Multi-dimensional service quality driven FSO / RF air-space-ground integrated robust resource allocation method

Abstract

The invention relates to a multi-dimensional service quality driven FSO / RF air-space-ground integrated robust resource allocation method, and belongs to the technical field of wireless communication networks. The problems that an existing FSO / RF SAGIN cloud edge cooperative computing method cannot give consideration to high efficiency and timeliness of computing task processing and reliability of computing task transmission and needs to depend on an accurate CSI condition for computing task processing are solved. The method comprises the following steps: firstly, establishing an FSO / RF SAGIN cloud edge cooperative computing network model, constructing an optimization problem OP1 based on the model, and solving the optimization problem OP1 by adopting a DDPG deep learning method to obtain an optimal solution, namely an FSO / RF space-air-ground integrated robust resource allocation scheme, by taking the purpose of maximizing the sum of long-time computing task handling capacities of all users as a target. According to the method disclosed by the invention, the realization of high-service-quality FSO / RF air-space-ground integrated cloud edge cooperative computing can be ensured from multiple dimensions under the condition of uncertain channels.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication networks, and particularly to a multi-dimensional quality-of-service-driven robust resource allocation method for FSO / RF space-air-ground integrated networks. Background Art

[0002] The future 6G network will not only provide seamless global connectivity but also integrate a vast amount of computing resources to support ubiquitous device interconnection and real-time data processing, thus promoting the development of emerging fields such as smart cities, telemedicine, and smart agriculture. The space-air-ground integrated network (SAGIN) empowered by cloud-edge collaborative computing plays an important role in the 6G network because it can meet these requirements. However, currently, SAGIN mainly uses radio frequency (RF) transmission links, which face the problems of scarce spectrum resources and difficulty in securely transmitting computing tasks between satellites and unmanned aerial vehicles. Therefore, applying free space optical (FSO) communication technology with ultra-wideband spectrum resources and high confidentiality to SAGIN cloud-edge collaborative computing, namely FSO / RF SAGIN cloud-edge collaborative computing technology, is an effective solution.

[0003] Achieving efficient, timely computing task processing and reliable computing task transmission are the basic requirements for FSO / RF SAGIN cloud-edge collaborative computing. However, only one of these aspects has been studied in the prior art. For example, only improving the processing efficiency of computing tasks is considered while ignoring the reliability and timeliness of task transmission, or only reliable computing task transmission is considered while ignoring the efficiency and timeliness of task processing. In addition, the realization of reliable, efficient, and timely FSO / RF SAGIN cloud-edge collaborative computing requires channel state information (CSI) of the transmission channel. Existing FSO / RF SAGIN cloud-edge collaborative computing methods are all analyzed under the assumption of known accurate CSI. However, due to the influence of channel estimation error and quantization error, the channel has uncertainty, and it is difficult to obtain accurate CSI information in practical applications. Therefore, the prior art cannot achieve reliable, efficient, and timely FSO / RF SAGIN computing task processing under uncertain CSI conditions. Summary of the Invention

[0004] To solve the problem that the existing FSO / RF SAGIN cloud-edge collaborative computing method cannot balance the efficiency, timeliness of computing task processing, the reliability of computing task transmission, and the need to rely on accurate CSI conditions for computing task processing, the present invention proposes a multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method, which can ensure the realization of high-quality FSO / RF space-air-ground integrated cloud-edge collaborative computing from multiple dimensions (reliable, efficient, timely) under uncertain channel conditions.

[0005] To solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method, which includes the following steps:

[0007] Step 1: Establish an FSO / RF SAGIN cloud-edge collaborative computing network model, which includes satellites, unmanned aerial vehicles (UAVs), cloud servers, and ground Internet of Things (IoT) users. The satellite is used to provide comprehensive services covering the area of interest. The UAV, as an edge server, is equipped with a caching device and a computing device, and is used to provide edge computing and task caching for ground IoT users. Each ground IoT user generates delay-sensitive computing tasks and offloads the computing tasks that are not placed in the user's computing cache queue to the UAV through the RF link. The UAV uses the FSO link to forward the computing tasks that are not placed in the UAV's computing cache queue to the cloud server through the satellite, where both the FSO link and the RF link adopt adaptive modulation and coding methods;

[0008] Step 2: Based on the model established in Step 1, construct an optimization problem OP1. The optimization problem OP1 is: under the constraints of the computing task transmission error frame rate and processing delay, and at the same time under the condition of uncertain CSI information, by optimizing the allocation of the bandwidth allocated by the UAV to each ground IoT user, the offloading location of the computing task, the UAV flight trajectory, and the modulation and coding format, achieve the goal of maximizing the total long-term computing task processing volume of all users;

[0009] Step 3: Use the DDPG deep learning method to solve the optimization problem OP1 to obtain the optimal solution, which is the FSO / RF space-air-ground integrated robust resource allocation scheme.

[0010] The method provided by the present invention can achieve high-quality FSO / RF SAGIN cloud-edge collaborative computing from multiple dimensions, and the method has the following beneficial effects:

[0011] (1) The present invention can simultaneously achieve reliable computing task transmission, the computing task can be executed within its allowed maximum delay, and the execution efficiency of the computing task is high;

[0012] (2) When optimizing the network parameters, the present invention takes into account the uncertainty of CSI information. Therefore, it can adjust the network parameters under the condition of uncertain CSI information, meeting the requirements of the actual communication environment;

[0013] (3) Since the ultra-long-distance mission transmission between the satellite and the UAV adopts the FSO communication method with the advantages of large capacity and secure transmission, the present invention can achieve the transmission of large-scale computing tasks and can keep the task information that does not want to be leaked confidential. Description of the Drawings

[0014] Figure 1 is a flowchart of the multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method described in the present invention;

[0015] Figure 2 is a schematic diagram of the FSO / RF SAGIN cloud-edge collaborative computing network model;

[0016] Figure 3 is an Actor-Critic network architecture diagram for computing task offloading, bandwidth allocation, and UAV flight trajectory optimization based on DDPG. Detailed Embodiments

[0017] The technical solution of the present invention will be described in detail below in conjunction with the drawings and preferred embodiments.

[0018] See Figure 1 , this embodiment provides a multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method, which specifically includes the following steps 1 to 3.

[0019] Step 1: Establish an FSO / RF SAGIN cloud-edge collaborative computing network model, as Figure 2 shown. This model includes: a satellite, which can provide comprehensive services covering the area of interest and connect the Internet of Things devices with the cloud server through the satellite backbone network; a UAV equipped with a caching device and a computing device, which serves as an edge server to provide edge computing and task caching for ground Internet of Things users (IOT users). The number of operations it can provide per second is ; a cloud server, and the number of operations the cloud server can provide per second is ; ground Internet of Things users with computing capabilities, and the set of ground Internet of Things users is represented as, , and the number of operations each user can provide per second is .

[0020] Each terrestrial Internet of Things (IoT) user generates latency-sensitive computing tasks and makes real-time decisions on whether to process the generated computing tasks locally. If the terrestrial IoT user can complete the computing task, the task is placed in the user's computing cache queue and waits to be executed. Otherwise, the computing task is sent to the user's transmission cache queue and waits to be offloaded to the drone edge server for execution. In this embodiment, the terrestrial IoT user uses an RF link to offload the computing tasks in the user's transmission cache queue to the drone. Then, the drone collects computing tasks from terrestrial IoT users and decides whether the collected computing tasks are to be executed locally or offloaded to the cloud server for execution. If executed locally, the computing task is placed in the drone's computing cache queue and waits to be executed. Otherwise, the computing task is sent to the drone's transmission cache queue and waits to be offloaded. In this embodiment, the drone uses an FSO link to forward the computing tasks in the drone's transmission cache queue to the cloud server via a satellite. This embodiment considers that the FSO link between the satellite and the cloud server can provide sufficient bandwidth resources and can forward the computing tasks to the cloud server in a timely manner. Therefore, no cache queue is set in the satellite. In addition, this embodiment also considers that the cloud server has sufficient computing devices, so no cache queue is set in the cloud server either.

[0021] In this embodiment, the modulation method and channel coding rate that can achieve reliable and efficient task transmission are adaptively selected according to the estimated CSI of the transmission link for task transmission. In this embodiment, the modulation method uses m RF -QAM, and the channel coding rate is , , where gives the selection range of the modulation and coding parameters of the RF link. The FSO links between the drone and the satellite and between the satellite and the cloud server use adaptive quadrature amplitude modulation m FSO -QAM, , gives the selection range of the modulation and coding parameters of the FSO link.

[0022] The present invention uses a time-slot analysis model to divide the total optimization time period into time slots, and the duration of each time slot is , and its set is represented as . The computing tasks generated by each terrestrial IoT user in each time slot are , . Among them, is the number of bits of the computing task data generated by the user, which follows a Poisson distribution with a parameter of ; is the number of operations required to execute each bit of data; is the maximum latency allowed for executing the task. The length of the computing cache queue of each user is , and the length of the transmission cache queue of the user is , where is an indication variable of whether the computing task is executed at the user . When is 1, it means that the task is executed at the user. When , it means taking the inverse of , that is, offloading to the UAV for execution; is the number of operations required to execute 1 bit of data; is the RF link task transmission rate between the user and the UAV; is the time of one time slot; represents the floor symbol. The length of the computing cache queue of the UAV in each time slot is , is an indication variable of whether the computing task is offloaded to the UAV edge server for execution, . When is 1, it means that the task is executed at the cloud server. The length of the transmission cache queue of the UAV

[0023] ;

[0024] is an indication variable of whether the computing task is offloaded to the cloud server for execution, . When is 1, it means that the task is executed at the cloud server; is the FSO link transmission rate between the UAV and the satellite;

[0025] Step 2: According to the FSO / RF SAGIN cloud-edge collaborative computing network model constructed in Step 1, in order to simultaneously achieve reliable task transmission and efficient and timely computing task processing under uncertain CSI conditions, the present invention, under the constraints of the computing task transmission error frame rate and processing latency, and also under the condition of uncertain CSI information, realizes the goal of maximizing the long-term computing task processing volume by optimizing and allocating the bandwidth allocated by the UAV to each user, the offloading location of the computing task (user, edge server, cloud server), the UAV flight trajectory, and the modulation and coding format. The optimization problem constructed in this step is:

[0026] OP1: (1)

[0027]

[0028]

[0029] In this problem optimization objective, is the amount of computing tasks executed by the t-time slot cloud server, is the transmission rate. Different links have different transmission rates. is the amount of computing tasks executed by the t-time slot edge server, is the t-time slot user 's amount of computing tasks executed. , , is the th modulation order and channel coding rate adopted by the user. and are the modulation orders of the FSO links between the UAV and the satellite, and between the satellite and the cloud server respectively. In the network model, the computing task sender selects the appropriate modulation and coding format and channel coding rate according to the uncertain CSI information. In addition, the present invention considers that as long as the user can process the newly generated computing tasks in time, the task is preferably executed at the user end (determined by whether the waiting delay in the computing cache queue exceeds the maximum delay of the task). Therefore, in the above optimization problem OP1, the decision on whether the computing task is executed locally is not optimized.

[0030] In the constraint conditions of the optimization problem, C1, C2, and C3 are computing task transmission reliability constraints, which respectively require that the maximum computing task transmission error frame rates of the RF link and the FSO links between the UAV and the satellite, and between the satellite and the cloud server be less than the threshold . is the task transmission error frame rate between the user and the UAV, is the task transmission error frame rate between the UAV and the satellite, is the task transmission error frame rate between the satellite and the cloud server. The uncertain ranges of their estimated CSI are , and , , and are the signal-to-noise ratios of the links between the user and the UAV, between the UAV and the satellite, and between the satellite and the cloud server respectively. C4 requires that the computing task processing delay of the user shall not exceed its allowed maximum processing delay ; C5 requires that the amount of computing tasks offloaded by the user to the UAV edge server shall not exceed the amount of tasks that the RF link can transmit ; C6 requires that the total bandwidth allocated to the user shall not exceed the maximum bandwidth B provided by the UAV, where is the bandwidth of the user ; C7 and C8 are the constraints on the flight angle and rate of the UAV respectively, where, is the maximum flight speed of the UAV; C9 constrains the flight area of the UAV, is the position coordinate of the UAV; C10 and C11 constrain the selection range of adaptive modulation and coding parameters; C12 - C14 constrain the computing offloading strategy, requiring that the task can only be offloaded to one of the cloud server or the edge server, is the offloading indication vector of the edge server, is the offloading indication vector of the cloud server.

[0031] The expression of the frame error rate is the premise for solving this optimization problem. The expressions of each parameter will be given below.

[0032] The user and the frame error rate of the RF link calculation task transmission between the user and the UAV is:

[0033] (2)

[0034] where, , and are fitting parameters, is the average mutual information amount of the modulation and coding scheme, and its calculation method can be found in "IEEE Journal on selected areas in communications, 2008, 26(8): 1599 - 1606.". is the user and the signal - to - noise ratio of the RF transmission link between the user and the UAV, is the transmission power of the user, is the received noise power of the UAV, is the RF channel attenuation, which is jointly determined by the fixed attenuation and the random attenuation . The random attenuation obeys the Nakagami - m distribution, and the fixed attenuation can be expressed as:

[0035] (3)

[0036] where, is the position vector of the user , H is the height of the UAV, It is the channel attenuation for transmitting 1 m.

[0037] The frame error rate of the FSO link calculation task transmission is:

[0038] (4)

[0039] When calculating the frame error rate of the FSO link between the UAV and the satellite , in formula (4), takes the value of ; when calculating the frame error rate of the FSO link between the satellite and the cloud server , in formula (4), takes the value of . Among them, , and are fitting parameters, M is the modulation order of the FSO link, is the signal-to-noise ratio of the FSO link. is the transmit power, is the received noise. is the FSO link attenuation, which is jointly determined by the fixed attenuation and the random attenuation . The random attenuation obeys the Log-normal distribution, and the fixed attenuation is determined by the transmit antenna gain , the receive antenna gain , the free space attenuation , the atmospheric attenuation and the transmission margin .

[0040] The transmission rate of the calculation task of the RF link between the user and the UAV is , is the baud rate of the RF link task data transmission, and are the adopted QAM modulation order and channel coding rate. The transmission rate of the FSO link calculation task between the UAV and the satellite is , is the baud rate of the FSO link task data transmission between the UAV and the satellite. The transmission rate of the FSO link calculation task between the satellite and the cloud server is , is the baud rate of the FSO link task data transmission between the UAV and the satellite.

[0041] The processing delay of the calculation task of the user is:

[0042] (5)

[0043] Among them, is the computing waiting delay in the edge server, is the transmission waiting delay of the user in, is the transmission waiting delay in the UAV.

[0044] Step 3: The optimization problem OP1 can be transformed into the following four sub-optimization problems:

[0045]

[0046]

[0047] (6)

[0048]

[0049]

[0050] The essence of OP1_1~OP1_3 is to adaptively adjust the transmission modulation and coding format to maximize the transmission rate under the condition that the maximum frame error rate does not exceed a specific threshold . Their solutions are actually to find the signal-to-noise ratio range corresponding to each transmission format, that is, the switching threshold of the transmission format. Taking the solution of the switching threshold of the transmission format in OP1_1 as an example for detailed description, where is the number of elements in the set . Since the frame error rate is inversely proportional to the signal-to-noise ratio, the maximum frame error rate corresponds to the minimum signal-to-noise ratio in . Therefore, by setting , the expression of can be deduced. Similarly, the expressions of the switching thresholds of the transmission formats in OP1_2 and OP1_3 ( and ) can be obtained.

[0051] Substituting the solution results of the optimization problems OP1_1~OP1_3 into OP1_4, OP1_4 can be transformed into:

[0052] OP2: (7)

[0053]

[0054]

[0055]

[0056] Among them, is the transmission rate of the user's computing tasks over the RF link between the user and the UAV when adaptive transmission is adopted; is the processing delay of the user's computing tasks under the adaptive transmission mode of the system.

[0057] Step 4: Use the DDPG (Deep Deterministic Policy Gradient) deep learning method to solve the optimization problem OP2. This first requires converting OP2 into a Markov process, which needs to construct a state space, an action space, and a reward function.

[0058] The state space includes the position vector of the UAV , the estimated CSI vectors of RF links, the FSO link CSI between the UAV and the satellite is , the FSO link CSI between the satellite and the cloud server is , the length of the user's transmission buffer queue , the length of the user's computing buffer queue , the length of the UAV's transmission buffer queue , the length of the UAV's computing buffer queue , the remaining average time of the computing tasks in the user's computing buffer queue , where represents the remaining average time of the user in the computing buffer queue of the th user, and the remaining average time of the computing tasks in the user's transmission buffer queue , where represents the remaining average time of the task in the transmission buffer queue of the th user, the remaining average time of the computing tasks in the UAV's transmission buffer queue , and the remaining average time of the computing tasks in the UAV's computing buffer queue . The state space constructed based on these variables is .

[0059] The action space includes the bandwidth allocated by the UAV to each user, the offloading location where the computing tasks are received from each user, the rotation angle of the UAV, and the flight speed , where is determined by the task offloading indication variables of the cloud server and the edge server. The action space constructed based on these variables is .

[0060] The reward function is constructed as , is the time when the processing delay of the th computing task has exceeded its maximum allowable delay; represents at time the number of computing tasks whose processing delays exceed their maximum allowable delays.

[0061] Based on the state space, action space, and reward function constructed above, the optimization problem OP2 can be transformed into an optimization problem OP3 based on DDPG, specifically:

[0062] OP3: (8)

[0063] where , is the discount factor, is the reward function, represents the statistical average operation.

[0064] When using DDPG to solve the optimization problem OP3, the UAV can act as an agent to determine the bandwidth allocation, computing task offloading, and UAV flight trajectory strategy. This method adopts the Actor-Critic network architecture, which is composed of an Actor network and a Critic network. The network architecture is as Figure 3 shown. In each time slot, the Actor network obtains the state space from the communication environment and outputs an action based on the action policy function , where are the Actor prediction neural network parameters. To enhance the exploration ability of the optimal policy, the agent superimposes Gaussian white noise on the output action. Therefore, the action output by the Actor network is:

[0065] (9)

[0066] After executing this action , the agent can obtain the immediate reward from the communication environment and transition to the next state . Then, the agent forms a matrix with the above results and stores it in the experience sample pool .

[0067] Since the action policy function of the Actor network is related to , it is necessary to optimize the parameter to obtain an action policy that can maximize the reward value, that is, to obtain the optimal action policy:

[0068] (10)

[0069] Among them, , is the statistical average operation. The present invention uses the gradient descent method to solve the optimization problem shown in formula (10), and the solution result can be expressed as:

[0070] (11)

[0071] Among them, is the learning rate of the Actor network, is with respect to gradient value, specifically expressed as:

[0072] (12)

[0073] Among them, is the state-action value output by the Critic prediction neural network, are the parameters of this network.

[0074] The selection of will affect the accuracy of the predicted value. Therefore, it is necessary to optimize the value. Its optimization criterion is to make the prediction error of

[0075] the smallest, that is:

[0076] can be expressed as:

[0077] (14)

[0078] Among them, and are the output values of the target neural networks of Actor and Critic respectively. The parameters of these two neural networks are and .

[0079] The optimization problem (13) can also be solved by the gradient descent method, and the result can be expressed as:

[0080] (15)

[0081] is the learning rate of the Critic network, is with respect to The gradient value, specifically expressed as:

[0082] (16)

[0083] and The optimization process of requires and , and the values of these two parameters can be calculated by the following iterative process:

[0084] (17)

[0085] (18)

[0086] where, is the update rate of the target neural network.

[0087] and The calculation of requires statistical averaging operations. The present invention performs random sampling on the experience samples in the experience sample pool , and estimates the statistical average values in and by taking the average of the sampled samples. For this purpose, an experience sample pool with a capacity of is equipped in the agent , and this experience sample pool is used to store the experience data . In each training time slot, the agent extracts M groups of experience samples from the experience sample pool . Based on the sampled M groups of experience samples (mini-batch samples), and can be respectively expressed as: (19)

[0088] (20)

[0089] (20)

[0090] The multi-dimensional quality-of-service-driven FSO / RF air-ground-space integrated robust resource allocation method proposed by the present invention is a multi-dimensional quality-of-service-driven FSO / RF SAGIN robust computing task offloading, bandwidth allocation, and UAV flight trajectory optimization method. This method can achieve high-quality FSO / RF SAGIN cloud-edge collaborative computing from multiple dimensions and has the following beneficial effects:

[0091] (1) The present invention can simultaneously achieve reliable transmission of computing tasks, the computing tasks can be executed within their allowed maximum time delay, and the execution efficiency of the computing tasks is high;

[0092] (2) When optimizing the network parameters, the present invention takes into account the uncertainty of CSI information. Therefore, it can adjust the network parameters under the condition of uncertain CSI information, meeting the requirements of the actual communication environment.

[0093] (3) Since the ultra-long-distance mission transmission between the satellite and the UAV adopts the FSO communication method with the advantages of large capacity and secure transmission, the present invention can achieve the transmission of large-scale computing tasks and keep the task information that does not want to be leaked confidential.

[0094] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0095] The above-described embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. A multi-dimensional quality of service-driven robust resource allocation method for FSO / RF air-space-ground integrated network, characterized in that It includes the following steps: Step 1: Establish an FSO / RF SAGIN cloud-edge collaborative computing network model, which includes satellites, drones, cloud servers, and ground Internet of Things users. The satellite is used to provide comprehensive services covering the area of interest. The drone, acting as an edge server, is equipped with a caching device and a computing device to provide edge computing and task caching for ground Internet of Things users. Each ground Internet of Things user generates latency-sensitive computing tasks and offloads the computing tasks that are not placed in the user's computing cache queue to the drone through RF links. The drone uses FSO links to forward the computing tasks that are not placed in the drone's computing cache queue to the cloud server through the satellite. Both the FSO link and the RF link adopt adaptive modulation and coding methods; Step 2: Based on the model established in Step 1, construct the optimization problem OP1. The optimization problem OP1 is as follows: Under the constraints of the computational task transmission error frame rate and processing delay, and under the condition of uncertain CSI information, by optimizing the allocation of the bandwidth assigned by the UAV to each ground Internet of Things user, the offloading location of the computational task, the UAV flight trajectory, and the modulation and coding format, achieve the goal of maximizing the total long-term computational task processing volume of all users; Step 3: Use the DDPG deep learning method to solve the optimization problem OP1 to obtain the optimal solution, which is the FSO / RF air-ground-space integrated robust resource allocation scheme.

2. The multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method according to claim 1, characterized in that The expression of the optimization problem OP1 is: ； ； ； In the optimization objective of optimization problem OP1, is the amount of computing tasks executed by the cloud server at time slot t, is the time of one time slot; is the amount of computing tasks executed by the edge server at time slot t, is the number of operations required to execute 1 bit of data for the user, is the user at time slot t 's amount of computing tasks executed, , , is the th modulation order and code rate adopted by the user, and are the modulation orders of the FSO links between the UAV and the satellite, and between the satellite and the cloud server, respectively; Among the constraints of the optimization problem OP1, C1, C2, and C3 are the constraints on the transmission reliability of computing tasks, which respectively require that the maximum transmission error frame rate of computing tasks for the RF link and the FSO links between the UAV and the satellite, and between the satellite and the cloud server be less than the threshold ; is the task transmission error frame rate between the user and the UAV, is the task transmission error frame rate between the UAV and the satellite, is the task transmission error frame rate between the satellite and the cloud server. The uncertain ranges of their estimated CSI information are respectively , and , , and are respectively the signal-to-noise ratios of the links between the user and the UAV, between the UAV and the satellite, and between the satellite and the cloud server; C4 requires that the computing task processing delay of the user shall not exceed its maximum allowed processing delay ; C5 requires that the amount of computing tasks offloaded by the user to the UAV edge server shall not exceed the amount of tasks that the RF link can transmit ; C6 requires that the total bandwidth allocated to the user shall not exceed the maximum bandwidth B provided by the UAV, where is the bandwidth of the user ; C7 and C8 are respectively the constraints on the flight angle and speed of the UAV. Among them, is the maximum flight speed of the UAV; C9 constrains the flight area of the UAV, is the position coordinate of the UAV; C10 and C11 constrain the selection range of adaptive modulation and coding parameters; C12~C14 constrain the computing offloading strategy, requiring that tasks can only be offloaded to one of the cloud server or the edge server, is the edge server offloading indication vector, is the cloud server offloading indication vector.

3. The multi-dimensional quality of service-driven FSO / RF air-space-ground integrated robust resource allocation method according to claim 2, characterized in that, User The frame error rate of the RF link calculation task transmission between the user and the drone is as follows: ； Among them, , and are fitting parameters; is the average mutual information of the modulation and coding scheme; is the user and the signal-to-noise ratio of the RF transmission link between the user and the UAV; is the transmit power of the user; is the received noise power of the UAV; is the RF channel attenuation, which is jointly determined by the fixed attenuation and the random attenuation .

4. The multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method according to claim 2, wherein Mission transmission error frame rate between UAV and satellite is ； Task transmission error frame rate between satellite and cloud server is ； Among them, , and are fitting parameters; M is the modulation order of the FSO link; and are the signal-to-noise ratio of the FSO link, and its value is ; is the transmit power; is the received noise; is the FSO link attenuation, which is jointly determined by the fixed attenuation and the random attenuation .

5. The multi-dimensional quality of service-driven FSO / RF air, space, and ground integrated robust resource allocation method according to any one of claims 2 to 4, characterized in that, In Step 3, when using the DDPG deep learning method to solve the optimization problem OP1, first transform the optimization problem OP1 into the following four sub-optimization problems: ； ； ； ； ； The sub-optimization problems OP1_1 to OP1_3 represent that, under the condition that the maximum frame error rate does not exceed the threshold , the transmission modulation and coding format is adaptively adjusted to maximize the transmission rate. Substitute the solution results of the sub-optimization problems OP1_1 to OP1_3 into the sub-optimization problem OP1_4, and convert the sub-optimization problem OP1_4 into the optimization problem OP2. The expression of the optimization problem OP2 is as follows: ； ； ； ； Among them, is the transmission rate of the computing task of the RF link between the user and the UAV when adaptive transmission is adopted; is the processing delay of the computing task of the user when the system adopts the adaptive transmission mode.

6. The multi-dimensional quality of service-driven FSO / RF air-ground-space integrated robust resource allocation method according to claim 5, characterized in that When solving the maximum transmission rate corresponding to the sub-optimization problems OP1_1~OP1_3, it is obtained by finding the switching threshold of each transmission format.

7. The multi-dimensional quality of service-driven FSO / RF air, space, and ground integrated robust resource allocation method according to claim 5, characterized in that When using the DDPG deep learning method to solve the optimization problem OP2, first construct the state space, action space, and reward function, and transform the optimization problem OP2 into a Markov process.

8. The multi-dimensional quality of service-driven FSO / RF air-space-ground integrated robust resource allocation method according to claim 7, characterized in that The constructed state space includes: the position vector of the UAV , the estimated CSI vectors of K RF links , the FSO link CSI between the UAV and the satellite , the FSO link CSI between the satellite and the cloud server , the length of the user transmission cache queue , the length of the user computing cache queue , the length of the UAV transmission cache queue , the length of the UAV computing cache queue , the remaining average time of the computing tasks in the user computing cache queue , where represents the remaining average time of the user in the computing cache queue of the th user, and the remaining average time of the computing tasks in the user transmission cache queue , where represents the remaining average time of the task in the transmission cache queue of the th user, the remaining average time of the computing tasks in the UAV transmission cache queue , the remaining average time of the computing tasks in the UAV computing cache queue ; The constructed action space includes: the bandwidth allocated by the UAV to each user , the offloading location of the computing tasks received from each user , the rotation angle of the UAV and the flight speed , where it is determined by the task offloading indicator variables of the cloud server and the edge server; The constructed reward function is , where is the task volume of the th computing task whose processing delay has exceeded its maximum allowable delay at time ; denotes that there are computing tasks whose processing delays exceed their maximum allowable delays at time 9. The multi-dimensional quality of service-driven FSO / RF space-air-ground integrated robust resource allocation method according to claim 8, wherein, Based on the constructed state space, action space, and reward function, transform the optimization problem OP2 into the DDPG-based optimization problem OP3. The expression of the optimization problem OP3 is: ； Among them, , is the discount factor, represents the statistical average operation.

10. The multi-dimensional quality-of-service-driven FSO / RF air-space-ground integrated robust resource allocation method according to claim 9, characterized in that When using DDPG to solve the optimization problem OP3, adopt the Actor-Critic architecture, and use the UAV as the agent to determine the bandwidth allocation, computational task offloading location, and UAV flight trajectory strategy.

Images