A task offloading method for rail transit 5G edge computing network system

Abstract

The present invention provides a task offloading method for a rail transit 5G edge computing network system, including: establishing a system task model, dividing the system task model into a URLLC task and an eMBB task; establishing a task offloading model for the URLLC task, and calculating the total processing delay of the URLLC task, and establishing a task offloading model for the eMBB task, and calculating the total processing delay of the eMBB task and the maximum transaction processing volume of the eMBB task; establishing a joint optimization model according to the total processing delay of the URLLC task, the total processing delay of the eMBB task and the maximum transaction processing volume of the eMBB task; solving the joint optimization model using the MATD3 algorithm to obtain an optimal task offloading strategy. The present invention can provide higher task service quality for the urban rail transit system.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a task offloading method for a rail transit 5G edge computing network system. Background Art

[0002] The integration of digitalization, networking, and intelligence is the development trend and characteristic of urban rail transit, accompanied by a surge in intensive computing tasks and real-time requirements. Existing research has established the resource allocation problem of URLLC (ultra-high reliability and ultra-low latency) and eMBB (enhanced mobile broadband) tasks in edge computing networks through 5G multi-channel wireless access technology, but has not considered the combination of URLLC and eMBB tasks in more complex and highly dynamic scenarios.

[0003] However, the service requirements of various intelligent application services in urban rail transit systems are heterogeneous. Delay-sensitive application services such as train autonomous control and operation tracking require low latency and high reliability, and related applications such as passenger information systems and closed-circuit television videos have large data transmission requirements. Therefore, how to use 5G edge computing networks to provide higher task service quality for urban rail transit systems is a technical problem that technicians in this field need to solve. Summary of the invention

[0004] The purpose of the present invention is to provide a task offloading method for a rail transit 5G edge computing network system to provide higher task service quality for the urban rail transit system.

[0005] A task offloading method for a rail transit 5G edge computing network system, the rail transit 5G edge computing network system includes a user layer, an edge layer and a digital twin layer, the user layer includes multiple user devices, the edge layer includes multiple edge servers, the user layer and the edge layer are both virtually mapped in the digital twin layer, and the method includes the following steps:

[0006] Step S1, establishing a system task model, dividing the system task model into URLLC tasks and eMBB tasks, URLLC tasks are offloaded through URLLC transmission links, and eMBB tasks are offloaded through eMBB transmission links;

[0007] Step S2, establishing a task offloading model for URLLC tasks and calculating the total processing delay of URLLC tasks, and establishing a task offloading model for eMBB tasks and calculating the total processing delay of eMBB tasks and the maximum transaction processing volume of eMBB tasks;

[0008] Step S3, establishing a joint optimization model according to the total processing delay of the URLLC task, the total processing delay of the eMBB task, and the maximum transaction processing volume of the eMBB task;

[0009] Step S4, using the MATD3 algorithm to solve the joint optimization model to obtain the optimal task offloading strategy.

[0010] According to the task offloading method of the rail transit 5G edge computing network system provided by the present invention, a digital twin layer is referenced, and both the user layer and the edge layer are virtually mapped in the digital twin layer. In the digital twin layer, a digital twin model is established, and the system task model is virtually mapped in the digital twin model. The system task model is centrally trained through the digital twin model, and then the trained system task model is deployed to each user device and each edge server for distributed execution. The method first calculates the total processing delay of the URLLC task, the total processing delay of the eMBB task, and the maximum transaction processing volume of the eMBB task, and then calculates the total processing delay of the URLLC task and the maximum transaction processing volume of the eMBB task according to the URLLC task. The total processing delay of RLLC tasks, the total processing delay of eMBB tasks and the maximum transaction processing volume of eMBB tasks are taken into consideration, and a joint optimization model is established. The short packet transmission of URLLC links and the large-capacity bandwidth of eMBB links meet the heterogeneous requirements of different services in urban rail transit systems, and the combination of URLLC and eMBB tasks is applied in more complex and highly dynamically changing scenarios. Finally, the MATD3 algorithm is used to solve the joint optimization model, which reduces the pressure of data transmission when the intelligent agent interacts online with the central controller, thereby realizing the dynamic resource allocation of URLLC and eMBB and improving the task processing performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is the structural block diagram of the rail transit 5G edge computing network system;

[0012] Figure 2 This is a flow chart of the task offloading method of the rail transit 5G edge computing network system of the present invention. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The embodiment of the present invention provides a task offloading method for a rail transit 5G edge computing network system. Figure 1The rail transit 5G edge computing network system includes a user layer, an edge layer and a digital twin layer. The user layer includes multiple user devices, the edge layer includes multiple edge servers, and the user layer and the edge layer are virtually mapped in the digital twin layer.

[0015] Specifically, in the user layer, the user equipment set is set as , The total number of user devices in M max , Middle m User devices Task characteristics ,in, represents the priority weight, Indicates the data transfer size. Indicates the number of CPU cycles required for task calculation. Indicates the maximum delay tolerance, Indicates the task type. , and They refer to URLLC tasks and eMBB tasks respectively.

[0016] In the edge layer, the edge server set is set as , The total number of edge servers in K max ,The edge servers in the edge server set are deployed beside the tracks and at urban rail transit stations. The edge servers are wired to the 5G base stations (BS), providing computing services for user devices on the train at the edge of the network.

[0017] In the digital twin layer, a digital twin model is established, the system task model is virtually mapped in the digital twin model, the system task model is centrally trained through the digital twin model, and then the trained system task model is deployed to each user device and each edge server for distributed execution. The expression is ,in, It is m A virtual task model of a user device, , Indicates m The total amount of local computing resources of each user device, and Respectively represent m The computing resources allocated to URLLC tasks and eMBB tasks by user equipment meet the requirements , It is kThe virtual task model of edge servers, , Indicates k The total computing resources of edge servers, and Respectively represent k The edge server is assigned to m The computing resources of URLLC tasks and eMBB tasks in the user equipment meet , is the communication resource management model, , Indicates k The available communication bandwidth of the base stations to which the edge servers are connected, and Respectively represent k The edge server is assigned to m The communication resources of URLLC tasks and eMBB tasks in the user equipment are satisfied. .

[0018] The above design can realize the virtual mapping of all physical devices (user devices and edge servers) and communication environments in the digital twin layer, optimize resource allocation, improve user experience, and eliminate the need to occupy actual physical device resources for model learning and testing.

[0019] See also Figure 2 , the method comprises steps S1 to S4:

[0020] Step S1, establish a system task model, divide the system task model into URLLC tasks and eMBB tasks, URLLC tasks are offloaded through the URLLC transmission link, and eMBB tasks are offloaded through the eMBB transmission link.

[0021] Step S2, establish a task offloading model for URLLC tasks and calculate the total processing delay of URLLC tasks, and establish a task offloading model for eMBB tasks and calculate the total processing delay of eMBB tasks and the maximum transaction processing volume of eMBB tasks.

[0022] In step S2, the total processing delay of the URLLC task is calculated using the following formula:

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] in, Indicates m Total processing delay of URLLC tasks for each user equipment, Indicates m URLLC tasks for user equipment t +1 time slot waiting delay, Indicates m URLLC tasks for user equipment t The waiting delay of the time slot, Indicates the m A binary variable indicating whether the URLLC task of a user device is in the waiting stage of the request buffer queue; =1, indicating the m The URLLC task of each user device is in the waiting stage of the request buffer queue; Indicates the length of a single time slot, Indicates m The local calculation delay of the URLLC task of each user equipment, Indicates m A binary variable indicating whether the URLLC task of each user device is completely offloaded to the local server processing, and ;when When m The URLLC tasks of each user device are completely offloaded to the local server for processing; Indicates the number of CPU cycles required for URLLC task calculation. Indicates m The latency of transmitting the URLLC task of a user device to the edge server, Indicates m User devices are uninstalled to k The data transmission size of the URLLC task of the edge server is Indicates m The data transmission rate of each user equipment performing URLLC tasks, Indicates m Whether the URLLC task of the first user equipment is completely offloaded to the k binary variables processed in edge servers, and ;when When m The URLLC task of each user device is completely offloaded to the k Processed in edge servers; Indicates the data transmission size of the URLLC task, Indicatesm User devices are uninstalled to k The number of CPU cycles required for URLLC task calculation of each edge server.

[0029] In step S2, the total processing delay of the eMBB task is calculated using the following formula:

[0030] ;

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] ;

[0037] in, Indicates m Total processing delay of eMBB tasks for each user equipment, Indicates m The total latency of the eMBB task calculated locally for each user equipment, , , , Respectively represent m The eMBB tasks of the UEs are offloaded to the first, second, and third k , K The total processing delay of edge servers is Indicates m The eMBB task of each user equipment is t +1 time slot waiting delay, Indicates m The eMBB task of each user equipment is t The waiting delay of the time slot, Indicates the m A binary variable indicating whether the eMBB task of a user equipment is in the waiting stage of the request buffer queue; When m The eMBB task of a user equipment is in the waiting stage of the request buffer queue; Indicates m The eMBB task of each user equipment is t +1 time slot local computation delay, Indicatesm The eMBB task of each user equipment is t The local computation delay of the time slot, Indicates the m A binary variable indicating whether the eMBB task of a user equipment is in the calculation phase of local calculation; When m The eMBB task of each user equipment is in the calculation stage of local calculation; Indicates m The eMBB task of each user equipment is t +1 The number of remaining CPU cycles required for local computation in the time slot, Indicates m User devices in t +1 time slot is allocated to computing resources for eMBB tasks, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot transmission delay, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t The transmission delay of the time slot, Indicates the m User devices are uninstalled to k A binary variable indicating whether the eMBB task of an edge server is in the transmission phase; When m User devices are uninstalled to k The eMBB task of the edge server is in the transmission phase; Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot remaining transmission data size, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot data transmission rate, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot calculation delay, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t The calculation delay of the time slot, Indicates the m User devices are uninstalled to k A binary variable indicating whether the eMBB task of an edge server is in the computing phase; When m User devices are uninstalled to k The eMBB tasks of the edge servers are in the computing stage; Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot calculates the number of remaining CPU cycles required, Indicated in t +1 time slot is assigned to m The computing resources for the eMBB services of each user equipment.

[0038] In step S2, the maximum transaction processing volume of the eMBB task is calculated using the following formula:

[0039] ;

[0040] in, Indicates m The eMBB task of each user equipment is t Maximum transaction volume of a time slot, Indicated in t Time slot m The computing resources allocated to eMBB tasks by user equipment, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t The data transmission rate of the time slot, Indicated in t Time slot k The edge server is assigned to m The computing resources for eMBB tasks in each user device.

[0041] Step S3: Establish a joint optimization model based on the total processing delay of the URLLC task, the total processing delay of the eMBB task, and the maximum transaction processing volume of the eMBB task.

[0042] Wherein, the expression of the joint optimization model is:

[0043] ;

[0044] ;

[0045] in, expresst The joint performance index of the time slot, and is the balance factor, and Respectively represent m The priority weight of the URLLC task and the priority weight of the eMBB task of each user equipment, Indicates the uninstallation policy. , Indicates m The offloading ratio of eMBB tasks of user equipment to the local server, Indicates m The eMBB tasks of the UE are offloaded to the k Offload ratio in edge servers, represents the communication resource allocation strategy, , represents the computing resource allocation strategy, , represents a set of time slots;

[0046] The joint optimization model satisfies the following constraints:

[0047] Constraint C1: , , ;

[0048] Constraint C2: , , ;

[0049] Constrained C3: ;

[0050] Constrained C4: ;

[0051] Constrained C5: ;

[0052] Constrained C6: ;

[0053] Constrained C7: , ;

[0054] Constrained C8: , ;

[0055] Constrained C9: , ;

[0056] Constrained C10: ;

[0057] Constrained C11: ;

[0058] in, Indicates the m Tasks for each user device A binary variable indicating whether the request is in the waiting phase of the request buffer queue; =1, indicating the m Tasks for each user device In the waiting stage of the request buffer queue; Indicates the m User devices are uninstalled to k Edge Server Tasks Binary variable indicating whether it is in the transmission phase; =1, indicating the m User devices are uninstalled to k Edge Server Tasks In the transmission phase; Indicates the m Tasks for each user device A binary variable indicating whether the local calculation is in the calculation phase; when =1, indicating the m Tasks for each user device In the calculation stage of local calculation; Indicates m User devices are uninstalled to k The remaining transmission data size of the eMBB task of the edge server, Indicates the eMBB task data transmission size. Indicates m User devices are uninstalled to k The remaining CPU cycles required for eMBB task calculation of each edge server, Indicates the number of CPU cycles required for eMBB task calculation. Indicates the maximum total processing delay of the URLLC task. Indicates the maximum total processing delay of eMBB tasks.

[0059] Among the above constraints, constraints C1 and C2 respectively restrict the offloading methods of URLLC tasks and eMBB tasks. Constraints C3 and C4 indicate that all tasks can only run in one stage. Constraints C5 and C6 restrict the remaining transmission data size and the number of CPU cycles required for the remaining calculation. Constraints C7, C8 and C9 restrict all allocated communication resources and computing resources from exceeding the maximum available communication resources and computing resources of the device itself. Constraints C10 and C11 indicate that URLLC tasks and eMBB tasks should meet their latency constraints.

[0060] Step S4, using the MATD3 algorithm to solve the joint optimization model to obtain the optimal task offloading strategy.

[0061] Specifically, firstly, a dual critic network and a corresponding target network are used. When calculating the target value, the smaller one of the dual critic network calculation result and the target network calculation result is selected. Then, a delayed policy update rule is adopted to stabilize the training process of the actor network updating the wrong policy in the initial stage. Finally, random clipping noise is added to the target action to solve the joint optimization model and obtain the optimal task offloading strategy.

[0062] In summary, according to the task offloading method of the rail transit 5G edge computing network system provided by the present invention, a digital twin layer is referenced, and both the user layer and the edge layer are virtually mapped in the digital twin layer. In the digital twin layer, a digital twin model is established, and the system task model is virtually mapped in the digital twin model. The system task model is centrally trained through the digital twin model, and then the trained system task model is deployed to each user device and each edge server for distributed execution. The method first calculates the total processing delay of the URLLC task, the total processing delay of the eMBB task, and the maximum transaction processing volume of the eMBB task, and then calculates the total processing delay of the URLLC task, the total processing delay of the eMBB task, and the maximum transaction processing volume of the eMBB task. According to the total processing delay of URLLC tasks, the total processing delay of eMBB tasks and the maximum transaction processing volume of eMBB tasks, a joint optimization model is established. The short packet transmission of URLLC links and the large-capacity bandwidth of eMBB links are used to meet the heterogeneous needs of different services in the urban rail transit system, and the combination of URLLC and eMBB tasks is applied in more complex and highly dynamically changing scenarios. Finally, the MATD3 algorithm is used to solve the joint optimization model, which reduces the pressure of data transmission when the intelligent agent interacts online with the central controller, thereby realizing the dynamic resource allocation of URLLC and eMBB and improving the task processing performance.

[0063] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0064] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A task offloading method for a rail transit 5G edge computing network system, characterized in that: The rail transit 5G edge computing network system includes a user layer, an edge layer and a digital twin layer, the user layer includes multiple user devices, the edge layer includes multiple edge servers, and the user layer and the edge layer are both virtually mapped in the digital twin layer. The method includes the following steps: Step S1, establishing a system task model, dividing the system task model into URLLC tasks and eMBB tasks, URLLC tasks are offloaded through URLLC transmission links, and eMBB tasks are offloaded through eMBB transmission links; Step S2, establishing a task offloading model for URLLC tasks and calculating the total processing delay of URLLC tasks, and establishing a task offloading model for eMBB tasks and calculating the total processing delay of eMBB tasks and the maximum transaction processing volume of eMBB tasks; Step S3, establishing a joint optimization model according to the total processing delay of the URLLC task, the total processing delay of the eMBB task, and the maximum transaction processing volume of the eMBB task; Step S4, using the MATD3 algorithm to solve the joint optimization model to obtain the optimal task offloading strategy; In the user layer, the user equipment set is set to , Middle m User devices Task characteristics ,in, represents the priority weight, Indicates the data transfer size. Indicates the number of CPU cycles required for task calculation. Indicates the maximum delay tolerance, Indicates the task type. , and They refer to URLLC tasks and eMBB tasks respectively; In the edge layer, the edge server set is set as ,The edge servers in the edge server set are deployed beside the tracks and at urban rail transit stations. The edge servers are connected to the 5G base stations by wire, providing computing services for the user devices on the train at the edge of the network; In the digital twin layer, a digital twin model is established, the system task model is virtually mapped in the digital twin model, the system task model is centrally trained through the digital twin model, and then the trained system task model is deployed to each user device and each edge server for distributed execution. The expression is ,in, It is m A virtual task model of a user device, , Indicates m The total amount of local computing resources of each user device, and Respectively represent m The computing resources allocated to URLLC tasks and eMBB tasks by user equipment meet the requirements , It is k The virtual task model of edge servers, , Indicates k The total computing resources of edge servers, and Respectively represent k The edge server is assigned to m The computing resources of URLLC tasks and eMBB tasks in the user equipment meet , is the communication resource management model, , Indicates k The available communication bandwidth of the base stations to which the edge servers are connected, and Respectively represent k The edge server is assigned to m The communication resources of URLLC tasks and eMBB tasks in the user equipment are satisfied. ; In step S2, the total processing delay of the URLLC task is calculated using the following formula: ； ； ； ； ； in, Indicates m Total processing delay of URLLC tasks for each user equipment, Indicates m URLLC tasks for user equipment t +1 time slot waiting delay, Indicates m URLLC tasks for user equipment t The waiting delay of the time slot, Indicates the m A binary variable indicating whether the URLLC task of a user device is in the waiting stage of the request buffer queue; =1, indicating the m The URLLC task of each user device is in the waiting stage of the request buffer queue; Indicates the length of a single time slot, Indicates m The local calculation delay of the URLLC task of each user equipment, Indicates m A binary variable indicating whether the URLLC task of each user device is completely offloaded to the local server processing, and ;when When m The URLLC tasks of each user device are completely offloaded to the local server for processing; Indicates the number of CPU cycles required for URLLC task calculation. Indicates m The latency of transmitting the URLLC task of a user device to the edge server, Indicates m User devices are uninstalled to k The data transmission size of the URLLC task of the edge server is Indicates m The data transmission rate of each user equipment performing URLLC tasks, Indicates m Whether the URLLC task of the first user equipment is completely offloaded to the k binary variables processed in edge servers, and ;when When m The URLLC task of each user device is completely offloaded to the k Processed in edge servers; Indicates the data transmission size of the URLLC task, Indicates m User devices are uninstalled to k The number of CPU cycles required to calculate the URLLC task of each edge server; In step S2, the total processing delay of the eMBB task is calculated using the following formula: ； ； ； ； ； ； ； in, Indicates m Total processing delay of eMBB tasks for each user equipment, Indicates m The total latency of the eMBB task calculated locally for each user equipment, , , , Respectively represent m The eMBB tasks of the UEs are offloaded to the first, second, and third k , K The total processing delay of edge servers is Indicates m The eMBB task of each user equipment is t +1 time slot waiting delay, Indicates m The eMBB task of each user equipment is t The waiting delay of the time slot, Indicates the m A binary variable indicating whether the eMBB task of a user equipment is in the waiting stage of the request buffer queue; When m The eMBB task of a user equipment is in the waiting stage of the request buffer queue; Indicates m The eMBB task of each user equipment is t +1 time slot local computation delay, Indicates m The eMBB task of each user equipment is t The local computation delay of the time slot, Indicates the m A binary variable indicating whether the eMBB task of a user equipment is in the calculation phase of local calculation; When m The eMBB task of each user equipment is in the calculation stage of local calculation; Indicates m The eMBB task of each user equipment is t +1 The number of remaining CPU cycles required for local computation in the time slot, Indicates m User devices in t +1 time slot is allocated to computing resources for eMBB tasks, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot transmission delay, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t The transmission delay of the time slot, Indicates the m User devices are uninstalled to k A binary variable indicating whether the eMBB task of an edge server is in the transmission phase; When m User devices are uninstalled to k The eMBB task of the edge server is in the transmission phase; Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot remaining transmission data size, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot data transmission rate, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot calculation delay, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t The calculation delay of the time slot, Indicates the m User devices are uninstalled to k A binary variable indicating whether the eMBB task of an edge server is in the computing phase; When m User devices are uninstalled to k The eMBB tasks of the edge servers are in the computing stage; Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t +1 time slot calculates the number of remaining CPU cycles required, Indicated in t +1 time slot is assigned to m Computing resources for eMBB services of user equipment; In step S2, the maximum transaction volume of the eMBB task is calculated using the following formula: ； in, Indicates m The eMBB task of each user equipment is t Maximum transaction volume of a time slot, Indicated in t Time slot m The computing resources allocated to eMBB tasks by user equipment, Indicates m User devices are uninstalled to k The eMBB tasks of the edge servers are t The data transmission rate of the time slot, Indicated in t Time slot k The edge server is assigned to m Computing resources for eMBB tasks in each user device; In step S3, the expression of the joint optimization model is: ； ； in, express t The joint performance index of the time slot, and is the balance factor, and Respectively represent m The priority weight of the URLLC task and the priority weight of the eMBB task of each user equipment, Indicates the uninstallation policy. , Indicates m The offloading ratio of eMBB tasks of user equipment to the local server, Indicates m The eMBB tasks of the UE are offloaded to the k Offload ratio in edge servers, represents the communication resource allocation strategy, , represents the computing resource allocation strategy, , represents a set of time slots; The joint optimization model satisfies the following constraints: ， ， ； ， ， ； ； ； ； ； ， ； ， ； ， ； ； ； in, Indicates the m Tasks for each user device A binary variable indicating whether the request is in the waiting phase of the request buffer queue; =1, indicating the m Tasks for each user device In the waiting stage of the request buffer queue; Indicates the m User devices are uninstalled to k Edge Server Tasks Binary variable indicating whether it is in the transmission phase; =1, indicating the m User devices are uninstalled to k Edge Server Tasks In the transmission phase; Indicates the m Tasks for each user device A binary variable indicating whether the local calculation is in the calculation phase; when =1, indicating the m Tasks for each user device It is in the calculation stage of local calculation; Indicates m User devices are uninstalled to k The remaining transmission data size of the eMBB task of the edge server, Indicates the eMBB task data transmission size. Indicates m User devices are uninstalled to k The remaining CPU cycles required for eMBB task calculation of each edge server, Indicates the number of CPU cycles required for eMBB task calculation. Indicates the maximum total processing delay of the URLLC task. Indicates the maximum total processing delay of eMBB tasks.

2. The task offloading method of the rail transit 5G edge computing network system according to claim 1 is characterized in that: Step S4 specifically includes: Firstly, a dual critic network and a corresponding target network are used. When calculating the target value, the smaller one between the dual critic network calculation result and the target network calculation result is selected. Then, a delay strategy is adopted to update the rule. Finally, random clipping noise is added to the target action to solve the joint optimization model and obtain the optimal task offloading strategy.

Images