Method and system for resource allocation in uav-assisted heterogeneous networks based on short packet communication

Abstract

The application discloses a resource allocation method and system in a UAV-assisted heterogeneous network based on short packet communication, and the method comprises the following steps: constructing a UAV-assisted heterogeneous network system, wherein the system comprises a base station, a UAV, a plurality of Internet of Things devices and a plurality of pairs of D2D devices; constructing an optimization problem with the target of maximizing the minimum effective information amount received by the Internet of Things devices in the whole system; fixing the UAV position and the channel allocation strategy, the block length allocation strategy and the channel allocation strategy, the block length allocation strategy and the UAV position respectively, converting the optimization problem into three sub-problems; adopting a two-layer iterative algorithm to jointly optimize the block length allocation strategy, the UAV position and the channel allocation strategy until the problem reaches the optimum. The application approximately converts a complex mixed integer non-convex original problem into an easy-to-handle problem, and proposes an alternating iterative algorithm, so that the maximization of the minimum effective information amount of the whole system is achieved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of wireless communication, and particularly relates to a resource allocation method for deploying a drone as a relay to assist a base station to communicate with multiple Internet of Things (IoT) devices in a heterogeneous network scenario facing high reliability and low latency (URLLC), and a system implementing the method. BACKGROUND

[0002] Recently, in the Internet of Things scenario, the traditional wired line connecting the central controller to the Internet of Things device is replaced by a wireless network due to low maintenance cost and flexible deployment, however, some key applications have high requirements on the quality of communication, and have high standards for reliability and delay. Ultra-reliable low latency communication (URLLC) is one of the three major pillar applications in the fifth generation network (5G), and its goal is that the network can operate under extremely high reliability standards, and the information transmission delay is less than 1ms. In order to reduce the delay, the channel block length of packet transmission is set to be limited, which will lead to the decline of transmission rate and higher decoding error probability, and at the same time, in such a case, the traditional Shannon formula is no longer applicable, and instead, the reachable data rate expression formula of finite block length information theory is more complex.

[0003] With the development of information technology, the application of Internet of Things devices is also more and more extensive. In remote areas, due to the difficulty of infrastructure coverage or due to high-rise buildings, mountains, etc., the direct link between the base station and the Internet of Things device is unavailable, in such a case, the drone is widely used due to its low price and flexible deployment. It can be used as a mobile base station to directly provide services for Internet of Things devices, or it can be used as a relay to assist the base station to communicate with the Internet of Things devices. Through the reasonable use of the drone, the fixed base station can be effectively converted into a mobile base station, which will improve the network connectivity and expand the coverage of the wireless network. At the same time, since the drone can adjust its flight height and horizontal position, it can ensure direct link communication with ground devices.

[0004] The widespread deployment of the fifth generation network (5G) in the world makes the composition of today's network more complex.

[0005] Therefore, it is necessary to design a resource allocation method and a system implementing the method to solve the above technical problems, and to solve the communication limited problem of the Internet of Things device and the base station caused by the communication infrastructure cost or the sudden natural disaster in the special environment. SUMMARY

[0006] The main purpose of the present application is to propose an optimization method for using a drone relay to assist a base station to transmit high reliability and low latency information, and to obtain the optimal fair and effective information amount.

[0007] To achieve the above object, the application provides a resource allocation method in a UAV-assisted heterogeneous network based on short packet communication, characterized in that it comprises the following steps:

[0008] Step (1), a UAV-assisted heterogeneous network system is constructed, the system comprising a base station, a UAV, a plurality of Internet of Things devices and a plurality of pairs of D2D devices;

[0009] Step (2), an optimization problem is constructed to maximize the minimum amount of effective information received by the Internet of Things devices in the entire system;

[0010] Step (3), the UAV position and the channel allocation strategy are fixed respectively, the block length allocation strategy and the channel allocation strategy, the block length allocation strategy and the UAV position, and the optimization problem in step (2) is converted into three sub-problems; and

[0011] Step (4), a two-layer iterative algorithm is used to jointly optimize the block length allocation strategy, the UAV position and the channel allocation strategy until the problem in step (2) reaches the optimum.

[0012] The further improvement of the application is that the optimization function in step (2) is:

[0013]

[0014] The constraint condition is:

[0015] s.t.L1+L2=L max

[0016]

[0017]

[0018] x min ≤q u [1]≤x max

[0019] y min ≤q u [2]≤y max

[0020] q u [3]==H

[0021]

[0022] Wherein:

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029] where R(γ u / i ) is the achievable rate under finite block length, L max is the total block length, L = {L1, L2}, L1, L2 are the block lengths between the base station and the UAV and between the UAV and the IoT device i, respectively, ε u / i are the decoding error probabilities from the base station to the UAV and from the UAV to the IoT device i, respectively, q u = [x u , y u , z u ] is the position of the UAV, x min , y min , x max , y max are the limits of the UAV coordinates, H is the constant height of the UAV flight, a i denotes the selected channel between the UAV and the IoT device i, denotes the set of channel allocation strategies, a denotes the selected channel of the jth pair of D2D, ψ(a i , a m ), denote whether the IoT device i has selected the same channel as the IoT device m, whether the IoT device i has selected the same channel as the jth pair of D2D, respectively, γ u / i are the signal-to-noise ratios from the base station to the UAV and from the UAV to the IoT device i, respectively, h u , h i , h j are the channel gains between the base station and the UAV, between the UAV and the IoT device i, and between the jth pair of D2D, respectively, p u , p i , p j are the transmit powers of the base station, the UAV, and the jth pair of D2D, respectively, β u , β i , β j are the channel gains between the base station and the UAV, between the UAV and the IoT device i, and between the jth pair of D2D at the reference distance d = 1 meter, respectively, d u , d i , d j are the distances between the base station and the UAV, between the UAV and the IoT device i, and between the jth pair of D2D, respectively, c u = ci = c j = 2 is the path loss parameter of LoS link.

[0030] Further improvement of the present application is that step (3) comprises the following steps:

[0031] Step 3.1, by fixing the UAV position and channel resource strategy, and converting the minimum problem of the optimization problem in step 2 into constraints, replacing the non-convex constraint by first-order Taylor expansion with its corresponding lower bound, and converting the original function into:

[0032]

[0033] The constraint condition is:

[0034]

[0035]

[0036]

[0037] Wherein, A i = (1-ε i )log2(1+γ i ), C u = (1-ε u )log2(1+γ u ), Then take as a whole, according to the monotonicity of the function and the root formula of the quadratic equation, the closed-form solution of block length allocation can be easily obtained:

[0038]

[0039] Step 3.2, by fixing the block length allocation strategy and the channel allocation strategy, and converting the minimum problem of the optimization problem in step (2) into constraints, replacing the non-convex constraint by introducing a slack variable and first-order Taylor expansion, and converting the original function into:

[0040]

[0041] Wherein, is used to process the minimization problem, is a slack variable, which is used to help convert the non-convex function into a convex function;

[0042] Step 3.3, by fixing the block length allocation strategy and the UAV position, using a Stochastic Learning Automaton (SLA) to converge to the optimal solution of the channel allocation strategy of the problem by constantly interacting with the unknown environment and adjusting the action probability through feedback.

[0043] Further improvements of the present application are that step (4) comprises the following steps:

[0044] Step 4.1, initialize the block length allocation strategy UAV position Channel allocation strategy Ψ 0 , iteration number t = 0;

[0045] Step 4.2, loop the following operations until the original optimization problem converges to the specified accuracy:

[0046] (a) Fix the UAV position and the channel allocation strategy, update the original problem using the continuous convex optimization technique, and through the closed-form solution in iteration step (31) until converges;

[0047] (b) Fix the block length allocation strategy and the channel allocation strategy, solve the transformed convex problem in step (32) using the cvx toolbox in Matlab until converges;

[0048] (c) Fix the block length allocation strategy and the UAV position, and use the SLA algorithm to obtain the optimal solution of the channel allocation strategy Ψ t+1 ;

[0049] (d) Update the iteration step, t <- t + 1;

[0050] Step 4.3, output the block length allocation strategy UAV position and channel allocation strategy Ψ * .

[0051] In order to achieve the above application purposes, the present application further provides a resource allocation system in a UAV-assisted heterogeneous network based on short packet communication, which can implement the method of any one of the preceding.

[0052] The beneficial effects of the present application: the resource allocation method and system in the unmanned aerial vehicle assisted heterogeneous network based on short packet communication designed by the present application considers that the unmanned aerial vehicle assisted base station sends ultra-reliable and low-latency control information to the Internet of Things device, and at the same time, there are multiple pairs of devices using device-to-device (D2D) communication in the scene, which constitutes a heterogeneous network with multiple communications. In order to ensure the fairness among multiple Internet of Things devices, the method takes maximizing the minimum effective information amount received by the Internet of Things devices in the whole system as the target, and jointly optimizes the block length allocation strategy, the unmanned aerial vehicle position and the channel allocation strategy through a two-layer iterative algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 The system model diagram of the present application is shown in the figure;

[0054] Figure 2 The block length allocation strategy algorithm flowchart is shown in the figure;

[0055] Figure 3 The unmanned aerial vehicle position algorithm flowchart is shown in the figure;

[0056] Figure 4 The channel allocation strategy algorithm flowchart is shown in the figure;

[0057] Figure 5 The joint resource allocation and unmanned aerial vehicle position algorithm flowchart is shown in the figure;

[0058] Figure 6 The convergence behavior diagram under different decoding error probabilities is shown in the figure;

[0059] Figure 7 The relationship diagram between the minimum effective information amount and the unmanned aerial vehicle flight height is shown in the figure;

[0060] Figure 8 The relationship between the minimum effective information amount and the total block length of the system. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be described in detail below in combination with the drawings and specific embodiments.

[0062] It should be emphasized that in the description of the present application, various formulas and constraints are distinguished by consistent symbols before and after, but different symbols may also be used to mark the same formula and / or constraint. The purpose of this setting is to make the features of the present application clearer.

[0063] As Figure 1 shown, the present application proposes a resource allocation method in an unmanned aerial vehicle assisted heterogeneous network based on short packet communication, which specifically includes the following steps:

[0064] Step (1), constructing a UAV-assisted heterogeneous network system, the system includes base station, UAV, multiple IoT devices and multiple pairs of D2D devices;

[0065] Step (2), constructing an optimization problem aiming at maximizing the minimum amount of effective information received by IoT devices in the whole system;

[0066] Step (3), fixing the UAV position and channel allocation strategy respectively, block length allocation strategy and channel allocation strategy, block length allocation strategy and UAV position, converting the optimization problem in step (2) into three sub-problems;

[0067] Step (4), using a two-layer iterative algorithm to jointly optimize the block length allocation strategy, UAV position and channel allocation strategy until the problem in step (2) reaches the optimal.

[0068] The following embodiments will be described in detail.

[0069] Step (1), constructing a UAV-assisted heterogeneous network system, the system includes base station, UAV, multiple IoT devices and multiple pairs of D2D devices.

[0070] As shown in Figure 1 , such a UAV-assisted heterogeneous network is constructed, in which the communication between the base station (position q b =[x b ,y b ,z b ]) and the IoT devices is hindered, so the UAV is needed as a relay to assist the base station to send control information to the IoT devices, and these information needs to be guaranteed to be ultra-reliable and low-latency. The whole network contains i IoT devices Their positions are q i =[x i ,y i ,z i ], there are j pairs of D2D devices Their positions are q jr =[x jr ,y jr ,z jr ], q js =[x js ,y js ,z js ]. We assume that the base station and the IoT devices and the D2D device pairs are equipped with single antennas, and the UAV is equipped with multiple antennas and can communicate with all IoT devices at the same time. The network adopts frequency division multiple access strategy, and the spectrum is divided into S channels, And D2D and Internet of Things devices occupy and only occupy one channel, if the channel selected by the UAV and the Internet of Things device is the same as the channel selected by the UAV and other Internet of Things devices or the same as the channel occupied by the D2D pair, interference will be generated.

[0071] Step (2), constructing an optimization problem aiming to maximize the minimum effective information amount received by the Internet of Things devices in the whole system.

[0072] In step (2), a mathematical optimization model is established with the block length allocation strategy, the UAV position and the channel allocation strategy as the design variables, aiming to maximize the minimum effective information amount received by the Internet of Things devices in the scene, under the premise of limited total block length, flight area and channel number, and with the constraint condition that the data transmitted by the base station to the UAV is greater than the sum of the data transmitted by the UAV to the Internet of Things devices, that is:

[0073]

[0074] The constraint condition is:

[0075] s.t.L1+L2=L max #(1)

[0076]

[0077]

[0078] x min ≤q u [1]≤x max #(4)

[0079] y min ≤q u [2]≤y max #(5)

[0080] q u [3]==H#(6)

[0081]

[0082] Among them:

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089] where R(γ u / i ) is the achievable rate under finite block length, L max is the total block length, L = {L1, L2}, L1, L2 are the block lengths between the base station and the UAV and between the UAV and the IoT device i, respectively, ε u / i are the decoding error probabilities from the base station to the UAV and from the UAV to the IoT device i, respectively, q u = [x u , y u , z u ] is the position of the UAV, x min , y min , x max , y max are the limits of the UAV coordinates, H is the constant height of the UAV flight, a i denotes the selected channel between the UAV and the IoT device i, denotes the set of channel allocation strategies, with denotes the selected channel for the jth pair of D2D, ψ(a i , a m ), denote whether the IoT device i has selected the same channel as the IoT device m and whether it has selected the same channel as the jth pair of D2D, respectively, γ u / i are the signal-to-noise ratios from the base station to the UAV and from the UAV to the IoT device i, respectively, h u , h i , h j are the channel gains between the base station and the UAV, between the UAV and the IoT device i, and between the jth pair of D2D, respectively, p u , p i , p j are the transmit powers of the base station, the UAV, and the jth pair of D2D, respectively, β u , β i , β j are the channel gains between the base station and the UAV, between the UAV and the IoT device i, and between the jth pair of D2D at the reference distance d = 1 meter, respectively, d u , d i , d j are the distances between the base station and the UAV, between the UAV and the IoT device i, and between the jth pair of D2D, respectively, c u = c i = c j = 2 are the path loss parameters for LoS links.

[0090] Step (3), fixing the UAV location and the channel allocation strategy, the block length allocation strategy and the channel allocation strategy, the block length allocation strategy and the UAV location, respectively, the optimization problem in step (2) is transformed into three sub-problems.

[0091] Since the above problem is a very complex mixed integer non-convex optimization problem, it cannot be handled by conventional methods, so according to each optimization variable, the other two optimization variables are fixed, and it is decomposed into three sub-problems to solve. For the block length allocation strategy and the UAV location, introduce a variable to transform the min problem into a constraint, and then in order to solve the non-convexity of the constraint, we solve it by using continuous convex optimization technology and increasing the relaxation variable. For the channel allocation strategy problem, we design an algorithm based on a stochastic learning automaton (SLA), the specific steps are as follows:

[0092] Step 3.1, fixing the UAV location q u With the channel allocation strategy Ψ, the block length allocation strategy L is optimized by solving the following problem. We use the variable to handle the minimization problem, and introduce the constraint:

[0093]

[0094] Then use continuous convex optimization technology to handle non-convex constraints (3), (8), and transform the block length sub-problem into:

[0095]

[0096] The constraint condition is:

[0097]

[0098]

[0099]

[0100] Where, A i = (1-ε i ) log2 (1+γ i ), C u = (1-ε u ) log2 (1+γ u ), Taking as a whole, then the constraint (11) is a monomial quadratic equation. We let Because then Because C ugreater than D u c1 can be greater than 0. We let and -C u L max write as can be easily obtained by the above method Then c2 is less than 0. The constraint (11) can be written as Because B 2 -4c1c2>0, so The range of L2 can be written as:

[0101]

[0102] Because the goal is to maximize Then The maximum is obtained when the constraint (12) is taken as equality, The constraint (8) is a monotonically increasing function about L2, which means the maximum is obtained when L2 takes the maximum. Then combined with (13), we can get the closed-form solution of the block length allocation strategy:

[0103]

[0104] where, denotes the floor function.

[0105] Step 3.2, fix the block length allocation strategy L and the channel allocation strategy Ψ, we optimize the UAV position q u Similarly, we introduce the variable to handle the minimization, and introduce the constraints:

[0106]

[0107] Because of the constraint (3), (15) is non-convex, so we handle it by introducing a slack variable and a continuous convex optimization technique, which transforms the UAV position subproblem into:

[0108]

[0109] The constraints are:

[0110] (4), (5), (6)

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120] where, Thus the UAV location subproblem is treated as a convex problem which can be solved efficiently by CVX, we use Mosek solver.

[0121] Step 3.3, distribute the strategy L with the UAV location q by fixing block length u Optimize the channel allocation strategy Ψ. Due to the constraints (3), (7), the original problem is a difficult integer non-convex problem which cannot be solved by general methods, we use Stochastic Learning Automaton (SLA), a reinforcement learning algorithm, which through constant interaction with the unknown environment, adjusts the action probability through feedback and finally converges to the optimal action.

[0122] We define P = (P1, P2,..., P I ) is the channel selection probability matrix of the UAV and each IoT device, for P i = (P i,1 , P i,2 ,... P i,S ) is the channel selection probability vector of the UAV IoT device i, P i,s represents the probability of the UAV and IoT device i selecting channel In each iteration, the channel is randomly selected according to the channel selection probability matrix P, and then R i (l) will be calculated according to the actual selected channel a i (l) = L2(1 - ε i )R(γ i ), that is, the effective information amount between the UAV and each IoT device. The reward function is defined as:

[0123]

[0124] The update of the channel selection probability matrix P is as follows:

[0125]

[0126] where η is the learning factor, is the normalized reward, a i,max is the most satisfying channel allocation scheme between UAVs and IoT devices i, which is as follows:

[0127]

[0128] To solve the constraint (3), we design a penalty method to force the channel selection away from the schemes that do not satisfy the constraint, i.e. i ′(l)=-R i (l), then the normalized reward is The update of the probability matrix P is:

[0129]

[0130] In this way, the algorithm will be away from the selection of random to the channel allocation scheme that does not satisfy the constraint.

[0131] Step (4), the algorithm shown in Figure 2 , Figure 3 , Figure 4 is used to iteratively optimize the block length allocation strategy, the UAV position and the channel allocation strategy, respectively, and the joint optimization method shown in Figure 5 is used to alternately optimize the optimal solution of each sub-problem until the problem in step 2 is optimized. The specific steps are as follows:

[0132] 1) Initialize the block length allocation strategy the UAV position the channel allocation strategy Ψ 0 , and the iteration number t = 0;

[0133] 2) Loop the following operations until the original optimization problem converges to the specified accuracy:

[0134] a) Input the block length allocation strategy the UAV position and the channel allocation strategy Ψ t , initialize Loop l1 = 1, 2...: solve the problem P1 by the closed-form solution (14) to obtain until Output

[0135] b) Input the block length allocation strategy the UAV position and the channel allocation strategy Ψ t , initialize Loop l2 = 1, 2...: solve the UAV position sub-problem P2 by cvx to obtain until Output

[0136] c) Input block length allocation strategy with UAV position Initialize η = 0.002, Loop l3 = 1, 2...: First at the beginning of the l3th iteration, a channel selection scheme is randomly generated according to the channel selection probability matrix P, and all effective information amounts are calculated according to the scheme The effective information amount calculated according to formula (16) updates the reward table Q i,s (l3), and the channel selection probability matrix P is updated according to formula (17), (20). Until any channel probability selection vector P i There is an element close to 1 or l3 reaches the maximum number of iterations, output Ψ t+1 = Ψ max .

[0137] d) Update iteration step, t <- t + 1.

[0138] 3) Output block length allocation strategy UAV position and channel allocation strategy Ψ * .

[0139] In MATLAB simulation, a 400x400 square meter scene is assumed, the flight height of the UAV is constant H = 160m, and the base station, IOT device and D2D device are all set to ground level position, i.e. height 0. We assume that the base station is located at (0, 0, 0), which is reasonable and helpful for our calculation. We assume that there are 3 pairs of D2D devices and 5 IOT devices in the network, which are randomly distributed in the scene, and there are only 3 channels available in the scene. The total bandwidth of the system is set to B = 1MHz, the transmission delay of the system is set to T max = 1ms, then the total block length of the system is L max = BT max = 1000, the minimum block length is set to L min = 0.2xL max . The channel power gain β u = β i = β j = 10 -5 , the noise at the UAV and the ground noise are set to -80dBm, i.e. σ u 2 = σ i 2 = -80dBm. The transmission power between the base station, UAV and D2D pair is set to 20dBm, i.e. p u = p i = pj = 20dBm. Since the decoding error probability is not our focus, we set the decoding error probability uniformly as ε u = ε i = 10 -6 This is reasonable.

[0140] Figure 6 The convergence behavior of the algorithm under different decoding error probabilities is studied, and it can be seen that the proposed alternating optimization algorithm converges quickly, and generally three iterations are sufficient for convergence, which means that the complexity of the designed algorithm is low.

[0141] We compared the performance of the fixed block length allocation strategy (the best scheme is adopted for the UAV position and channel allocation strategy), the fixed UAV position (the best scheme is adopted for the block length allocation strategy and channel allocation strategy), the fixed channel allocation strategy (the best scheme is adopted for the block length allocation strategy and UAV position), the exhaustive algorithm and our algorithm under different flight heights and different total block lengths of the system, as shown in Figure 7 , Figure 8 It can be seen that the results of our algorithm and the exhaustive algorithm are almost the same. As can be seen from Figure 7 , with the increase of the UAV height, the minimum effective information amount continues to decrease. This is reasonable because the channel gain decreases with the increase of the distance between the UAV and the ground Internet of Things device. In addition, as can be seen from Figure 8 , with the increase of the total block length of the system, the minimum effective information amount increases continuously, which is also reasonable because the transmitted information increases.

[0142] Based on the above inventive concept, the application further discloses a cross-warehouse facial expression recognition system based on spatiotemporal feature point motion attention and subfield adaptation, comprising at least one computing device, the computing device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the computer program is loaded into the processor, the above-mentioned cross-warehouse facial expression recognition method based on spatiotemporal feature point motion attention and subfield adaptation can be realized.

[0143] The application studies a resource allocation method in a UAV-assisted heterogeneous network based on short packet communication. Because short packet communication technology is adopted to reduce delay to ensure high-quality communication, the traditional Shannon formula is no longer applicable, and the finite block length information theory is adopted. The problem is described as a very complex mixed integer non-convex optimization problem. Through continuous convex optimization technology, the block length allocation sub-problem and the UAV position sub-problem are converted into easy-to-solve convex problems, and a closed-form solution of the block length allocation sub-problem is derived. Finally, the channel allocation sub-problem is solved by a stochastic learning automaton (SLA), and a two-layer iterative algorithm is designed to maximize the minimum effective information amount received by the Internet of Things device in the scene.

[0144] The above embodiments are only used to illustrate the technical solutions of the present application, not limit the present application. Even though the present application has been described in detail with reference to the preferable embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced, without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A resource allocation method in a UAV-assisted heterogeneous network based on short packet communication, characterized in that: Includes the following steps: Step (1): Construct a drone-assisted heterogeneous network system, which includes base stations, drones, multiple IoT devices, and multiple pairs of D2D devices; Step (2): Construct an optimization problem with the goal of maximizing the minimum amount of effective information received by IoT devices in the entire system; Step (3): Fix the UAV position and channel allocation strategy, block length allocation strategy and channel allocation strategy, and block length allocation strategy and UAV position respectively, and transform the optimization problem in step (2) into three sub-problems; as well as Step (4): Use a two-layer iterative algorithm to jointly optimize the block length allocation strategy and the UAV position and channel allocation strategy until the problem in step (2) is optimal; Step (3) includes the following steps: Step 3.1: By fixing the UAV position and channel resource strategy, and transforming the optimization problem in Step 2 into a minimum problem, the non-convex constraints are replaced with their corresponding lower bounds through a first-order Taylor expansion. The original function is then transformed into: The constraints are: Where t is the number of iterations. A i =(1-ε i log2(1+γ) i ), C u =(1-ε u log2(1+γ) u ), Then Treating it as a whole, the closed-form solution for block length distribution can be easily obtained based on the monotonicity of the function and the quadratic formula: Step 3.2: Using a fixed block length allocation strategy and a channel allocation strategy, and transforming the optimization problem in step (2) into a minimum problem, the non-convex constraint is obtained by introducing slack variables and a first-order Taylor expansion. The original function is then transformed into: Among them, R γf Used to handle minimization problems, x1, x2, x3 i,m x4 i x5 i,m x6 i x7 i These are slack variables used to help transform non-convex functions into convex functions; Step 3.3: Using a fixed block length allocation strategy and the UAV's location, a stochastic learning automaton (SLA) is employed to continuously interact with the unknown environment and adjust the action probability through feedback until it converges to the optimal action, thus obtaining the optimal solution for the channel allocation strategy of the problem.

2. The method according to claim 1, characterized in that: The optimization function in step (2) is: The constraints are: stL1+L2=L max x min ≤q u [1]≤x max and min ≤q u [2]≤y max q u [3]==H in: Wherein, R(γ) u R(γ) represents the achievable rate from the base station to the drone with a finite block length. i Let L be the achievable speed from the drone to the i-th IoT device with a finite block length. max Let L be the total block length, L = {L1, L2}, where L1 and L2 are the block lengths between the base station and the drone, and between the drone and IoT device i, respectively. u / i Let q represent the decoding error probabilities from the base station to the drone and from the drone to the i-th IoT device, respectively. u =[x u ,y u ,z u [x] represents the location of the drone. min y min x max y max Here, H represents the limit value of the UAV's coordinates, H is the constant altitude of the UAV's flight, and a is the limit value of the coordinates. i This indicates the channel selected between the drone and the IoT device i. Represents the set of channel allocation strategies, using Let ψ(a) represent the channel selected in the j-th pair of D2D operations. i ,a m ), These respectively indicate whether IoT device i and IoT device m have selected the same channel, whether i and j-th D2D pair have selected the same channel, and γ. u γ represents the signal-to-noise ratio from the base station to the drone. i h represents the signal-to-noise ratio from the drone to the i-th IoT device. u h i h j These represent the channel gains between the base station and the drone, between the drone and IoT device i, and between the j-th D2D pair, respectively. u p i p j These represent the transmit power of the base station, the drone, and the D2D pair, respectively, β. u ,β i ,β j These represent the channel gains at a reference distance d = 1 meter between the base station and the drone, between the drone and IoT device i, and between the j-th D2D pair. u d i d j These are the distances between the base station and the drone, between the drone and IoT device i, and between the j-th D2D pair, c. u =c i =c j =2 is the path loss parameter for the Loss of Path (LoS) link.

3. The method according to claim 2, characterized in that: Step (4) includes the following steps: Step 4.1: Initialize the block length allocation strategy drone location Channel allocation strategy Ψ 0 The number of iterations t = 0; Step 4.2: Repeat the following operations until the original optimization problem converges to the specified accuracy: (a) Fix the UAV position and channel allocation strategy, update the original problem using continuous convex optimization techniques, and iterate through the closed-form solution in step (31) until... convergence; (b) Fixed block length allocation strategy and channel allocation strategy, solved using the convex problem transformed in step (32) and the cvx toolbox in Matlab, until... convergence; (c) Using a fixed block length allocation strategy and UAV location, the optimal channel allocation strategy Ψ is obtained using the SLA algorithm. t+1 ; (d) Update iteration steps, t←t+1; Step 4.3, Output Block Length Allocation Strategy drone location and channel allocation strategy Ψ * .

4. A resource allocation system in a drone-assisted heterogeneous network based on short packet communication, the system comprising a base station, drones, multiple IoT devices, and multiple pairs of D2D devices, and capable of implementing the method as described in any one of claims 1 to 3, wherein, The drone is equipped with multiple antennas, while the base station, IoT device and D2D device are equipped with a single antenna. The system also includes a block length optimization module, a drone position optimization module and a channel allocation module. The block length optimization module is used to perform the closed-form solution iterative calculation in step 3.1, the drone position optimization module is used to perform the CVX solution process in step 3.2, and the channel allocation module is used to perform the SLA algorithm in step 3.3.

Images