Resource allocation strategy generation method, device and equipment based on satellite edge calculation

Abstract

The invention discloses a resource allocation strategy generation method, device and equipment based on satellite edge computing, which are applied to a satellite-ground mixed edge computing system, and comprises the following steps: calculating a signal-to-noise ratio, a decoding rate, a transmission data volume and data transmission energy consumption of signal transmission of a satellite and a user terminal according to an unloading decision parameter, the time slot calculation task load and the time slot calculation energy consumption are calculated; calculating time slot total energy consumption based on the data transmission energy consumption and the time slot calculation energy consumption, and constructing an average energy consumption objective function; constructing a target optimization model by adopting a Lyapunov algorithm in combination with the average energy consumption target function and the decision constraint condition; and performing hierarchical decision solving on the target optimization model based on a Markov decision and an SCA algorithm to obtain a target resource allocation strategy. According to the resource allocation method and device, the technical problem that actual resource allocation is lack of accuracy, reliability and applicability due to the fact that factors such as task scheduling, resource allocation, communication link states, user heterogeneous characteristics and various index performance of a system are not comprehensively considered in the prior art can be solved.

Description

Technical Field

[0001] The present application relates to the field of distributed computing technology, and in particular to a method, device and equipment for generating a resource allocation strategy based on satellite edge computing. Background Art

[0002] Although some networked devices are equipped with powerful central processing units, they still cannot meet the processing requirements of computing-intensive tasks. Therefore, how to efficiently use computing resources has become an important challenge for today's network technology. As a centralized computing model, cloud computing provides important support for networked devices through powerful computing power and on-demand resource supply. However, the cloud computing model also has defects such as high latency, high energy consumption, data security risks and insufficient user experience. Therefore, an extended computing model of cloud computing - edge computing EC (Edge Computing) was proposed.

[0003] Edge computing significantly reduces transmission delays and energy consumption and improves service quality by sinking computing resources to the edge of the network, close to the user end. Since ground-based edge computing has limited coverage and is easily damaged by disasters, low-orbit satellites (LEO) and satellite-borne edge computing have more advantages because they can seamlessly cover a larger area of ​​user ends, and the distributed processing of data on satellite edge servers makes information distribution more dispersed and more difficult to attack; in addition, it can also achieve collaborative work between satellites and the ground, and realize dynamic allocation of tasks.

[0004] However, the current edge computing of low-orbit satellites (LEO) still has some shortcomings. For example, it independently handles task scheduling and resource allocation, and fails to consider the coupling effect between the two, resulting in low overall system performance and insufficient resource utilization. It also fails to consider heterogeneous characteristics such as communication conditions between user terminals, resulting in resource allocation that is not objective and reliable enough. In addition, most existing technologies optimize a single goal, which is bound to ignore other system performance, resulting in a lack of accuracy in the overall resource allocation of the system. Furthermore, current technologies are unable to adapt to complex scenarios where both communication link status and resource allocation change, so their applications are limited and it is difficult to meet actual application needs. Summary of the invention

[0005] The present application provides a method, device and equipment for generating a resource allocation strategy based on satellite edge computing, which is used to solve the technical problem that the existing technology does not comprehensively consider factors such as task scheduling, resource allocation, communication link status, user heterogeneous characteristics and multiple indicators of the system performance, resulting in the lack of accuracy, reliability and applicability of actual resource allocation.

[0006] In view of this, the first aspect of the present application provides a resource allocation strategy generation method based on satellite edge computing, which is applied to a satellite-ground hybrid edge computing system, including:

[0007] Determine the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy, and obtain the offloading decision parameter;

[0008] Calculate the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of satellite and user terminal signal transmission according to the offloading decision parameters;

[0009] Calculating the time slot computing task amount and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameter and the actual computing capacity, wherein the actual computing capacity includes the user computing capacity and the satellite computing capacity;

[0010] Calculate the total energy consumption of the time slot based on the data transmission energy consumption and the time slot calculation energy consumption, and construct an average energy consumption objective function according to the total energy consumption of the time slot;

[0011] Constructing a plurality of decision constraints according to the offloading decision parameters, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variables;

[0012] After constructing an initial queue energy consumption optimization model by combining the average energy consumption objective function and the decision constraint condition, the initial queue energy consumption optimization model is simplified by using the Lyapunov algorithm to obtain a target optimization model;

[0013] The target optimization model is solved by hierarchical decision making based on Markov decision making and SCA algorithm to obtain the target resource allocation strategy.

[0014] Preferably, the method of determining the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy to obtain the offloading decision parameter also includes:

[0015] Dynamic communication resource allocation strategy, task offloading strategy and computing resource allocation strategy are formulated according to the mobility of user terminals, task requirements and link status.

[0016] Preferably, the calculating of the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of the satellite and user terminal signal transmission according to the offloading decision parameters includes:

[0017] Calculate the signal-to-noise ratio of each signal sent from the user terminal to the satellite according to the offloading decision parameter, the channel model and the data transmission power;

[0018] Calculating a decoding rate of the information stream according to the signal-to-noise ratio and the total bandwidth of the system;

[0019] Calculate the transmission data volume of the information flow according to the decoding rate and the task transmission time;

[0020] The time slot transmission energy consumption of the user terminal is calculated in combination with the offloading decision parameter and the data transmission power to obtain the data transmission energy consumption.

[0021] Preferably, the calculating of the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of the satellite and user terminal signal transmission according to the offloading decision parameters further includes:

[0022] The RSMA technology is used to send the information flow of the user terminal to the satellite for calculation according to the transmission data volume and the data transmission energy consumption;

[0023] The satellite decodes the received information stream and allocates computing resources based on SIC technology and computing resource allocation strategy.

[0024] Preferably, the calculating the time slot computing task amount and the time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameter and the actual computing capacity includes:

[0025] Calculating the computing task amount of the user terminal according to the offloading decision parameter and the user computing capability to obtain the user computing task amount;

[0026] Calculating the computing energy consumption of the user terminal according to the offloading decision parameter, the user computing capability and the transmission computing time to obtain the user computing energy consumption;

[0027] Calculating the satellite computing task amount according to the satellite computing capability and the satellite processing density to obtain the satellite computing task amount;

[0028] Calculating the satellite computing energy consumption according to the satellite computing capability and the satellite computing time to obtain the satellite computing energy consumption;

[0029] The user computing task amount and the satellite computing task amount constitute a time slot computing task amount;

[0030] The user computing energy consumption and the satellite computing energy consumption constitute the time slot computing energy consumption.

[0031] Preferably, the multiple decision constraints are constructed according to the offloading decision parameters, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variables, and the above also includes:

[0032] Calculate the user dynamic evolution state of the user terminal queue according to the transmission data volume, the time slot calculation task volume and the time slot task arrival volume;

[0033] Calculating the user task discarding amount of the user terminal queue according to the user dynamic evolution state, the time slot task arrival amount and the user queue upper limit value;

[0034] Calculating the satellite dynamic evolution state of the satellite queue according to the transmission data volume and the time slot calculation task volume;

[0035] Calculating the satellite mission discard amount of the satellite queue according to the satellite dynamic evolution state, the transmission data amount and the satellite queue upper limit value;

[0036] The user dynamic evolution state and the satellite dynamic evolution state are queue dynamic evolution states;

[0037] The user task discard amount and the satellite task discard amount are queue task discard amounts.

[0038] Preferably, after the initial queue energy consumption optimization model is constructed by combining the average energy consumption objective function and the decision constraint condition, the initial queue energy consumption optimization model is simplified by using the Lyapunov algorithm to obtain a target optimization model, including:

[0039] Constructing an initial queue energy consumption optimization model in combination with the average energy consumption objective function and the decision constraints;

[0040] Defining a virtual packet loss queue according to the queue task loss amount;

[0041] A Lyapunov function is defined according to the virtual packet loss queue and the dynamic evolution state of the queue using a Lyapunov algorithm;

[0042] Defining a drift penalty term according to the Lyapunov function, the total energy consumption of the time slot and the energy consumption weight parameter;

[0043] The drift penalty term is minimized and optimized through a defined queue stability cost function, and the initial queue energy consumption optimization model is simplified into a target optimization model.

[0044] Preferably, the target optimization model is subjected to hierarchical decision solving based on Markov decision and SCA algorithm to obtain the target resource allocation strategy, including:

[0045] Dividing the target optimization model into an upper optimization model and a lower optimization model according to the hierarchical reinforcement learning method guided by the Lyapunov algorithm;

[0046] Based on Markov decision making, a global queue state space, a global queue action space and a global queue reward function are defined according to the upper-level optimization model to obtain an upper-level queue stable decision model;

[0047] The extended SCA algorithm is used to optimize and solve the upper queue stability decision model to obtain an unloading strategy, a transmission power allocation strategy and a decoding strategy;

[0048] Based on the Markov decision making, a local queue state space, a local queue action space and a local queue reward function are defined according to the lower optimization model to obtain a lower queue stable decision model;

[0049] The standard SCA algorithm is used to optimize and solve the lower queue stable decision model to obtain a computing resource allocation strategy, wherein the computing resource allocation strategy includes a user computing resource allocation strategy and a satellite computing resource allocation strategy;

[0050] The target resource allocation strategy includes the offloading strategy, the transmit power allocation strategy, the decoding strategy and the computing resource allocation strategy.

[0051] The second aspect of the present application provides a resource allocation strategy generation device based on satellite edge computing, which is applied to a satellite-ground hybrid edge computing system, including:

[0052] An offloading decision determination unit, used to determine the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy, and obtain an offloading decision parameter;

[0053] A transmission parameter calculation unit, used to calculate the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of satellite and user terminal signal transmission according to the unloading decision parameters;

[0054] A task processing and analysis unit, used to calculate the time slot computing task amount and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameters and the actual computing capacity, wherein the actual computing capacity includes the user computing capacity and the satellite computing capacity;

[0055] An energy consumption target construction unit, configured to calculate the total energy consumption of a time slot based on the data transmission energy consumption and the time slot calculation energy consumption, and to construct an average energy consumption target function according to the total energy consumption of the time slot;

[0056] A constraint condition construction unit, used to construct a plurality of decision constraint conditions according to the offloading decision parameter, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variable;

[0057] A model optimization construction unit, configured to construct an initial queue energy consumption optimization model by combining the average energy consumption objective function and the decision constraint condition, and then simplify the initial queue energy consumption optimization model by using a Lyapunov algorithm to obtain a target optimization model;

[0058] The model optimization solving unit is used to perform hierarchical decision solving on the target optimization model based on Markov decision and SCA algorithm to obtain the target resource allocation strategy.

[0059] A third aspect of the present application provides a resource allocation strategy generation device based on satellite edge computing, the device comprising a processor and a memory;

[0060] The memory is used to store program code and transmit the program code to the processor;

[0061] The processor is used to execute the resource allocation strategy generation method based on satellite edge computing described in the first aspect according to the instructions in the program code.

[0062] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0063] In the present application, a resource allocation strategy generation method based on satellite edge computing is provided, which is applied to a satellite-ground hybrid edge computing system, including: determining the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy to obtain the offloading decision parameters; calculating the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of the satellite and user terminal signal transmission according to the offloading decision parameters; calculating the time slot computing task volume and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameters and the actual computing power, and the actual computing power includes the user computing power and the satellite computing power; calculating the total time slot energy consumption based on the data transmission energy consumption and the time slot computing energy consumption, and constructing an average energy consumption objective function according to the total time slot energy consumption; constructing multiple decision constraints based on the offloading decision parameters, the actual computing power, the dynamic evolution state of the queue, the queue task discard amount, the data transmission power, and the decoding sorting variables; after constructing the initial queue energy consumption optimization model in combination with the average energy consumption objective function and the decision constraints, the Lyapunov algorithm is used to simplify the initial queue energy consumption optimization model to obtain the target optimization model; hierarchical decision solving is performed on the target optimization model based on the Markov decision and SCA algorithms to obtain the target resource allocation strategy.

[0064] The resource allocation strategy generation method based on satellite edge computing provided in this application can not only determine whether the computing task is left in the local computing or unloaded to the satellite for edge computing according to the offloading decision parameters, but also calculate some transmission parameters of the user terminal transmission signal, such as signal-to-noise ratio, etc.; this process fully considers the heterogeneous characteristics of the user terminal, so that the computing resource allocation operation based on this is more in line with the actual situation. Moreover, the total energy consumption of the time slot caused by task scheduling is calculated under the consideration of various performance parameters of the user terminal and the satellite, and the average energy consumption objective function is constructed based on this, taking into account the energy consumption optimization problem of the system; and the decision constraints are configured according to parameters such as the dynamic evolution state of the queue and the amount of task discards, which fully considers the optimization problem of queue stability; the decision constraints are configured according to the actual computing power, data transmission power and decoding sorting variables, which take into account the dynamic changes of the communication link and the time-varying characteristics of resource allocation; furthermore, the process of optimizing the model and then solving it using the Lyapunov algorithm can weigh multiple optimization objectives to achieve the optimal solution, and can also adapt to the calculation of data in various formats in the target optimization model; it can also optimize and solve multi-stage optimization problems such as dynamically changing link states and resource allocation to adapt to complex scene changes. The whole process comprehensively considers the influence between various situations, various parameters and multiple optimization goals of the edge computing system, and can obtain accurate and reliable resource allocation strategies with strong applicability. Therefore, this application can solve the technical problem that the existing technology does not comprehensively consider factors such as task scheduling, resource allocation, communication link status, user heterogeneity characteristics and various system performance indicators, resulting in the lack of accuracy, reliability and applicability of actual resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A schematic diagram of a flow chart of a method for generating a resource allocation strategy based on satellite edge computing provided in an embodiment of the present application;

[0066] Figure 2 A schematic diagram of the structure of a resource allocation strategy generation device based on satellite edge computing provided in an embodiment of the present application;

[0067] Figure 3 A schematic diagram of the structure of a satellite-ground hybrid edge computing system provided in an embodiment of the present application;

[0068] Figure 4 A schematic diagram of a satellite-to-ground transmission signal flow in a time slot provided in an embodiment of the present application;

[0069] Figure 5 A schematic diagram of the hierarchical information obtained after stratifying the model based on the Lyapunov method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0070] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0071] For easier understanding, see Figure 1 , the embodiment of the resource allocation strategy generation method based on satellite edge computing provided by the present application is applied to the satellite-ground hybrid edge computing system, including:

[0072] Step 101: Determine the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy, and obtain the offloading decision parameter.

[0073] Furthermore, step 101, before that, also includes:

[0074] Dynamic communication resource allocation strategy, task offloading strategy and computing resource allocation strategy are formulated according to the mobility of user terminals, task requirements and link status.

[0075] It should be noted that the resource allocation strategy generation method is applied to the satellite-ground hybrid edge computing system proposed in this embodiment. The system combines the resources of low-orbit satellite LEO and ground network to improve resource utilization and service quality through collaborative work. Figure 3 The system includes a low-orbit satellite LEO and N ground user terminal groups GU. The low-orbit satellite LEO is used as an onboard edge computing node, and an edge computing MEC server is deployed to process computing tasks and provide communication services; the ground user terminal groups orthogonally occupy N resource blocks NB respectively, and all users in the group share the same resource blocks. Due to the heterogeneity, mobility of user terminals and the randomness of task arrival, the system exhibits dynamic time-varying characteristics in communication link status and resource allocation.

[0076] The system provided in this embodiment works in a time slot operation mode, and the time is expressed as , the time slot length is ,The system workflow can be divided into four stages. The first stage :The system dynamically formulates communication resource allocation strategy, task offloading strategy and computing resource allocation strategy according to the mobility of user terminals, task requirements and link status. :The user terminal can transmit and process the computing tasks according to the communication resource allocation strategy and task offloading strategy formulated by the system. It can choose to transmit to the low-orbit satellite LEO, or keep it on the local device for direct calculation. This choice can generate the offloading decision parameters Phase 3 :The satellite LEO and user terminal allocate computing resources to each user according to the computing resource allocation strategy. :The satellite LEO returns the calculation results to the user terminal device. Since the volume of the calculation results is much smaller than the original mission data, the time of this stage can be ignored; and, , Are much smaller than and , so you don't have to consider The above uninstall decision parameters represents the parameter of the i-th user terminal in the n-th ground user terminal group; the number of ground user terminal groups is expressed as , each ground user terminal group includes multiple user terminals, the number of which is expressed as The user's computing task messages can be divided into Information flows are then transmitted. Offloading decision parameters Expressed as:

[0077]

[0078] Step 102: Calculate the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of satellite and user terminal signal transmission according to the offloading decision parameters.

[0079] Furthermore, step 102 includes:

[0080] Calculate the signal-to-noise ratio of each signal sent from the user terminal to the satellite according to the offloading decision parameter, the channel model and the data transmission power;

[0081] Calculate the decoding rate of the information stream based on the signal-to-noise ratio and the total system bandwidth;

[0082] Calculate the transmission data volume of the information flow according to the decoding rate and the task transmission time;

[0083] The time slot transmission energy consumption of the user terminal is calculated by combining the offloading decision parameters and the data transmission power to obtain the data transmission energy consumption.

[0084] Furthermore, step 102 further includes:

[0085] RSMA technology is used to send the information flow of the user terminal to the satellite for calculation based on the amount of data transmitted and the energy consumption of data transmission;

[0086] The satellite decodes the received information stream and allocates computing resources based on SIC technology and computing resource allocation strategy.

[0087] It should be noted that in order to alleviate the transmission interference between multiple different user terminals and improve the frequency utilization and system capacity, this embodiment selects the Rate-Splitting Multiple Access (RSMA) technology to realize the information transmission from the user to the satellite. The user terminal is configured with a single antenna and uses the RSMA technology to realize the communication at the physical layer. The information flow is segmented and decoded, and information is transmitted by sharing the same spectrum resources, thereby alleviating transmission interference between users and improving spectrum utilization.

[0088] See also Figure 4 , if the i-th user terminal in a group of n ground user terminals The transmitted signal is:

[0089]

[0090] in, Indicates user terminal The data transmission power allocated to information flow j is, is the sending signal of the corresponding information flow. If the signal received by the satellite LEO is expressed as:

[0091]

[0092] in, is additive white Gaussian noise, is the channel model, which can be expressed as:

[0093]

[0094] in, , are the antenna gains of satellite LEO and user terminal respectively, is the Rayleigh fading of a complex Gaussian variable, is the log-normally distributed shadow attenuation, is the unit path loss constant, is the path attenuation, is the path loss exponent, The distance from the user terminal to the satellite LEO for each time slot.

[0095] The satellite LEO uses serial interference cancellation SIC technology to decode the received signal, and then can allocate computing resources to the received computing tasks according to the computing resource allocation strategy. Specifically, the satellite LEO preferentially selects a certain information stream for decoding according to the specific decoding strategy, and regards other information streams to be decoded as interference; then removes the impact of the current information stream that has been decoded from the received signal to reduce interference with subsequent decoding; and then decodes the remaining information streams in turn. This decoding strategy can effectively improve the communication efficiency and decoding accuracy in multi-user terminal scenarios. The decoding strategy is determined by the decoding sorting variable, which is the decoding order, expressed as ,if , then it represents the signal flow wall Decode first, on this basis, the signal-to-noise ratio of each signal stream is expressed as :

[0096]

[0097] in, is the variance of Gaussian additive white noise, that is, the noise power, is the offloading decision parameter, indicating that the user chooses to leave the computing task to local computing or offload it to the satellite LEO. , it is left to local calculation, then the signal-to-noise ratio caused by communication is 0, that is, , at this time, the user will not interfere with the transmission operation of other user terminals, that is, the interference is 0, expressed as:

[0098]

[0099] if , the calculation task is offloaded to the satellite LEO calculation, then the signal-to-noise ratio of the signal needs to meet the following conditions for correct decoding:

[0100]

[0101] in, Represents the decoding threshold, which is the minimum signal-to-noise ratio (SINR) required for the system to successfully decode the signal. When the received signal-to-noise ratio SINR is greater than or equal to the threshold, the system can consider that the signal is correctly decoded.

[0102] Based on the above, the user terminal can be calculated based on the signal-to-noise ratio and the total system bandwidth Information flow Decoding rate:

[0103]

[0104] in, is the total bandwidth available to the system, that is, the total system bandwidth, is the number of ground user terminal groups, which is also the number of orthogonal shared resource blocks. ,So , then the decoding rate .

[0105] User Terminal The total data decoding rate can be expressed as:

[0106]

[0107] Then the total amount of transmitted data can be expressed as:

[0108]

[0109] in, It is the time when the user terminal sends the information flow to the satellite LEO, that is, the mission transmission time. In time slot The time slot transmission energy consumption, that is, the data transmission energy consumption can be expressed as:

[0110]

[0111] Step 103: Calculate the time slot computing task volume and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameters and the actual computing capacity, where the actual computing capacity includes the user computing capacity and the satellite computing capacity.

[0112] Further, step 103 includes:

[0113] Calculating the computing task amount of the user terminal according to the offloading decision parameter and the user computing capability to obtain the user computing task amount;

[0114] Calculate the computing energy consumption of the user terminal according to the offloading decision parameter, the user computing capability and the transmission computing time to obtain the user computing energy consumption;

[0115] The satellite computing task volume is calculated according to the satellite computing capability and the satellite processing density to obtain the satellite computing task volume;

[0116] The satellite computing energy consumption is calculated according to the satellite computing capability and the satellite computing time to obtain the satellite computing energy consumption;

[0117] The user computing workload and the satellite computing workload constitute the time slot computing workload;

[0118] User computing energy consumption and satellite computing energy consumption constitute the time slot computing energy consumption.

[0119] It should be noted that this embodiment takes into account both local computing and satellite edge computing, so the local computing mode and satellite edge computing mode should be divided into analysis during the calculation and analysis process. , then the user terminal The computing tasks are calculated locally, and the actual computing power allocated by the user is the user computing power. The user computing capacity of a time slot is expressed as , the unit is cycles / s, and the maximum value of the user's computing power is expressed as Then the user terminal In time slot The amount of user computing tasks can be expressed as:

[0120]

[0121] in, It is the processing density of the user's CPU, in cycles / bit. The decision transmission time of the information, is the user's local computing time, and the sum of the two represents the local processing transmission computing time. In time slot The user computing energy consumption is expressed as:

[0122]

[0123] in, For user terminals The CPU effective energy coefficient.

[0124] like , then the user terminal The calculation task is calculated on the satellite LEO, so at this time the satellite LEO is in the time slot The amount of satellite computing tasks processed is expressed as:

[0125]

[0126] in, Assigned to user terminals for satellite LEO The actual computing power, that is, the satellite computing power, can be expressed as , is the CPU processing density of satellite LEO, that is, satellite processing density. The satellite computing energy consumption is expressed as:

[0127]

[0128] in, is the CPU effective energy coefficient of the LEO satellite.

[0129] Step 104: Calculate the total energy consumption of the time slot based on the data transmission energy consumption and the time slot calculation energy consumption, and construct an average energy consumption objective function according to the total energy consumption of the time slot.

[0130] The time slot computing energy consumption includes the user computing energy consumption and satellite computing energy consumption , and the data transmission energy consumption is expressed as , so the user terminal In time slot The system time slot energy consumption is expressed as:

[0131]

[0132] Then the system is in the time slot The overall energy consumption of the time slot for the terrestrial user terminal group n is expressed as:

[0133]

[0134] Then the system is in the time slot The total energy consumption of the time slot can be expressed as:

[0135]

[0136] In order to be able to perform joint optimization based on energy consumption optimization in the future, this embodiment takes minimizing the long-term average total energy consumption of the system as the optimization goal, and then constructs the average energy consumption objective function:

[0137]

[0138] Step 105: construct multiple decision constraints based on the offloading decision parameters, actual computing capacity, queue dynamic evolution state, queue task discard amount, data transmission power, and decoding sorting variables.

[0139] Furthermore, step 105, before that, also includes:

[0140] Calculate the user dynamic evolution state of the user terminal queue according to the amount of transmitted data, the amount of time slot calculation tasks and the amount of time slot task arrival;

[0141] Calculate the user task discard quantity of the user terminal queue according to the user's dynamic evolution state, the time slot task arrival quantity and the user queue upper limit value;

[0142] Calculate the satellite dynamic evolution state of the satellite fleet according to the amount of transmitted data and the amount of time slot calculation tasks;

[0143] The satellite mission discard quantity of the satellite queue is calculated according to the satellite dynamic evolution state, the amount of transmitted data and the upper limit of the satellite queue;

[0144] The user dynamic evolution state and the satellite dynamic evolution state are the queue dynamic evolution state;

[0145] The user task discard amount and satellite task discard amount are the queue task discard amount.

[0146] It should be noted that in order to build an optimization model, it is also necessary to configure corresponding decision constraints for the average energy consumption objective function so as to meet the needs of joint optimization. Joint optimization needs to consider factors such as offloading decisions, decoding strategies, transmission power allocation and computing resources. Therefore, in order to ensure the stability of long-term task queues, improve computing efficiency and reduce system energy consumption at the same time, it is necessary to build decision constraints based on multiple parameters such as offloading decision parameters, actual computing power, dynamic evolution state of queues, amount of queue task discards, data transmission power, decoding sorting variables, etc.

[0147] Among the above factors, the dynamic evolution state of the queue and the amount of queue task discards can reflect the state of the system queue. Configuring decision constraints with queue parameters can enable the system optimization model to consider both queue stability and system energy consumption issues. The solution can be used to optimize the balance between the two and achieve the optimal state.

[0148] Both the user terminal and the satellite LEO are equipped with task queues to store the arriving computing tasks and track the dynamic status. The user-side task queue is mainly used to store the random computing tasks generated in each time slot. The computing tasks generated by the user at the beginning of each time slot. The computing tasks that reach the task queue need to satisfy independent and identical distribution (iid) and bounded second-order distance. Its distribution must satisfy:

[0149]

[0150] The user's task queue is represented as , where the user terminal The user dynamics evolution state is expressed as:

[0151]

[0152] in, is the number of time slot task arrivals for each time slot, Indicates user terminal The output task volume of the queue, It mainly includes two parts, namely the local user computing task volume and the amount of data transmitted and offloaded to the LEO satellite ; So we can get:

[0153]

[0154] User Terminal The upper limit of the user queue is recorded as , tasks that exceed the queue limit will be discarded, so the user task discard amount is expressed as:

[0155]

[0156] in, For user terminals Maximum task discard tolerance.

[0157] The satellite side allocates a management task queue for each user terminal, which can also be called a satellite task queue, specifically represented as , then the satellite LEO manages the user terminal The satellite fleet evolution state, that is, the satellite dynamic evolution state is expressed as:

[0158]

[0159] in, For user terminals The amount of information sent, that is, the amount of data transmitted, and the output task volume is the satellite computing task volume of the satellite task queue , so we know The upper limit of the satellite fleet managed by the satellite LEO is , if it exceeds the upper limit, it will be discarded, so the satellite mission discard amount is expressed as:

[0160]

[0161] in, is the maximum mission abandonment tolerance at the satellite LEO.

[0162] After the above calculations, decision constraints can be constructed based on multiple parameters such as offloading decision parameters, actual computing power, dynamic evolution state of queues, queue task discards, data transmission power, decoding sorting variables, etc. Specifically, the offloading decision parameters need to be limited to 0 / 1 binary variables; the actual computing power allocated on the user side and the satellite side needs to be no greater than their maximum computing power; the allocated data transmission power cannot exceed its maximum power; and the decoding sorting variables need to be in discrete form.

[0163] The constraints constructed based on the dynamic evolution state of the queue and the amount of queue task discards are mainly to ensure that the queue is strongly stable and the number of packet losses is not too large. Specifically, the queue length can be restricted from growing indefinitely based on the dynamic evolution state of the user and satellite queues. Under this constraint, each received data packet will be processed within a limited queuing delay, thereby achieving long-term stability of the data queue. This queue stability not only ensures that the backlog of each task queue is always kept within a limited range to avoid exceeding its specified capacity, but also ensures that the task can be processed within a limited time to meet the system quality of service (QoS) requirements. Based on the amount of queue task discards, the amount of packet loss data can be limited to not exceed a specified threshold.

[0164] Step 106: After constructing the initial queue energy consumption optimization model by combining the average energy consumption objective function and the decision constraints, the Lyapunov algorithm is used to simplify the initial queue energy consumption optimization model to obtain the target optimization model.

[0165] Further, step 106 includes:

[0166] The initial queue energy consumption optimization model is constructed by combining the average energy consumption objective function and decision constraints;

[0167] Define a virtual packet loss queue based on the queue task drop volume;

[0168] The Lyapunov algorithm is used to define the Lyapunov function according to the virtual packet loss queue and the queue dynamic evolution state;

[0169] The drift penalty term is defined according to the Lyapunov function, the total energy consumption of the time slot and the energy consumption weight parameter;

[0170] The drift penalty term is minimized and optimized through the defined queue stability cost function, and the initial queue energy consumption optimization model is simplified to the target optimization model.

[0171] Combining the above average energy consumption objective function and the configuration of decision constraints, the initial queue energy consumption optimization model can be expressed as:

[0172]

[0173] in, The maximum data transmission power that can be allocated.

[0174] Since the initial queue energy consumption optimization model contains binary variables and continuous variables, it is a long-term random mixed integer nonlinear programming problem, and is affected by the dynamics and randomness of the environment, the model problem will only become more complicated and difficult to solve. Therefore, this embodiment proposes to use the Lyapunov algorithm to simplify the initial queue energy consumption optimization model, or to convert the solution problem of the model into a series of deterministic optimization problems for each time slot, thereby reducing the complexity of the solution and enabling it to adapt to the dynamically changing environment.

[0175] It is still divided into local computing of user terminals and satellite edge computing. Virtual packet loss queues can be defined according to the discard amount of each queue task. Then the user terminal The user virtual packet loss queue at time slot t is expressed as:

[0176]

[0177] The satellite virtual packet loss queue of satellite LEO in time slot t is expressed as:

[0178]

[0179] These two virtual packet loss queues can be used to enforce packet loss limits on the user side and the satellite side.

[0180] The total queue can be expressed as:

[0181]

[0182] The queue in each ground user terminal group is expressed as:

[0183]

[0184] The user-side queue is expressed as:

[0185]

[0186] The satellite side queue is expressed as:

[0187]

[0188] The Lyapunov function can be defined as:

[0189]

[0190] The conditional Lyapunov drift can be written as:

[0191]

[0192] Assume that the current system queue state in the time slot is , then the drift is the expected change of the function over a time slot, so the Lyapunov drift The drift penalty term can be obtained by weighted calculation:

[0193]

[0194] in, It is a weight parameter that emphasizes the importance of system energy consumption, recorded as energy consumption weight parameter, which can be used to control the balance between system energy consumption and queue stability.

[0195] The upper bound on the right side of the drift penalty term can be optimized:

[0196]

[0197] in, is a constant, expressed as:

[0198]

[0199] in, , for the nth group of user terminals, the user-side queue constant is:

[0200]

[0201] The satellite side queue constants are:

[0202]

[0203] The queue stability cost function is defined as:

[0204]

[0205] in,

[0206]

[0207] The queue stability cost function on the user side is:

[0208]

[0209] The satellite side queue stabilization cost function is:

[0210]

[0211] By minimizing the drift penalty, we can effectively control system congestion and stabilize the queue state, thereby reducing the amount of tasks waiting in the buffer, helping to maintain a low level of task backlog and minimize task delays. After the above optimization, the solution of the initial queue energy consumption optimization model is converted into the solution of the following target optimization model:

[0212]

[0213] In order to get closer to the optimal solution of the initial queue energy consumption optimization model, the model needs more time to meet the average energy consumption constraint; at the same time, the larger the V value, the more significant the system's ability to achieve lower average energy consumption. The target optimization model is suitable for dynamic optimization of each time slot, and does not require prior knowledge of environmental information, nor does it rely on the probability distribution of random events.

[0214] Step 107: Perform hierarchical decision solving on the target optimization model based on Markov decision and SCA algorithm to obtain the target resource allocation strategy.

[0215] Further, step 107 includes:

[0216] According to the hierarchical reinforcement learning method guided by Lyapunov algorithm, the target optimization model is divided into an upper optimization model and a lower optimization model;

[0217] Based on Markov decision making and the upper optimization model, the global queue state space, the global queue action space and the global queue reward function are defined to obtain the upper queue stable decision model;

[0218] The extended SCA algorithm is used to optimize and solve the upper-layer queue stability decision model, and the unloading strategy, transmission power allocation strategy and decoding strategy are obtained;

[0219] Based on Markov decision making, the local queue state space, local queue action space and local queue reward function are defined according to the lower-level optimization model to obtain a stable decision model for the lower-level queue;

[0220] The standard SCA algorithm is used to optimize and solve the lower-level queue stability decision model to obtain the computing resource allocation strategy, which includes the user computing resource allocation strategy and the satellite computing resource allocation strategy.

[0221] The target resource allocation strategy includes offloading strategy, transmission power allocation strategy, decoding strategy and computing resource allocation strategy.

[0222] It should be noted that since the target optimization model belongs to the complex convex optimization and numerical optimization, and cannot be solved directly, this embodiment chooses to model based on Markov decision making. Before modeling, in order to consider the coupling between variables and the characteristics of phased optimization, this embodiment combines the hierarchical reinforcement learning method guided by Lyapunov to hierarchically process the target optimization model. Please refer to Figure 5 ,The upper optimization model is responsible for the optimization of offloading decisions, transmit power allocation and decoding decisions, while the lower optimization model is responsible for the optimization of computing resource allocation on the user side and the satellite side.

[0223] Markov decision can be used to optimize the upper and lower layers respectively. Although Markov decision modeling includes four basic elements, namely, state, action, state transition probability and reward, the solution model of this embodiment does not require prior knowledge of the environmental state and does not rely on the probability distribution of random events, so the state transition probability is unknown.

[0224] The upper-level decision modeling needs to obtain global queue information, optimize the stability of the global queue by controlling the amount of data sent from the user-side queue to the satellite-side queue, and adjust the stability values ​​of the queues on both sides. Therefore, the global queue state space can be defined:

[0225]

[0226] That is, the global real-time queue status , User task arrival volume and channel status The global queue action space can be expressed as:

[0227]

[0228] Uninstallation decision parameters , decode sort variables and data transmission power The global queue reward function can be expressed as:

[0229]

[0230] in, is the penalty for exceeding and violating the constraint, , , , is the weight coefficient used to balance the order of magnitude.

[0231] Since the action space of the upper layer contains binary data, discrete variables and continuous variables, this embodiment adopts an extended SCA algorithm to solve the constructed Markov model, wherein the Bernoulli distribution is used to model and sample the binary data to ensure that the gradient does not disappear; the Gumbel-Softmax technique is used to process discrete variables so that their output action gradients can be back-propagated; the continuous variables are processed by the re-parameterization method to ensure the continuity of the gradient during the sampling process; finally, the optimal solution of the upper-layer queue stability decision model can be solved, which specifically includes the unloading strategy, the transmission power allocation strategy and the decoding strategy.

[0232] Since the lower-layer queue stability decision model is to optimize the computing resource allocation on the user side and the satellite side, the modeling process of the lower-layer queue stability decision model needs to be divided into user-side modeling and satellite-side modeling. For user-side modeling, it is mainly to obtain user queue information and adjust the computing output of the user queue to optimize the queue stability and energy consumption on the user side. The local queue state space on the user side is defined as:

[0233]

[0234] That is, the real-time queue status from the user side and user task arrival constitute.

[0235] The user-side local queue action space is defined as:

[0236]

[0237] Calculate resource allocation for the user side.

[0238] The user-side local queue reward function is defined as:

[0239]

[0240] Since the action space on the user side only contains continuous variables, this embodiment directly uses the standard SCA algorithm to perform optimization and solve to obtain the user computing resource allocation strategy.

[0241] For satellite-side modeling, the main purpose is to obtain satellite fleet information and adjust the calculation output of the satellite fleet to optimize the fleet stability and energy consumption on the satellite side. Therefore, the local fleet state space on the satellite side is defined as:

[0242]

[0243] That is, the real-time queue status from the satellite side and the amount of data arriving at the satellite constitute.

[0244] The satellite-side local queue action space is defined as:

[0245]

[0246] Calculate resource allocations for satellites.

[0247] The satellite-side local fleet reward function is defined as:

[0248]

[0249] Similarly, the action space on the satellite side only contains continuous variables, so the standard SCA algorithm is directly sampled for optimization and solution to obtain the satellite computing resource allocation strategy.

[0250] The solution provided in this embodiment can dynamically adjust the allocation of computing resources based on real-time information to achieve stability and energy efficiency of the user-side and satellite-side queues. The system can make dynamic decisions based on real-time information. The lower-layer network optimizes the computing output and energy consumption of the queues to achieve queue stability and energy consumption optimization on the user side and satellite side respectively; the upper-layer network dynamically balances the queue status on both sides by adjusting the transmission volume from the user-side queue to the satellite-side queue. Through the collaborative iterative optimization of the upper and lower-layer networks, the system can gradually converge to the global optimal strategy, thereby achieving efficient operation in a complex dynamic environment.

[0251] The resource allocation strategy generation method based on satellite edge computing provided in the embodiment of the present application can not only determine whether the computing task is left in the local computing or unloaded to the satellite for edge computing according to the offloading decision parameters, but also calculate some transmission parameters of the user terminal transmission signal, such as signal-to-noise ratio, etc.; this process fully considers the heterogeneous characteristics of the user terminal, so that the computing resource allocation operation based on this is more in line with the actual situation. Moreover, the total energy consumption of the time slot caused by the task scheduling is calculated under the consideration of various performance parameters of the user terminal and the satellite, and the average energy consumption objective function is constructed based on this, taking into account the energy consumption optimization problem of the system; and the decision constraints are configured according to parameters such as the dynamic evolution state of the queue and the amount of task discards, which fully considers the optimization problem of the queue stability; the decision constraints are configured according to the actual computing power, data transmission power and decoding sorting variables, which take into account the dynamic changes of the communication link and the time-varying characteristics of resource allocation; furthermore, the process of optimizing the model and then solving it using the Lyapunov algorithm can weigh multiple optimization objectives to achieve the optimal solution, and can also adapt to the calculation of data in various formats in the target optimization model; it can also optimize and solve multi-stage optimization problems such as dynamically changing link states and resource allocation to adapt to complex scene changes. The whole process comprehensively considers the influence between various situations, various types of parameters and multiple optimization goals of the edge computing system, and can obtain accurate and reliable resource allocation strategies with strong applicability. Therefore, the embodiments of the present application can solve the technical problem that the prior art does not comprehensively consider factors such as task scheduling, resource allocation, communication link status, user heterogeneity characteristics and various index performance of the system, resulting in the lack of accuracy, reliability and applicability of actual resource allocation.

[0252] For easier understanding, see Figure 2 The present application provides an embodiment of a resource allocation strategy generation device based on satellite edge computing, and the resource allocation strategy generation device based on satellite edge computing is characterized in that it is applied to a satellite-ground hybrid edge computing system, including:

[0253] The offloading decision determination unit 201 is used to determine the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy, and obtain the offloading decision parameter;

[0254] The transmission parameter calculation unit 202 is used to calculate the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of the satellite and user terminal signal transmission according to the unloading decision parameters;

[0255] The task processing analysis unit 203 is used to calculate the time slot computing task amount and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameter and the actual computing capacity, where the actual computing capacity includes the user computing capacity and the satellite computing capacity;

[0256] The energy consumption target construction unit 204 is used to calculate the total energy consumption of the time slot based on the data transmission energy consumption and the time slot calculation energy consumption, and to construct an average energy consumption target function according to the total energy consumption of the time slot;

[0257] The constraint condition construction unit 205 is used to construct multiple decision constraint conditions according to the offloading decision parameters, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variables;

[0258] The model optimization construction unit 206 is used to construct an initial queue energy consumption optimization model by combining the average energy consumption objective function and the decision constraint condition, and then simplify the initial queue energy consumption optimization model by using the Lyapunov algorithm to obtain a target optimization model;

[0259] The model optimization solving unit 207 is used to perform hierarchical decision solving on the target optimization model based on Markov decision and SCA algorithm to obtain the target resource allocation strategy.

[0260] It can be understood that the specific operation process of the above-described devices and units can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0261] The present application also provides a resource allocation strategy generation device based on satellite edge computing, the device including a processor and a memory;

[0262] The memory is used to store the program code and transmit the program code to the processor;

[0263] The processor is used to execute the resource allocation strategy generation method based on satellite edge computing in the above method embodiment according to the instructions in the program code.

[0264] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0265] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0266] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0267] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions for executing all or part of the steps of the method described in each embodiment of the present application through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0268] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A resource allocation strategy generation method based on satellite edge computing, characterized in that: Applied to satellite-ground hybrid edge computing systems, including: Determine the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy, and obtain the offloading decision parameter; Calculate the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of satellite and user terminal signal transmission according to the offloading decision parameters; Calculating the time slot computing task amount and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameter and the actual computing capacity, wherein the actual computing capacity includes the user computing capacity and the satellite computing capacity; Calculate the total energy consumption of the time slot based on the data transmission energy consumption and the time slot calculation energy consumption, and construct an average energy consumption objective function according to the total energy consumption of the time slot; Constructing a plurality of decision constraints according to the offloading decision parameters, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variables; After constructing an initial queue energy consumption optimization model by combining the average energy consumption objective function and the decision constraint condition, the initial queue energy consumption optimization model is simplified by using the Lyapunov algorithm to obtain a target optimization model; The target optimization model is solved by hierarchical decision making based on Markov decision making and SCA algorithm to obtain the target resource allocation strategy.

2. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: The method of determining the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy to obtain the offloading decision parameter also includes: Dynamic communication resource allocation strategy, task offloading strategy and computing resource allocation strategy are formulated according to the mobility of user terminals, task requirements and link status.

3. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: The calculating, according to the offloading decision parameters, the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of the satellite and user terminal signal transmission comprises: Calculate the signal-to-noise ratio of each signal sent from the user terminal to the satellite according to the offloading decision parameter, the channel model and the data transmission power; Calculating a decoding rate of the information stream according to the signal-to-noise ratio and the total bandwidth of the system; Calculate the transmission data volume of the information flow according to the decoding rate and the task transmission time; The time slot transmission energy consumption of the user terminal is calculated in combination with the offloading decision parameter and the data transmission power to obtain the data transmission energy consumption.

4. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: The step of calculating the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of satellite and user terminal signal transmission according to the unloading decision parameters also includes: The RSMA technology is used to send the information flow of the user terminal to the satellite for calculation according to the transmission data volume and the data transmission energy consumption; The satellite decodes the received information stream and allocates computing resources based on SIC technology and computing resource allocation strategy.

5. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: The calculating the time slot computing task amount and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameter and the actual computing capacity includes: Calculating the computing task amount of the user terminal according to the offloading decision parameter and the user computing capability to obtain the user computing task amount; Calculating the computing energy consumption of the user terminal according to the offloading decision parameter, the user computing capability and the transmission computing time to obtain the user computing energy consumption; Calculating the satellite computing task amount according to the satellite computing capability and the satellite processing density to obtain the satellite computing task amount; Calculating the satellite computing energy consumption according to the satellite computing capability and the satellite computing time to obtain the satellite computing energy consumption; The user computing task amount and the satellite computing task amount constitute a time slot computing task amount; The user computing energy consumption and the satellite computing energy consumption constitute the time slot computing energy consumption.

6. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: The method of constructing a plurality of decision constraints based on the offloading decision parameters, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variables also includes: Calculate the user dynamic evolution state of the user terminal queue according to the transmission data volume, the time slot calculation task volume and the time slot task arrival volume; Calculating the user task discarding amount of the user terminal queue according to the user dynamic evolution state, the time slot task arrival amount and the user queue upper limit value; Calculating the satellite dynamic evolution state of the satellite queue according to the transmission data volume and the time slot calculation task volume; Calculating the satellite mission discard amount of the satellite queue according to the satellite dynamic evolution state, the transmission data amount and the satellite queue upper limit value; The user dynamic evolution state and the satellite dynamic evolution state are queue dynamic evolution states; The user task discard amount and the satellite task discard amount are queue task discard amounts.

7. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: After the initial queue energy consumption optimization model is constructed by combining the average energy consumption objective function and the decision constraint condition, the initial queue energy consumption optimization model is simplified by using the Lyapunov algorithm to obtain a target optimization model, including: Constructing an initial queue energy consumption optimization model in combination with the average energy consumption objective function and the decision constraints; Defining a virtual packet loss queue according to the queue task loss amount; A Lyapunov function is defined according to the virtual packet loss queue and the dynamic evolution state of the queue using a Lyapunov algorithm; Defining a drift penalty term according to the Lyapunov function, the total energy consumption of the time slot and the energy consumption weight parameter; The drift penalty term is minimized and optimized through a defined queue stability cost function, and the initial queue energy consumption optimization model is simplified into a target optimization model.

8. The method for generating a resource allocation strategy based on satellite edge computing according to claim 1, characterized in that: The target optimization model is subjected to hierarchical decision solving based on Markov decision and SCA algorithm to obtain the target resource allocation strategy, including: Dividing the target optimization model into an upper optimization model and a lower optimization model according to the hierarchical reinforcement learning method guided by the Lyapunov algorithm; Based on Markov decision making, a global queue state space, a global queue action space and a global queue reward function are defined according to the upper-level optimization model to obtain an upper-level queue stable decision model; The extended SCA algorithm is used to optimize and solve the upper queue stability decision model to obtain an unloading strategy, a transmission power allocation strategy and a decoding strategy; Based on the Markov decision making, a local queue state space, a local queue action space and a local queue reward function are defined according to the lower optimization model to obtain a lower queue stable decision model; The standard SCA algorithm is used to optimize and solve the lower queue stable decision model to obtain a computing resource allocation strategy, wherein the computing resource allocation strategy includes a user computing resource allocation strategy and a satellite computing resource allocation strategy; The target resource allocation strategy includes the offloading strategy, the transmit power allocation strategy, the decoding strategy and the computing resource allocation strategy.

9. A resource allocation strategy generation device based on satellite edge computing, characterized in that: Applied to satellite-ground hybrid edge computing systems, including: An offloading decision determination unit, used to determine the decision of the user terminal computing task based on the communication resource allocation strategy and the task offloading strategy, and obtain an offloading decision parameter; A transmission parameter calculation unit, used to calculate the signal-to-noise ratio, decoding rate, transmission data volume and data transmission energy consumption of satellite and user terminal signal transmission according to the unloading decision parameters; A task processing and analysis unit, used to calculate the time slot computing task amount and time slot computing energy consumption of the user terminal and the satellite respectively according to the offloading decision parameters and the actual computing capacity, wherein the actual computing capacity includes the user computing capacity and the satellite computing capacity; An energy consumption target construction unit, configured to calculate the total energy consumption of a time slot based on the data transmission energy consumption and the time slot calculation energy consumption, and to construct an average energy consumption target function according to the total energy consumption of the time slot; A constraint condition construction unit, used to construct a plurality of decision constraint conditions according to the offloading decision parameter, the actual computing capacity, the dynamic evolution state of the queue, the amount of queue task discards, the data transmission power, and the decoding sorting variable; A model optimization construction unit, configured to construct an initial queue energy consumption optimization model by combining the average energy consumption objective function and the decision constraint condition, and then simplify the initial queue energy consumption optimization model by using a Lyapunov algorithm to obtain a target optimization model; The model optimization solving unit is used to perform hierarchical decision solving on the target optimization model based on Markov decision and SCA algorithm to obtain the target resource allocation strategy.

10. A resource allocation strategy generation device based on satellite edge computing, characterized in that: The device comprises a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the resource allocation strategy generation method based on satellite edge computing as described in any one of claims 1-8 according to the instructions in the program code.

Images