Cognitive unmanned aerial vehicle MEC system cooperation energy consumption optimization method

Abstract

The invention discloses a cognitive unmanned aerial vehicle MEC system cooperation energy consumption optimization method, which comprises the steps of 1, establishing an LoS path loss model, an unloading rate model, a system energy consumption model and a time delay model of unmanned aerial vehicle cluster air-ground communication according to a deployment position of a cognitive unmanned aerial vehicle cluster and geographic positions and communication requirements of a ground primary user and a secondary base station; step 2, based on an energy consumption model and a time delay model, establishing a joint optimization problem of user association and unmanned aerial vehicle deployment position, communication and computing resource allocation by taking minimization of the total energy consumption of the system as a target under the limited conditions of communication and computing resource constraints of a master user and the unmanned aerial vehicle, master user service quality requirements and the like; and step 3, converting and solving the optimization problem by using a K-Medoids algorithm, Taylor expansion and a continuous convex approximation method to obtain a constraint-based distribution scheme, and performing cognitive unmanned aerial vehicle MEC system cooperative energy consumption optimization according to the scheme.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular, to a method for optimizing the collaborative energy consumption of a cognitive unmanned aerial vehicle (UAV) mobile edge computing (MEC) system. Background Art

[0002] With the rapid development of the sixth-generation mobile communication technology, the popularization of intelligent mobile devices has given rise to a large number of emerging applications, but also brought challenges of high computational intensity and latency sensitivity. Limited by the computing power and battery capacity of mobile devices, the traditional computing mode is difficult to meet the requirements of high-demand tasks. For this reason, mobile edge computing has emerged. By sinking computing and information services to the network edge, it significantly reduces task processing latency, alleviates network congestion, and extends the device's battery life, thereby improving system performance and user experience. In the context of the low-altitude economy, UAVs, with their flexibility, mobility, and low-cost deployment advantages, have become a key carrier to break through the terrain limitations of traditional ground networks and provide high-quality communication services. By deeply integrating UAVs with mobile edge computing systems and optimizing the location of airborne mobile edge computing servers, efficient line-of-sight links and high-quality communication channels can be established, which not only meet the strict requirements of low-latency task processing but also provide efficient data computing services for scenarios such as intelligent inspection, logistics distribution, and emergency communication in the low-altitude economy, further promoting the high-quality development of the low-altitude economy.

[0003] There is an existing Internet of Things system that uses energy harvesting and opportunistic access to overlay the licensed spectrum of the primary communication link for data sensing offloading to a mobile edge computing server, and an optimization strategy based on an online algorithm is proposed to maximize the long-term average sensing rate. The deficiencies of this scheme are as follows: It only considers the case where the mobile edge computing server is deployed at a fixed location and does not consider using the flexible, mobile, and easy-to-deploy characteristics of UAVs to deploy the mobile edge computing server on the UAV and optimize the location to establish an efficient line-of-sight link and high-quality communication channel. Moreover, this article does not further consider collaborative cognitive edge computing.

[0004] There is an existing cognitive UAV-assisted mobile edge computing system where the cognitive UAV is equipped with an edge computing server, uses non-orthogonal multiple access, serves enhanced mobile broadband communication and massive machine-type communication users, and establishes a problem of jointly optimizing the user transmit power. The processing latency of a fixed-location cognitive UAV is significantly reduced through the Rosen gradient projection algorithm. The deficiencies of this method are as follows: The algorithm in this article does not fully consider the scenario where the primary user needs to rely on the cognitive UAV to provide computing support due to latency requirements and computing power limitations. Moreover, this method only considers the single-UAV scenario. Summary of the Invention

[0005] Objective of the Invention: The technical problem to be solved by the present invention is to provide a method for optimizing the collaborative energy consumption of a cognitive UAV MEC system in view of the deficiencies of the prior art.

[0006] To solve the above technical problem, the present invention discloses a method for optimizing the collaborative energy consumption of a cognitive UAV MEC (Mobile Edge Computing) system, including the following steps:

[0007] Step 1, establish a LoS path loss model, an offloading rate model, a system energy consumption model, and a delay model for the air-ground communication of the UAV cluster according to the deployment location of the cognitive UAV cluster, the geographical locations of the ground primary users and secondary base stations, and the communication requirements.

[0008] Step 2, based on the energy consumption model and the delay model, under the restricted conditions such as the communication and computing resource constraints between the primary users and the UAVs and the quality of service (QoS) requirements of the primary users, with the goal of minimizing the total system energy consumption, establish a joint optimization problem of user association, UAV deployment location, and communication and computing resource allocation.

[0009] Step 3, use the K-Medoids algorithm, Taylor expansion, and successive convex approximation method to transform and solve the above optimization problem, obtain an allocation scheme based on the above constraints, and optimize the collaborative energy consumption of the cognitive UAV MEC system according to this scheme.

[0010] The deployment location of the cognitive UAV cluster in Step 1 and the geographical locations of the ground primary users and secondary base stations are as follows:

[0011] There are U primary users randomly distributed in the ground area. The primary base station is located at the edge of the area, far from the users and with extremely poor communication conditions with the users. The secondary base station is deployed in the center of the area. M cognitive UAVs carrying edge computing servers hover and remain stationary at a height of H above the ground area.

[0012] The communication requirements in Step 1 are as follows: In each task completion cycle, the UAV cluster needs to transmit the data monitored in the area to the secondary base station; the primary users offload part of the computing tasks to the cognitive UAVs. Single antennas are equipped at the users, UAVs, and secondary base stations. The orthogonal frequency division multiple access (OFDM) technology is used for offloading the computing tasks from the users to the computing UAVs. The same and orthogonal bandwidths are allocated to each primary user. The channels between the UAVs and the ground devices and between the UAVs and the ground secondary base stations are all line-of-sight channels. Let the label of the primary user be u, the label of the computing UAV be m, and the label of the ground secondary base station be SBS. The task execution time T is divided into N sub-slots, and the frame length of each slot n is small enough.

[0013] The line-of-sight (LoS) path loss model for the UAV swarm's air-ground communication in Step 1 includes the LoS path loss model between the user and the UAV and the LoS path loss model between the UAV and the ground secondary base station;

[0014] The LoS path loss model between the user and the UAV (the cognitive UAV carrying the edge computing server) is:

[0015]

[0016] where u represents the index of the primary user, with a range of 1, 2,..., U; m represents the index of the UAV, with a range of 1, 2,..., M; represents the channel gain between the primary user u and the UAV m; h0 is the channel gain when the distance between the primary user and the UAV is 1 m; represents the Euclidean distance between the primary user labeled u and the UAV labeled m, H is the flight altitude of the computing UAV, q u represents the position vector of the primary user labeled u, q m represents the position vector of the computing UAV labeled m, ||·|| represents the vector norm;

[0017] The LoS path loss model between the UAV and the ground secondary base station is:

[0018]

[0019] where, represents the channel gain between the UAV m and the ground secondary base station, represents the Euclidean distance between the UAV labeled m and the ground secondary base station SBS, q sbs represents the position vector of the ground secondary base station SBS.

[0020] The offloading rate model described in Step 1 includes:

[0021] The ground base station BS task offloading rate model calculated based on the LoS path loss model between the UAV and the ground secondary base station is:

[0022]

[0023] where R m (n) represents the task offloading rate received by the ground secondary base station SBS from the UAV m in the nth time slot, P m (n) represents the offloading power of the UAV m, represents the bandwidth obtained by the UAV m;

[0024] LoS path loss model between user and UAV Offloading rate model of UAV swarm's air-to-ground communication calculated as follows:

[0025]

[0026] Among them, R u (n) represents the total offloading amount from primary user u to UAV m in the nth time slot, N0 represents the noise power spectral density, and P u (n) represents the offloading power of primary user u in the nth time slot; B u is the spectral bandwidth adopted by each primary user; log2(·) represents the logarithmic operation with base 2; is a binary variable representing the offloading decision from primary user u to UAV m.

[0027] In step 1, the total system energy consumption model in the nth time slot of the system energy consumption model is as follows:

[0028]

[0029] Among them, λ1 and λ2 represent weighting factors, and their value ranges are both from 0 to 1, local energy consumption model of the primary user, local delay model of the primary user, energy consumption model for the UAV to assist the primary user in processing computing tasks, transmission energy consumption model for the UAV to transmit data to the secondary base station after obtaining communication bandwidth;

[0030] The local energy consumption model of the primary user is:

[0031]

[0032] Among them, represents the computing task of local computing by primary user u in the nth time slot, and f u (n) represents the CPU clock frequency of primary user u, and c u represents the number of CPU cycles required for the UAV to process 1 bit of computing task; D u (n) represents the total computing task of primary user u in the nth time slot, and ω u is the effective coefficient of the primary user's capacitance;

[0033] The transmission energy consumption model of the primary user is as follows:

[0034]

[0035] Among them, Denote the time ratio of the primary user \(u\) offloading part of the computing task to the associated UAV in the \(n\)th time slot;

[0036] The energy consumption model of the UAV assisting the primary user in processing the computing task is as follows:

[0037]

[0038] where, Denote the computing task offloaded from the primary user \(u\) to the UAV in the \(n\)th time slot, \(f\) m (n) denote the CPU clock frequency of the UAV, \(c\) m Denote the number of CPU cycles required for the UAV to process 1 bit of computing task, \(\omega\) m Is the effective coefficient of the capacitance of the primary user;

[0039] After the UAV obtains the communication bandwidth, the transmission energy consumption model for transmitting data to the secondary base station is as follows:

[0040]

[0041] where, Denote the time ratio of the UAV \(m\) transmitting the monitoring data of its own area to the secondary base station SBS in the \(n\)th time slot.

[0042] The delay model described in step 1 includes:

[0043] The local delay model of the primary user:

[0044]

[0045] where, Denote the computing delay of the primary user's local computing, Denote the computing task of the primary user \(u\)'s local computing in the \(n\)th time slot, \(f\) u (n) denote the CPU clock frequency of the primary user, \(c\) u Denote the number of CPU cycles required for the UAV to process 1 bit of computing task;

[0046] The delay model of the UAV assisting the primary user in processing the computing task:

[0047]

[0048] where, The delay of the UAV assisting the primary user in processing the computing task, \(f\) m (n) denote the CPU clock frequency of the UAV, Is a binary variable, representing the offloading decision of the primary user \(u\) to the UAV \(m\).

[0049] In Step 2, an optimization problem of minimizing the total system energy consumption \(E(n)\) is established under the restricted conditions such as the communication and computing resource constraints between the primary user and the UAV, as well as the QoS requirements of the primary user. The objective function is as follows:

[0050]

[0051] where \(min\) represents the minimization operation, time allocation user association power allocation \(P = \{P u (n), P m (n)\}\), task allocation CPU clock frequency allocation \(f = \{f u (n), f m (n)\};

[0052] The constraint conditions of this objective function include:

[0053] Constraint condition (b): gives the maximum offloading power limit of the user and the UAV in each time slot:

[0054]

[0055] where \(s.t.\) represents the constraint condition;

[0056] Constraint condition (c): the time constraint for each sub-time slot is:

[0057]

[0058] Constraint condition (d): for different users \(u, v\in U\) associated with the same UAV, the time ratio and the time ratio have the same total duration:

[0059]

[0060] Constraint condition (e): constraints the local and offloading computing task sizes of the user in each time slot:

[0061]

[0062] Constraint condition (f): constraints the total computing task of the user:

[0063]

[0064] Constraint condition (g): the CPU clock frequency constraints of the primary user and the cognitive UAV:

[0065]

[0066] Constraint (h): Latency constraint on local computing of the primary user and the cognitive UAV:

[0067]

[0068] Constraint (i): Restrictions are imposed on the amount of data for the primary user to offload tasks to the cognitive UAV and for the cognitive UAV to transmit tasks to the secondary base station:

[0069]

[0070] Constraint (j): Each primary user can only choose between two offloading decisions: associating or not associating.

[0071]

[0072] Constraint (k): Each primary user can associate with at most one cognitive UAV, and the maximum number of primary users that each cognitive UAV can serve does not exceed U max :

[0073]

[0074] In all of the above formulas, is a binary variable representing the offloading decision of primary user u to UAV m, is a binary variable representing the offloading decision of primary user v to UAV m, U max represents the maximum number of users, B u is the spectrum bandwidth adopted by each primary user, B m represents the bandwidth obtained by UAV m, c m represents the number of CPU cycles required for the UAV to process 1 bit of computing task, d m (n) represents the data transmitted by the cognitive UAV to the secondary base station in the nth time slot, represents the computing task offloaded by primary user u to the UAV in the nth time slot, D u (n) is the total computing task volume of primary user u in the nth time slot, f m (n) represents the CPU clock frequency of the UAV, P m (n) represents the offloading power of UAV m, P u (n) represents the offloading power of primary user u in the nth time slot, represents the time ratio of primary user u partially offloading computing tasks to the associated UAV in the nth time slot, represents the time ratio of primary user v partially offloading computing tasks to the associated UAV in the nth time slot, represents the time ratio of the UAV m processing the computing task of user u, Denotes the time ratio of the cognitive UAV m to process the computing tasks of user v. Denotes the time ratio of UAV m to transmit the monitoring data of its own area to the secondary base station SBS in the nth time slot. Denotes the local delay model of the primary user. The delay of the UAV assisting the primary user to process computing tasks, R u R(n) denotes the total offloading amount from the primary user u to the UAV m in the nth time slot. m ω(n) represents the task offloading rate received by the ground secondary base station SBS from the UAV m in the nth time slot. m Is the effective coefficient of the capacitance of the primary user.

[0075] The specific method of step 3 is to decouple the problem Into three sub-problems:

[0076] User association and UAV deployment sub-problem

[0077] Joint optimization sub-problem of power, time slot and task allocation

[0078] Computing resource allocation sub-problem

[0079] The joint optimization of the three sub-problems is specifically as follows:

[0080] Step 3.1, solve the problem To obtain the UAV deployment location and user association α.

[0081] Step 3.2, initialize the iteration number k = 0 and the convergence thresholds ε and ζ.

[0082] Step 3.3, by solving the problem To obtain the optimal power allocation p k Of the CPU clock frequency allocation f k+1 At the given kth iteration, time allocation t k+1 And task allocation d k+1

[0083] Step 3.4, initialize the iteration number l to 0.

[0084] Step 3.5, by solving the problem To obtain the optimal CPU clock frequency allocation f k Of the power allocation p k At the given kth iteration, time allocation t k And task allocation d l+1 Until the condition

[0085] Step 3.6, set f k+1 = f l+1 ;

[0086] Step 3.7, loop and execute Steps 4.2 to 4.6 until the target value that meets the condition is less than the convergence threshold ε;

[0087] Step 3.8, output the optimal results: user association α, time allocation t, task allocation d, and CPU clock frequency allocation f.

[0088] The solution to the user association and UAV deployment sub-problem is specifically as follows:

[0089] Step 3-1-1, use the K-Medoids algorithm to solve this problem. The dissimilarity index of users is defined as follows:

[0090]

[0091] where, represents the distance from user u to x, and x represents the UAV.

[0092] Step 3-1-2, calculate the dissimilarity index Z of all users u ;

[0093] Step 3-1-3, after arranging the dissimilarity index Z u , select the first O minimum values as the initial centroid clusters o. Each centroid corresponds to a cluster (i.e., user grouping). The cluster o contains the centroid and the users assigned to it. Finally, UAVs need to be deployed to serve this cluster;

[0094] Step 3-1-4, for each cluster, sort the distance G u,o from u = 1 to U in ascending order, and assign the first U max users to the cluster o. The offloading decision of the corresponding primary user u to the UAV m is set to

[0095] Step 3-1-5, for each cluster, calculate the sum of distances G max between the centroid in the current cluster and the users from user index l = 1 to user U l,o ;

[0096] Step 3-1-6, for the current cluster, from j = 1 to U max -1 (excluding the current centroid), calculate the sum of distances G max between the temporary centroid of j and the users from user index l = 1 to U l,j ;

[0097] Step 3-1-7, if it satisfies Then update the current centroid to user j, and repeat steps 3-1-4 to 3-2-6;

[0098] Step 3-1-8, for each cluster, output the coordinates (X o , Y o ) of each final centroid as the horizontal position coordinates of the currently calculated UAV, i.e., the deployment position of the UAV, and the user association α;

[0099] The solution to the joint optimization sub-problem of power, time slot, and task allocation is specifically as follows:

[0100] Step 3-2-1, under the given CPU clock frequency allocation f, transform the joint optimization sub-problem of power, time slot, and task allocation into sub-problem

[0101]

[0102] Its constraint conditions are: s.t. (b)-(i) (constraint conditions b to i) (1.1b)

[0103] Among them, (i) is a concave condition, and the remaining conditions are convex conditions;

[0104] Step 3-2-2, use variable substitution to simplify the problem, and define auxiliary variables Auxiliary variables Rewrite problem P2.1 as problem P2.2:

[0105]

[0106] Its constraint conditions are: s.t. (b)-(h) (1.2b)

[0107] Step 3-3-3, introduce an auxiliary function to simplify the non-convex constraints (1.2c) and (1.2d):

[0108]

[0109]

[0110] Define the auxiliary function:

[0111]

[0112] Step 3-3-4, use the first-order Taylor expansion to linearly approximate the concave function in (1.2c), and approximate the auxiliary function That is,

[0113]

[0114] Among them, and At the l-th iteration point, for the auxiliary function respectively for and the first-order derivatives of, denote the function value of the auxiliary function at the l-th iteration point, that is

[0115]

[0116] Similarly, use the first-order Taylor expansion to linearly approximate the concave function in (1.2d), and approximate the auxiliary function to obtain

[0117] Step 3-3-5, construct an approximate optimization problem, substitute the approximate constraint expression in Step 4-3-4 and rewrite the problem to obtain the problem as follows:

[0118]

[0119] This problem is a standard linear programming problem, which is effectively solved by the CVX tool to obtain the power allocation p, time allocation t and task allocation d.

[0120] The calculation resource allocation sub-problem is solved specifically as follows:

[0121] Given the power allocation p, time allocation t and task allocation d, transform the calculation resource allocation sub-problem into

[0122]

[0123] s.t.(g),(h)(1.4b)

[0124] This problem is a standard convex optimization problem, which is effectively solved by the convex optimization toolbox (CVX, Convex Optimization Toolbox) to obtain the CPU clock frequency allocation f.

[0125] Beneficial effects:

[0126] (1) The present invention first proposes a collaborative energy consumption optimization scheme for a cognitive UAV MEC system based on resource exchange;

[0127] (2) Under the conditions of limited device computing power, limited device communication capacity, and limited device data transmission power, etc., the present invention optimizes the total system energy consumption by adjusting the deployment position of the UAV and the computing resources, communication resources, and transmission power of the device.

[0128] (3) The convergence speed of the cyclic iterative optimization algorithm based on K-Medoids of the present invention is fast and the efficiency is high;

[0129] (4) The present invention can provide highly reliable technical guidance for the cooperative computing of cognitive UAV-assisted communication systems. Description of the Drawings

[0130] Figure 1 It is a schematic model diagram of a cooperative mobile edge computing network based on the location deployment grouping of cognitive UAVs in the present invention.

[0131] Figure 2 It is a diagram of the system time slot allocation scheme in the present invention.

[0132] Figure 3 It is a diagram showing the relationship between the total system energy consumption and the number of iterations under different schemes of the present invention.

[0133] Figure 4 It is a diagram showing the relationship between the total system energy consumption and the total number of tasks of each user under different schemes of the present invention.

[0134] Figure 5 It is a diagram showing the relationship between the total system energy consumption and the bandwidth size of each user under different schemes of the present invention. Detailed Embodiment

[0135] In view of the deficiencies of the prior art, the present invention provides a cooperative energy consumption optimization scheme for a cognitive UAV-assisted edge computing system, which effectively reduces the total system energy consumption and is applied to an actual cognitive UAV-assisted edge computing network.

[0136] This embodiment combines simulation conditions:

[0137] A method for optimizing the cooperative energy consumption of a cognitive UAV MEC system includes the following steps:

[0138] Step 1, according to the deployment locations of the cognitive UAV cluster, the geographical locations and communication requirements of the ground primary users and secondary base stations, establish a LoS path loss model, an offloading rate model, a system energy consumption model, and a delay model for the air-ground communication of the UAV cluster;

[0139] Step 2, based on the energy consumption model and the delay model, under the restricted conditions such as the communication and computing resource constraints between the primary users and the UAVs and the quality of service (QoS) requirements of the primary users, with the goal of minimizing the total system energy consumption, establish a joint optimization problem of user association, UAV deployment location, and communication and computing resource allocation;

[0140] Step 3: Use the K-Medoids algorithm, Taylor expansion, and continuous convex approximation method to transform and solve the above optimization problem, and obtain an allocation scheme based on the above constraints.

[0141] The deployment locations of the cognitive UAV swarm described in Step 1, the geographical locations of the ground primary users and the secondary base station are as Figure 1 shown: There are U primary users randomly distributed in the ground area. The primary base station is located at the edge of the area, far from the users, and the communication conditions with the users are extremely poor. The secondary base station is deployed in the center of the area. M cognitive UAVs carrying edge computing servers hover and remain stationary at a height of H above the ground area.

[0142] The communication requirements described in Step 1 are as follows: In each task completion cycle, the UAV swarm needs to transmit the data monitored in the area to the secondary base station; the primary users offload part of the computing tasks to the cognitive UAVs. Single antennas are equipped at the users, UAVs, and secondary base stations. The orthogonal frequency division multiple access (OFDM, Orthogonal Frequency Division Multiplexing) technology is used for the computing tasks to be offloaded from the users to the computing UAVs. Each primary user is assigned the same and orthogonal bandwidth. The channels between the UAVs and the ground devices and between the UAVs and the ground secondary base station are all line-of-sight channels. Let the label of the primary user be u, the label of the computing UAV be m, and the label of the ground secondary base station be SBS. Assume that the system operates in a time-slot synchronization mode. As Figure 2 shown, the task execution time T is divided into N sub-time slots. The frame length of each time slot n is small enough, with a length of T / N. Each time slot is further divided into three sub-time slots. The proportion of the first sub-time slot is for the primary user u∈U to offload part of the computing tasks to the associated cognitive UAV m∈M; the proportion of the second sub-time slot is for the cognitive UAV m∈M to process the computing tasks of the user u∈U; after the cognitive UAV assists all the associated users to complete the computing tasks and returns the results, it then obtains their authorized spectrum and, in the third sub-time slot (with a proportion of ), transmits its own monitored data d m (n) to the ground secondary base station. Since the data volume of the computing results is much smaller than the data volume of the offloaded tasks, the time for the cognitive UAV to transmit the computing results back can be ignored. Assume that for the users associated with the same cognitive UAV, the UAV needs to process all their computing tasks before it can obtain the spectrum resources for its own data transmission. Therefore, for different users associated with the same UAV, and have the same total duration.

[0143] 1. The common simulation parameters are set as follows:

[0144] Assume that U = 30 terrestrial primary users are randomly distributed in a 50*50m 2 square area. The users are divided into M = 5 clusters, and each cluster is provided with computing services by a cognitive drone. A secondary base station is deployed at [25m, 25m, 0]. In addition, the maximum number of users U max for cluster services is 6. After the positions of the cognitive drones are deployed, they are fixed in the air and remain stationary. The weighting coefficients λ1 and λ2 are 0.5, and the convergence thresholds ε and ζ are 10 -3 . The total number of time slots N = 20, the bandwidth B = 2MHz, the noise power N0 = -150dbm / Hz, and the total number of tasks D u (n) = 6Mbit processed by each user per time slot. The total number of tasks d m (n) = 6Mbit transmitted by each cognitive drone per time slot. The channel gain power h0 at a distance of 1m is 50dB, the flight altitude H of the drone is 12m, and the maximum CPU clock frequency of the primary user The maximum CPU clock frequency of the cognitive drone The time slot length T = 0.3s. The number of CPU cycles required for the primary user and the cognitive drone to process 1 bit of computing task is 1000cycles / bit. The capacitance coefficient ω of the primary user u = 10 -27 , the capacitance coefficient ω of the cognitive drone m = 10 -28 , the maximum power of the primary user The maximum power of the cognitive drone

[0145] To illustrate the performance advantages of the algorithm proposed in the present invention, three benchmarks are defined for comparison as follows:

[0146] (1) Only local computing scheme: The terrestrial primary users do not cooperate with the cognitive drones, and all computing tasks are processed locally.

[0147] (2) Fixed task allocation scheme: The computing tasks offloaded from each primary user to the cognitive drone account for 50% of the total tasks of each primary user.

[0148] (3) Drone random position deployment scheme: The positions of the cognitive drones are randomly generated.

[0149] 2. Simulation content

[0150] Figure 3Shows the relationship between the total system energy consumption and the number of iterations under different scenarios. As can be seen from the figure, the total system energy consumption of the algorithm proposed in this paper, the fixed task allocation scheme, and the random UAV position deployment scheme all show a rapid downward trend in the first 5 iterations and then gradually tend to be stable. In contrast, for the local computing only scheme, due to the lack of cooperation between users and UAVs, its total system energy consumption only depends on the local computing overhead of users and is thus independent of the number of iterations, presenting as a horizontal line. The results show that the algorithm proposed in this paper is always superior to the other three benchmark schemes in terms of total system energy consumption. This performance advantage is mainly due to the joint optimization of resource allocation and UAV position deployment in the algorithm, thus achieving the global minimization of system energy consumption.

[0151] Figure 4 Presents the relationship between the total system energy consumption and the task volume of each user under different scenarios. As the task size of each user increases, the total system energy consumption under each scenario also increases accordingly. For the algorithm in this paper, the fixed task allocation scheme, and the random UAV position deployment scheme, when the total task volume of each user is less than 5 megabits, the system computing resources are sufficient to handle the current computing load, so the total system energy consumption increases relatively slowly. However, subsequently, due to the limited resources of cognitive UAVs, more computing tasks are transferred to local processing, increasing the energy consumption cost and resulting in a rapid increase in the total system energy consumption.

[0152] Figure 5 Is for the relationship between the total system energy consumption and the bandwidth of each user under different scenarios. As shown by Figure 5 As the bandwidth of each user increases, for a certain amount of transmission, the power required for users and cognitive UAVs to transmit data decreases, and the energy consumption reduces. When the bandwidth increases to a certain value, according to the Shannon capacity formula, the channel capacity will tend to a fixed value, that is, the decrease in power is limited, and the total system energy consumption will eventually converge. In the local computing only scheme, users are not involved in transmitting data to cognitive UAVs, so the computing energy consumption remains unchanged. Generally speaking, the algorithm proposed in this paper shows a lower total system energy consumption level in various situations.

[0153] The present invention provides a method for optimizing the collaborative energy consumption of a cognitive UAV MEC system. There are many methods and ways to specifically implement this technical solution. The above description is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Each component not clearly defined in this embodiment can be implemented by existing technologies.

Claims

1. A collaborative energy consumption optimization method for cognitive UAV MEC systems, characterized in that, It includes the following steps: Step 1: According to the deployment location of the cognitive UAV cluster, the geographical locations of the ground primary users and secondary base stations, and the communication requirements, establish the LoS path loss model, offloading rate model, system energy consumption model, and delay model for the air-ground communication of the UAV cluster. Step 2: Based on the energy consumption model and delay model, under the limited conditions such as the communication and computing resource constraints between the primary users and the UAVs and the quality of service requirements of the primary users, with the goal of minimizing the total system energy consumption, establish a joint optimization problem of user association, UAV deployment location, and communication and computing resource allocation. Step 3: Use the K-Medoids algorithm, Taylor expansion, and successive convex approximation method to transform and solve the above optimization problem, obtain the allocation scheme based on the above constraints, and optimize the collaborative energy consumption of the cognitive UAV MEC system according to this scheme.

2. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, wherein, The deployment location of the cognitive UAV cluster described in Step 1 and the geographical locations of the ground primary users and secondary base stations are as follows: There are U primary users randomly distributed in the ground area. The primary base station is located at the edge of the area, far from the users, and the communication conditions with the users are extremely poor. The secondary base station is deployed in the center of the area. M cognitive UAVs carrying edge computing servers hover and remain stationary at a height of H above the ground area.

3. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, wherein The communication requirements described in Step 1 are as follows: In each task completion cycle, the UAV cluster needs to transmit the data monitored in the area to the secondary base station; the primary users offload part of the computing tasks to the cognitive UAVs. Each primary user is assigned the same and orthogonal bandwidth. The channels between the UAVs and the ground devices and between the UAVs and the ground secondary base stations are all line-of-sight channels. Let the label of the primary user be u, the label of the computing UAV be m, and the label of the ground secondary base station be SBS. The task execution time T is divided into N sub-slots, and the frame length of each slot n is small enough.

4. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, characterized in that, The line-of-sight path loss model for the air-ground communication of the UAV cluster described in Step 1 includes the LoS path loss model between the user and the UAV and the LoS path loss model between the UAV and the ground secondary base station. The LoS path loss model between the user and the UAV is: where \(u\) represents the index of the primary user, ranging from \(1, 2, \cdots, U\); \(m\) represents the index of the UAV, ranging from \(1, 2, \cdots, M\). represents the channel gain between the primary user \(u\) and the UAV \(m\); \(h_0\) is the channel gain when the distance between the primary user and the UAV is \(1m\). represents the Euclidean distance between the primary user labeled \(u\) and the UAV labeled \(m\), \(H\) is the flight altitude of the computing UAV, \(q\) u represents the position vector of the primary user labeled \(u\), \(q\) m represents the position vector of the computing UAV labeled \(m\), \(\|\cdot\|\) represents the vector norm. The LoS path loss model between the UAV and the ground secondary base station is: Among them, represents the channel gain between the unmanned aerial vehicle m and the ground secondary base station, represents the Euclidean distance between the unmanned aerial vehicle labeled m and the ground secondary base station SBS, q sbs represents the position vector of the ground secondary base station SBS.

5. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, characterized in that The offloading rate model described in Step 1 includes: LoS path loss model between UAV and ground secondary base station Calculated task offloading rate model of ground base station BS Among them, R m (n) represents the task offloading rate of the ground secondary base station SBS receiving from the unmanned aerial vehicle m at the nth time slot, and P m (n) represents the offloading power of the unmanned aerial vehicle m, represents the bandwidth obtained by the unmanned aerial vehicle m; LoS path loss model based on the distance between the user and the UAV Offloading rate model for UAV swarm air-ground communication calculated based on the above: Among them, R u (n) represents the total offloading volume from the primary user u to the UAV m in the nth time slot, N 0 represents the noise power spectral density, P u (n) represents the offloading power of the primary user u in the nth time slot; B u is the spectral bandwidth adopted by each primary user; log2(·) represents the logarithmic operation with base 2; is a binary variable representing the offloading decision from the primary user u to the UAV m.

6. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, characterized in that In Step 1, the total system energy consumption model in the nth time slot of the system energy consumption model is as follows: where, λ1 and λ2 represent weighting factors, is the local energy consumption model of the primary user, is the local latency model of the primary user, is the energy consumption model for the UAV to assist the primary user in processing computing tasks, is the transmission energy consumption model for the UAV to transmit data to the secondary base station after obtaining communication bandwidth; The local energy consumption model of the primary user is: Among them, represents the computing task locally computed by the primary user u in the nth time slot, f u (n) represents the CPU clock frequency of the primary user, c u represents the number of central processor cycles required for the UAV to process 1-bit computing task; D u (n) represents the total computing task of the primary user u in the nth time slot, ω u is the capacitance effective coefficient of the primary user; The transmission energy consumption model of the primary user is as follows: Among them, represents the time ratio of the primary user u partially offloading the computing task to the associated UAV in the nth time slot; The energy consumption model for the UAV to assist the primary user in processing the computing task is as follows: Among them, represents the computing task unloaded from the primary user u to the UAV in the nth time slot, and f m (n) represents the CPU clock frequency of the UAV, and c m represents the number of central processor cycles required for the UAV to process 1 bit of computing task, and ω m is the capacitance effective coefficient of the primary user; After the UAV obtains the communication bandwidth, the transmission energy consumption model for transmitting data to the secondary base station is as follows: Among them, represents the time ratio of the UAV m transmitting the monitoring data of its own area to the secondary base station SBS in the nth time slot.

7. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, wherein The delay model described in Step 1 includes: The local delay model of the primary user: Among them, represents the computing delay of the primary user's local computing, represents the computing task of the primary user u's local computing in the nth time slot, f u (n) represents the CPU clock frequency of the primary user, c u represents the number of central processor cycles required for the UAV to process 1-bit computing tasks; The delay model for the UAV to assist the primary user in processing the computing task: Among them, The latency for the UAV to assist the primary user in processing computing tasks, f m (n) represents the CPU clock frequency of the UAV, is a binary variable representing the offloading decision of the primary user u to the UAV m.

8. A method for collaborative energy consumption optimization of a cognitive UAV MEC system according to claim 1, characterized in that, In step 2, under the restricted conditions such as the communication and computing resource constraints between the primary user and the UAV, and the primary user QoS requirements, an optimization problem of minimizing the total system energy consumption E(n) is established as follows The objective function is as follows: Among them, min represents the minimization operation, and time allocation User association Power allocation P = {P u (n), P m (n)}, task allocation CPU clock frequency allocation f = {f u (n), f m (n)}; The constraint conditions of this objective function include: Constraint condition (b): Gives the maximum offloading power limit of the user and the UAV in each time slot: In the formula, s.t. represents the constraint condition; Constraint condition (c): The time constraint for each sub-slot is: Constraint (d): For different users u, v ∈ U associated with the same drone, the time ratio and the time ratio have the same total duration: Constraint condition (e): Constrains the size of the local and offloaded computing tasks of the user in each time slot: Constraint condition (f): Constrains the total computing task of the user: Constraint (g): CPU clock frequency constraint for the primary user and the cognitive UAV: Constraint (h): Latency constraint for local computing of the primary user and the cognitive UAV: Constraint (i): Constraints are imposed on the amount of data for the primary user to offload tasks to the cognitive UAV and for the cognitive UAV to transmit tasks to the secondary base station: Constraint (j): Each primary user can only choose between two offloading decisions: association or non - association: Constraint (k): Each primary user can be associated with at most one cognitive UAV, and the number of primary users that each cognitive UAV can serve is at most U max : In all of the above formulas, is a binary variable representing the offloading decision of primary user u to UAV m, is a binary variable representing the offloading decision of primary user v to UAV m, U max represents the maximum number of users, B u is the spectrum bandwidth adopted by each primary user, B m represents the bandwidth obtained by UAV m, c m represents the number of CPU cycles required for the UAV to process 1 bit of computing task, d m $(n)$ represents the data transmitted by the cognitive UAV to the secondary base station in the $n$-th time slot, represents the computing task offloaded by primary user u to the UAV in the $n$-th time slot, D u $(n)$ is the total computing task volume of primary user u in the $n$-th time slot, f m $(n)$ represents the CPU clock frequency of the UAV, P m $(n)$ represents the offloading power of UAV m, P u $(n)$ represents the offloading power of primary user u in the $n$-th time slot, represents the time ratio of primary user u offloading part of the computing task to the associated UAV in the $n$-th time slot, represents the time ratio of primary user v offloading part of the computing task to the associated UAV in the $n$-th time slot, represents the time ratio of the cognitive UAV m processing the computing task of user u, represents the time ratio of the cognitive UAV m processing the computing task of user v, represents the time ratio of UAV m transmitting its own area monitoring data to the secondary base station SBS in the $n$-th time slot, represents the local delay model of the primary user, The delay of the UAV assisting the primary user in processing the computing task, R u $(n)$ represents the total offloading volume from primary user u to UAV m in the $n$-th time slot, R m $(n)$ represents the task offloading rate received by the ground secondary base station SBS from UAV m in the $n$-th time slot, ω m is the effective coefficient of the primary user's capacitance.

9. The collaborative energy consumption optimization method for a cognitive UAV MEC system according to claim 1, wherein The specific method of Step 3 is to decouple the problem into three sub-problems: User Association and Drone Deployment Sub - problem Joint optimization sub-problem of power, time slot and task allocation Computing resource allocation sub-problem The joint optimization of the three sub - problems is specifically as follows: Step 3.1, from the problem Solve to obtain the UAV deployment location and the user association α; Step 3.2, Initialize the iteration number k = 0, and the convergence thresholds ε and ζ; Step 3.3, by solving the problem obtain the optimal power allocation p k for the CPU clock frequency allocation f k+1 at the given k-th iteration, time allocation t k+1 and task allocation d k+1 Step 3.4, Initialize the iteration number l to 0; Step 3.5, by solving the problem obtain the power allocation p k , time allocation t k , task allocation d k and the optimal CPU clock frequency allocation f l+1 until the condition Step 3.6, set f k+1 = f l+1 ; Step 3.7, loop through steps 4.2 to 4.6 until the target value is less than the convergence threshold ε; Step 3.8, Output the optimal results: user association α, time allocation t, task allocation d, and CPU clock frequency allocation f.

10. A method for optimizing the collaborative energy consumption of a cognitive UAV MEC system according to claim 9, characterized in that, The user association and UAV deployment sub-problem is solved specifically as follows: Step 3 - 1 - 1, The dissimilarity index of users is defined as follows: Among them, represents the distance from user u to x, where x represents the drone; Step 3-1-2, calculate the dissimilarity index Z of all users u ; Step 3-1-3, for the dissimilarity index Z u After arranging, select the first O minimum values as the initial centroid cluster o. Each centroid corresponds to a cluster. Cluster o contains the centroid and the users assigned to it. Finally, it is necessary to deploy drones to serve this cluster; Step 3-1-4, for each cluster, for the distance G u,o sort in ascending order from u = 1 to U, and assign the first U max users to cluster o, and set the offloading decision of the corresponding primary user u to the drone m as Step 3-1-5, for each cluster, calculate the sum of distances G between the centroid in the current cluster and users from user index l = 1 to user U max and l,o ; Step 3-1-6, for the current cluster, from j = 1 to U max -1, calculate the sum of distances G from the user index l = 1 to U max users with j as the temporary centroid l,j ; Step 3-1-7, if satisfied then update the current centroid to user j, and repeat Steps 3-1-4 to 3-2-6; Step 3-1-8, for each cluster, output the coordinates (X o , Y o ) of each final centroid as the horizontal position coordinates of the currently calculated UAV, i.e., the deployment position of the UAV, and the user association α; The solution to the joint optimization sub - problem of power, time - slot, and task allocation is specifically as follows: Step 3-2-1, under the given CPU clock frequency allocation f, transform the joint optimization sub-problem of power, time slot, and task allocation into a sub-problem Its constraint conditions are: s.t.(b)-(i) (1.1b) Among them, (i) is a concave condition, and the rest are convex conditions; Step 3-2-2, simplify the problem using variable substitution and define auxiliary variables Auxiliary variable The problem Rewrite it as the problem Its constraint conditions are: s.t.(b)-(h) (1.2b) Step 3 - 3 - 3, Introduce an auxiliary function to simplify the non - convex constraints (1.2c) and (1.2d): Define the auxiliary function: Step 3-3-4: Use the first-order Taylor expansion to linearly approximate the concave function in (1.2c), and approximate the auxiliary function That is, Among them, and are the first-order derivatives of the auxiliary function with respect to and at the l-th iteration point, represents the function value of the auxiliary function at the l-th iteration point, that is, Similarly, the concave function in (1.2d) is linearly approximated by the first-order Taylor expansion, and the auxiliary function is approximated to obtain Step 3-3-5: Construct an approximate optimization problem, substitute the approximate constraint expressions in Step 4-3-4, and rewrite the problem. Obtain the problem as follows: This problem is a standard linear programming problem, which is effectively solved by the CVX tool to obtain the power allocation p, time allocation t, and task allocation d; The sub-problem of computing resource allocation The solution is specifically as follows: Given the power allocation \(p\), time allocation \(t\) and task allocation \(d\), the sub-problem of computing resource allocation is transformed into s.t.(g),(h)(1.4b) This problem is a standard convex optimization problem, which is effectively solved through the convex optimization toolbox to obtain the CPU clock frequency allocation f.

Images