iteration node, that is, the allocation matrix obtained from the current optimal solution node, where i min The index of the iterative node that has the minimum energy consumption estimate (lowest lower bound);

[0217] The identity matrix represents the transpose of the matrix, with a size of (N+1). The element at the k-th position is 1, and the other elements are 0. It is used to indicate the addition of a new access point allocation based on the current iteration node.

[0218] Accordingly, by resetting the branch and updating the access point allocation based on the existing optimal solution, the branch algorithm generates a new task unloading scheme, thereby promoting the optimization process, realizing branch exploration, effectively narrowing the search space and gradually approaching the minimum energy consumption solution.

[0219] S302. Determine whether the iteration node of the current branch is a leaf node: If not, use the minimum energy consumption estimation algorithm to calculate, update the minimum energy consumption estimation value of the iteration node of the current branch according to the calculation result, and execute step S305; if yes, execute step S303.

[0220] Specifically, the minimum energy consumption estimate for the current branch's iteration node is represented as follows:

[0221]

[0222] In the formula, lb (N+1)i′+k This represents the minimum energy consumption estimate for the iterative node (N+1)i′+k of the current branch.

[0223] S303. The RIS phase optimization algorithm is used to update the phase of the iteration node of the current branch to obtain the minimum energy consumption estimate and the optimal task unloading scheme of the iteration node of the current branch.

[0224] Specifically, the minimum energy consumption estimate of the iterative node of the current branch is expressed as follows:

[0225]

[0226] In the formula, θ (N+1)i′+k The phase adjustment coefficient vector representing the RIS reconcilable unit corresponding to the optimal task offloading scheme of the current branch is obtained through the phase update, and its specific calculation is expressed as follows:

[0227]

[0228] In the formula, Opt ROS The RIS phase update algorithm is used to update the RIS phase based on the user allocation matrix of the iterative node of the current branch. Its specific steps are as follows:

[0229] Please also refer to Figure 7 , Figure 7 This is a simplified pseudocode flowchart of the RIS phase update algorithm described in this invention.

[0230] S30301. Calculate the current phase update parameters based on the current phase adjustment coefficient vector of the RIS and the corresponding cascaded channel information.

[0231] Specifically, the current phase update parameters include data transmission rate, signal-to-noise ratio gain, and channel gain correction factor;

[0232] The data transmission rate, used to measure the communication performance between the i-th user and the j-th access point in the u-th iteration, is specifically expressed as follows:

[0233]

[0234] In the formula, R ij (θ (u) ) represents the data transmission rate corresponding to the phase adjustment coefficient vector of the u-th iteration; θ (u) Let θ be the phase adjustment coefficient vector for the u-th iteration. When u = 0, θ (0) Each element has a modulus of 1, used to constrain the randomly initialized phase adjustment coefficient vector; where u represents the number of iterations of the RIS phase update algorithm;

[0235] The signal-to-noise ratio gain is used to measure the enhancement effect of RIS on the signal, and it is specifically expressed as follows:

[0236]

[0237] In the formula, This represents the signal-to-noise ratio gain in the u-th iteration; A matrix representing the cascaded channel information between all users and all access points, and transposed by Hermite, used to comprehensively represent the channel gain between all users and access points;

[0238] The channel gain correction factor is used to guide the optimization direction of the phase during the iterative update process, and is specifically expressed as follows:

[0239]

[0240] In the formula, This represents the channel gain correction factor for the u-th iteration; The expected value of the signal gain is used to represent the amplification factor of the signal by the channel gain. This is the noise power term, used to ensure stable signal energy calculation and avoid unreasonable signal amplification.

[0241] S30302. Based on the current phase update parameters, the coupled influence optimization algorithm is used to perform phase optimization on all elements of the current RIS phase adjustment coefficient vector to obtain the optimized RIS phase adjustment coefficient vector.

[0242] Please also refer to Figure 8 , Figure 8 This is a simplified pseudocode flowchart of the coupling effect optimization algorithm described in this invention.

[0243] Specifically, for the q-th element of the current RIS phase adjustment coefficient vector The phase optimization representation of the coupling effect optimization algorithm in the (t+1)th iteration is as follows:

[0244]

[0245] In the formula, This represents the q-th element of the phase adjustment coefficient vector in the (t+1)-th iteration of the coupling effect optimization algorithm, with an initial value of φ. (0) Let θ be the current phase adjustment coefficient vector of RIS. (u) ; Let be the optimization direction vector of the q-th phase in the t-th iteration of the coupling effect optimization algorithm. It is used to determine the optimization direction of the current RIS reflection unit. The phase angle is adjusted according to the optimization direction to maximize the signal gain. Its specific representation is as follows:

[0246]

[0247] In the formula, Gain effect, used to represent the signal-to-noise ratio, is the enhancement of signal quality by increasing the signal-to-noise ratio.

[0248] The coupling effect term reflects the interaction between the q-th element of the RIS phase adjustment coefficient vector and other reflection units p, ensuring that the optimization process is not limited to local optima. The coupling effect factor Γ between the p-th and q-th elements of the phase adjustment coefficient vector is also considered. pq , its Γ pq Specifically, it is expressed as follows:

[0249]

[0250] In the formula, This represents the square of the channel gain correction factor, used to quantify the signal enhancement effect; The outer product of the channel gain matrix is ​​used to reflect the channel quality from all users to all access points, and to characterize the interaction channel impact between users and access points.

[0251] The superscript (*) indicates complex conjugate, which is used to ensure that the phase adjustment is within the unit modulus range to meet physical constraints.

[0252] Next, the iterative conditions for the coupling effect optimization algorithm are:

[0253] Determine ||φ (t+1) -φ (t) || Is it less than a threshold value ε? If so, then adjust the phase adjustment coefficient vector φ of the RIS from the previous iteration. (t) As the phase adjustment coefficient vector of the optimized RIS, i.e., θ (u+1) =φ (t) If not, then complete phase optimization; otherwise, adjust φ. (t+1) As the phase adjustment coefficient vector of RIS in the current coupled influence optimization algorithm, the coupled influence optimization algorithm is used to further optimize the phase of all elements, which can be specifically expressed as:

[0254]

[0255] Wherein, the threshold value ε∈[0,1] is set to 0.01 by default to ensure that the phase optimization process does not have excessive fluctuations; ‖·‖ represents the norm of the vector, and the L2 norm is used by default.

[0256] Accordingly, this invention uses a coupling effect optimization algorithm to consider the mutual influence between reflection units in the RIS phase optimization process by employing a nested iterative approach. This effectively avoids the performance degradation of RIS after phase optimization due to local optimization and ensures the global optimization of the system. Furthermore, it improves system energy efficiency and task offloading efficiency by maximizing global signal gain.

[0257] S30303. Determine whether the current RIS phase adjustment coefficient vector has converged: If not, continue to execute step S30301; if yes, output the current RIS phase adjustment coefficient vector and complete the iteration of the RIS phase update algorithm.

[0258] Specifically, the convergence is determined by the target energy consumption value corresponding to the optimal energy consumption objective function, which is expressed as follows:

[0259]

[0260] In the formula, δ represents the target energy consumption value corresponding to the phase adjustment coefficient vector of the (u+1)th iteration; δ is the convergence threshold of the RIS phase update algorithm, which is δ∈[0,1] and is used to determine whether the RIS phase adjustment coefficient vector of the current RIS phase update algorithm has converged.

[0261] In addition, vector transformation method or other conditions can be used to determine whether the current RIS phase update algorithm has converged. Therefore, this invention does not specifically limit the choice of convergence conditions.

[0262] S304. Is the minimum energy consumption estimate of the iteration node of the current branch less than the target energy consumption of the current optimal task unloading scheme? If yes, update the current optimal task unloading scheme and execute step S305; otherwise, execute step S305 directly.

[0263] Specifically, the update of the current optimal task unloading scheme is expressed as follows:

[0264] lflb (N+1)i ′ +k <v OPT then

[0265] v OPT =lb (N+1)i ′ +k

[0266]

[0267] θ OPT =θ (N+1)i ′ +k

[0268] Therefore, during the iteration of the lower bound branch algorithm, when using the coupled influence optimization algorithm instead of the minimum energy consumption estimation algorithm, it can be directly determined whether the minimum energy consumption estimate obtained through the coupled influence optimization algorithm is less than the target energy consumption value of the current optimal task offloading scheme. If so, the current optimal task offloading scheme is directly updated to ensure that the scheme with the lowest energy consumption is selected in each iteration, thus driving the optimization process to gradually approach the global optimum.

[0269] S305. Determine whether the current iteration number of the lower bound branch algorithm based on user allocation is less than an iteration condition: if yes, then execute step S301; if no, then complete the iteration of the lower bound branch algorithm.

[0270] The iteration condition is k < (N+1). By using the number of access points as the lower bound for the maximum number of iterations of the branch algorithm, the algorithm can ensure that it can branch based on all access points to fully explore the corresponding task offloading schemes and find the optimal resource allocation scheme.

[0271] The task unloading scheme pruning unit 4 is used to perform step S4: according to the updated target energy consumption value of the optimal task unloading scheme, prune the task unloading allocation iteration node to obtain a refined task unloading allocation iteration node.

[0272] Specifically, the pruning operation is represented as follows:

[0273]

[0274] In the formula, status i Indicates the derived scheme state of the i-th task unloading and allocation iteration node; lb i Let represent the minimum energy consumption estimate for the i-th task to be unloaded and assigned to the iterative node, and This indicates that the task unloading and allocation iteration nodes are traversed through n task unloading and allocation iteration nodes, where n represents the total number of task allocation iteration nodes.

[0275] If lb exists i ≥v OPT This indicates that the minimum energy consumption estimate of the current node is greater than the existing optimal solution, meaning that the node cannot generate a better solution. Therefore, its derived solution status can be changed. i Setting it to 0 prevents its iteration nodes from continuing to participate in branch iterations.

[0276] The task unloading scheme iteration judgment unit 5 is used to execute step S5: determine whether the refined task unloading allocation iteration node satisfies the one-branch delimitation iteration condition: if yes, then unload and allocate the user's computing tasks according to the current optimal task unloading scheme to complete the unloading scheduling of the user's tasks; if no, then continue to call the task unloading scheme branch unit 3.

[0277] Specifically, the branch-bound iteration condition is determined as follows:

[0278]

[0279] Among them, if all derived schemes are in status i If all values ​​are 0, it can be assumed that all branches have been traversed and no new, better solutions have been generated. The algorithm can then directly use the existing optimal task unloading scheme {X}. OPT ,θ OPT The offloading and scheduling is performed based on the phase adjustment coefficient vector θ of the current optimal RIS coordinating unit. OPT The adjustable units of the reconfigurable metasurface are adjusted according to the current optimal user allocation matrix X. OPT Access points are allocated to user equipment, and task offloading and scheduling for user equipment are completed.

[0280] Compared to existing technologies, this invention ensures that the adjustment of user allocation and RIS reflection unit through multi-layer iterative optimization can achieve the global optimal solution for task offloading allocation. In particular, it adopts a lower bound branch algorithm based on user allocation to effectively expand the search space through branch reset and minimum energy consumption estimation algorithm, ensuring the full exploration of the user allocation matrix and avoiding getting trapped in local optima, thereby significantly improving task offloading efficiency and system resource utilization.

[0281] Furthermore, in the RIS phase optimization algorithm, this invention utilizes the user allocation matrix of the corresponding iteration node to optimize the phase of the RIS reflection unit, effectively improving signal quality, reducing interference and attenuation between reflection units, thereby significantly improving the communication performance of the system and further reducing system energy consumption.

[0282] Meanwhile, this invention also uses a coupling effect optimization algorithm to ensure that the adjusted reflection unit avoids multipath effects during the RIS phase optimization process, thereby significantly enhancing the signal quality control capability during the RIS-assisted task offloading process, ensuring the efficiency and stability of task offloading, and ultimately effectively improving the battery life of user equipment.

[0283] Please see Figure 5 , Figure 5 This diagram illustrates a comparison between the joint RIS phase optimization of this invention and other phase selection methods. Local computation refers to the user performing computation solely on their local device, without offloading the task to a remote server; fixed phase refers to the algorithm optimizing only the user-assigned matrix X, while the phase coefficient θ = 1. 16×1 Random selection refers to a feasible solution to the optimal energy consumption objective function under certain constraints.

[0284] Compared to schemes using local computing, fixed phase selection, and random phase selection, the method of this invention not only has significantly lower total energy consumption than other methods, but also has a lower energy consumption growth rate as the number of users increases. In other words, as the number of users increases, the method of this invention can effectively control the growth of system energy consumption, thereby significantly reducing the energy consumption of user devices, improving the utilization rate of computing resources in the MEC system, and thus effectively extending the battery life of user devices.

[0285] Based on the same inventive concept, this application also provides an electronic device, which can be a server, desktop computing device, or mobile computing device (e.g., laptop computing device, handheld computing device, tablet computer, netbook, etc.). The device includes one or more processors and a memory, wherein the processor is used to execute a program to implement the RIS-based user task offloading and scheduling method of this invention; the memory is used to store computer programs executable by the processor.

[0286] Based on the same inventive concept, this application also provides a computer-readable storage medium corresponding to the aforementioned embodiment of a RIS-based user task offloading and scheduling method. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the RIS-based user task offloading and scheduling method described in any of the above embodiments.

[0287] This application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0288] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and the present invention also intends to include these modifications and variations.

Claims

1. A method for user task offloading scheduling based on RIS cooperation, characterized in that, Includes the following steps: S1: Obtain the cascaded channel information and noise power between the user and the access point; S2: performing task offloading initialization according to the cascade channel information and the noise power and an optimal energy consumption target function, to generate an initial task offloading allocation iteration node and an optimal task offloading scheme; wherein the initial task offloading allocation iteration node comprises an initialized user allocation matrix, an initialized derivative scheme state and an initialized minimum energy consumption estimation value; and the optimal energy consumption target function Specifically: In the formula, represents the target energy consumption value corresponding to the optimal task offloading scheme; represents the current optimal user allocation matrix, and the element represents the user selects the access point ; represents the current optimal phase adjustment coefficient vector of the RIS coordination unit. For locally computed energy consumption models, when Then it means the first Some users choose to perform calculations locally; Energy consumption correlation coefficient; To calculate the complexity correlation coefficient; For the first Local computing frequency for each user; Indicates the first The task length for each user; M represents the total number of users; for a remote computing model, for an uplink overhead factor, for a device transmit power of the th user; denotes the total number of access points; denotes the actual data transmission rate of the th user with the th access point at which is calculated as follows: wherein represents the bandwidth allocated to the current user by the th access point; is a channel gain vector between the th user and the th access point, and is transposed by Hermite; is the noise power; The method for calculating the initial minimum energy consumption estimate is as follows: ， wherein is the minimum energy estimate for initialization; is the current user allocation matrix; is the minimum energy estimation algorithm; S3: A user-assigned lower bound branching algorithm is used to branch the current task unloading allocation iteration node and update the optimal task unloading scheme to obtain the updated task unloading allocation iteration node and the optimal task unloading scheme; the lower bound branching algorithm includes: Get the iteration node of the current branch; The RIS phase optimization algorithm is used to update the phase of the iterative nodes that belong to the leaf nodes, to obtain the optimal task unloading scheme and the corresponding target energy consumption value of the current branch. When the target energy consumption value is less than the target energy consumption value of the current optimal task unloading scheme, the current optimal task unloading scheme is updated. The minimum energy consumption estimation method is used to estimate the minimum energy consumption value of the iterative nodes that do not belong to the leaf nodes, and the iterative nodes are updated. Subsequently, if the number of iterations is less than one iteration condition, the lower bound branch algorithm operation is performed again; otherwise, the lower bound branch algorithm iteration is completed. S4: Based on the updated target energy consumption value of the optimal task unloading scheme, prune the task unloading allocation iteration nodes to obtain refined task unloading allocation iteration nodes. S5: Determine whether the refined task unloading and allocation iteration node satisfies the one-branch bounding iteration condition: if yes, then unload and allocate the user's computing tasks according to the current optimal task unloading scheme to complete the unloading and scheduling of user tasks; if no, then continue to execute step S3.

2. The RIS-collaborative-based user task offloading scheduling method according to claim 1, characterized in that, The optimal energy consumption objective function includes several constraints, including user selection constraints, maximum access point allocation constraints, user allocation element constraints, and RIS phase constraints. The specific representation of the user-selected constraints is as follows: The maximum allocation constraint for access points is specifically expressed as follows: In the formula, represents the total number of user connections of the first access point; represents the maximum number of connections of the first access point; The specific constraints on the user-assigned elements are as follows: The RIS phase constraint condition is specifically expressed as follows: wherein denotes the phase adjustment coefficient, i.e. the phase angle, of the j-th coordinated unit of the RIS; is a complex number of unit modulus in the phase coefficient; denotes the total number of coordinated units, i.e. the total number of reflecting units, of the RIS.

3. The RIS-collaborative-based user task offloading scheduling method according to claim 2, characterized in that, The initialized minimum energy consumption estimation value is calculated by using a minimum energy consumption estimation algorithm, and the specific steps are as follows: S201. Based on the size of the user allocation matrix of the current task unloading allocation iteration node, generate a simulated user allocation index array and an initial energy consumption estimate. S202. Generate a simulated user selection submatrix based on the current simulated user allocation index array; S203. Concatenate the simulated user selection submatrix with the current user allocation matrix and determine whether the maximum allocation constraint of an access point is satisfied: if yes, calculate the energy consumption estimate of the concatenated user allocation matrix according to the user allocation energy consumption model and execute step S204; if no, execute step S205. S204. Determine whether the energy consumption estimate of the spliced ​​user allocation matrix is ​​less than the current energy consumption estimate: if yes, update the current energy consumption estimate and execute step S205; if no, directly execute step S205. S205, judging whether the current iteration number meets an energy consumption estimation condition; if yes, taking the current energy consumption estimation value as the minimum energy consumption estimation value of the iteration node for the current task offloading distribution; if no, updating the current simulation user distribution index array and performing step S202.

4. The RIS-collaborative based user task offloading scheduling method of claim 3, wherein, The generation of the simulation user selection sub-matrix is represented as follows: wherein is the sub-matrix selected by the user, denotes the size of the matrix ; is the number of rows of the matrix assigned to the current user; denotes each element of the matrix and the rule that the user selects the access point is expressed as follows: In the formula, This indicates iterating through all users; For the first The simulated user allocation index for each user is calculated as follows: In the formula, the first element of the array of assigned indices for the simulation user; for representing traversing all elements in the array of assigned indices for the simulation user;​ The judgment expression of the access point maximum distribution constraint condition is as follows: In the formula, for indicating traversing all access points; based on the spliced allocation matrix The energy consumption estimation value of is expressed as follows: is the upper bound on the data transfer rate for the nth user with the mth access point, which is calculated as follows: is the upper bound on the data transfer rate for the nth user with the mth access point, which is calculated as follows:​ The specific representation of updating the current simulation user distribution index array in step S205 is as follows: In the formula, represents the first element in the current simulated user allocation index array.

5. The RIS-collaborative based user task offloading scheduling method according to claim 4, characterized in that, The user distribution based lower bound branch algorithm includes the following steps: S301, performing branch reset on the user distribution matrix of the iteration node for the task offloading distribution, and obtaining the iteration node of the current branch; S302, judging whether the iteration node of the current branch is a leaf node; if no, performing calculation by using the minimum energy consumption estimation algorithm, updating the minimum energy consumption estimation value of the iteration node of the current branch according to the calculation result, and performing step S305; if yes, performing step S303; S303, performing phase updating on the iteration node of the current branch by using the RIS phase optimization algorithm, and obtaining the optimal task offloading scheme of the current branch and the corresponding target energy consumption value; S304, judging whether the target energy consumption value of the current branch is less than the target energy consumption value of the current optimal task offloading scheme; if yes, updating the current optimal task offloading scheme and performing step S305; if no, directly performing step S305; S305, judging whether the iteration number of the current user distribution based lower bound branch algorithm is less than an iteration condition; if yes, performing step S301; if no, completing the iteration of the lower bound branch algorithm.

6. The RIS-collaborative based user task offloading scheduling method of claim 5, wherein, The branch reset of step S301 is specifically represented as follows: wherein denotes the user allocation matrix of the task offloading allocation iteration node, which denotes the iteration number of the current task offloading allocation iteration node; is the iteration number of the lower bound based user allocation branch and bound algorithm. Indicates the first The user allocation matrix of the transposed task unloading allocation iteration node, The index of the iterative node that has the minimum energy consumption estimate; denotes the identity matrix of size whose element in the th position is 1 and whose other elements are 0; The updating of the minimum energy consumption estimation value of the iteration node of the current branch according to the calculation result of step S302 is specifically calculated as follows: wherein represents the iteration node of the current branch is the minimum energy estimation of the current branch; is the minimum energy estimation algorithm.

7. The RIS-collaborative-based user task offloading scheduling method according to claim 6, characterized in that, The specific calculation representation of the phase updating of step S303 is as follows: In the formula, represents the phase adjustment coefficient vector of the RIS coordination unit corresponding to the optimal task offloading scheme of the most current branch; The RIS phase updating algorithm is as follows: S30301, calculating the current phase updating parameter according to the phase adjustment coefficient vector of the current RIS and the corresponding cascaded channel information; wherein the current phase updating parameter includes a data transmission rate, a signal noise ratio gain and a channel gain correction factor; The specific representation of the data transmission rate is as follows: In the formula, the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; and the phase adjustment coefficient vector corresponding to the data transmission rate of the first iteration; and The specific representation of the signal noise ratio gain is as follows: wherein denotes the signal-to-noise ratio gain of the denotes the matrix of the concatenated channel information between all users and all access points and is transposed by Hermite.​ The specific representation of the channel gain correction factor is as follows: wherein denotes the channel gain correction factor of the denotes the channel gain correction factor of the denotes the expected term of the signal gain; is the noise power term; S30302, performing phase optimization on all elements of the phase adjustment coefficient vector of the current RIS by using the coupling influence optimization algorithm according to the current phase updating parameter, and obtaining the optimized phase adjustment coefficient vector of the RIS; S30303, judging whether the phase adjustment coefficient vector of the current RIS converges; if no, continuing to perform step S30301; if yes, outputting the phase adjustment coefficient vector of the current RIS and completing the iteration of the user distribution based lower bound branch algorithm; The convergence is judged by using the target energy consumption value corresponding to the optimal energy consumption objective function, and the specific representation is as follows: In the formula, represents the phase adjustment coefficient vector corresponding to the target energy consumption value of the first iteration; represents the phase adjustment coefficient vector corresponding to the target energy consumption value of the first iteration; is a convergence threshold value of the RIS phase update algorithm.

8. The RIS-collaborative based user task offloading scheduling method of claim 7, wherein, The coupling influence optimization algorithm represents the phase optimization of the i-th element of the current RIS's phase-adjusted coefficient vector as follows: ​​ In the formula, In the optimization algorithm representing the coupling effect, the first... The phase adjustment coefficient vector of the nth iteration Each element has an initial value. The current phase adjustment coefficient vector of RIS ; In the optimization algorithm for coupling effects, the first The iteration of the ... The optimized direction vector for each phase is specifically represented as follows: In the formula, Gain effect used to represent the signal-to-noise ratio; This represents the coupling effect term, for the first term of the phase adjustment coefficient vector. The element and the first Coupling influence factor of individual elements ,That Specifically, it is expressed as follows: wherein denotes the square of the channel gain correction factor; denotes the outer product of the channel gain matrix; wherein the superscript denotes the complex conjugate; the coupling influence optimization algorithm's iteration condition is: determining whether the value is less than a threshold value : if yes, the phase adjustment coefficient vector of the RIS in the last round of iteration is taken as the phase adjustment coefficient vector of the optimized RIS, that is , the phase optimization is completed; if no, the is taken as the phase adjustment coefficient vector of the RIS in the current coupling influence optimization algorithm, and the coupling influence optimization algorithm is continued to be used to perform phase optimization on all elements; wherein the threshold value .

9. A user task offloading scheduling apparatus based on RIS cooperation, characterized in that, The user's cascaded channel information acquisition unit is configured to execute step S1 in the user task offloading scheduling method based on RIS cooperation according to any one of claims 1-8. The task offloading scheme initialization unit is configured to execute step S2 in the user task offloading scheduling method based on RIS cooperation according to any one of claims 1-8. The task offloading scheme branching unit is configured to execute step S3 in the user task offloading scheduling method based on RIS cooperation according to any one of claims 1-8. The task offloading scheme pruning unit is configured to execute step S4 in the user task offloading scheduling method based on RIS cooperation according to any one of claims 1-8. The task offloading scheme iteration judgment unit is configured to execute step S5 in the user task offloading scheduling method based on RIS cooperation according to any one of claims 1-8. The reconfigurable intelligent surface (RIS), a base station, and a task offloading scheduling device based on RIS cooperation in communication connection with the reconfigurable intelligent surface and the base station are included. 10.A mobile edge computing system based on reconfigurable intelligent surface, characterized in that, The reconfigurable intelligent surface is provided with a plurality of tunable units, which are configured to receive a computing task offloading request from a user equipment, and to transmit signal information to the base station by adjusting the reflection phase through the tunable units; wherein the user equipment is configured to generate a computing task and to send the computing task and corresponding computing requirements to the reconfigurable intelligent surface. The base station is provided with an access point connected with an edge computing server, which is configured to calculate the cascaded channel information and noise power between the current user and the access point according to the computing task and the corresponding computing requirements, and to input them into the task offloading scheduling device based on RIS cooperation. The task offloading scheduling device based on RIS cooperation is configured to execute the user task offloading scheduling method based on RIS cooperation according to any one of claims 1-8. ​

Images