CN112996121B - U2U distributed dynamic resource allocation method for intra-cluster communication - Google Patents

Abstract

The invention provides a U2U distributed dynamic resource allocation method facing to intra-cluster communication, which comprises the steps of constructing an IaS-U2U communication scene model and a communication flow; constructing an objective function of IaS-U2U communication to obtain a multi-objective optimization problem of the IaS-U2U communication; decomposing the multi-objective optimization problem, and carrying out a UAV matching process by using the H-UAV and the R-UAV by adopting a classical GS algorithm to complete UAV pairing; according to the UAV pairing result, completing the position optimization of the R-UAV by adopting an SPO algorithm, and completing the solution of a multi-objective optimization problem; and according to the solving result, completing the distribution of the U2U distributed dynamic resources. The invention provides a U2U distributed dynamic resource allocation method facing to intra-cluster communication, provides a distributed U2U communication flow, is suitable for IaS-U2U transmission in an off-line state, and overcomes the defects that actual conditions such as UAV (unmanned aerial vehicle) propulsion power neglect, UAV interference neglect and QoS (quality of service) constraint neglect in the existing U2U literature.

Description

U2U distributed dynamic resource allocation method for intra-cluster communication

Technical Field

The invention relates to the technical field of communication, in particular to a U2U distributed dynamic resource allocation method for intra-cluster communication.

Background

With the rapid development of communication technology, people can enjoy high-speed and reliable mobile communication services in most land areas. However, in remote areas where the eNB cannot effectively cover, and in a scenario where the number of users in a cell is overloaded, the eNB is not sufficient to meet the communication requirements of all users. The UAV plays a remarkable role in expanding the wireless coverage of the eNB and improving the communication quality of cell users.

In recent years, UAVs are also gradually being put into new application scenarios and cover different needs of military and commercial applications. The role played by UAVs is also gradually changing from relaying assisted ground communications to users of air or ground network cells. UAVs have a variety of application scenarios including building planning, rescue implementation, cargo transportation, real-time monitoring, etc. The types of services supported by UAVs are also gradually going from singular to plural, making the industry increasingly demanding communications for UAVs. However, UAVs are cell users and are not necessarily able to achieve high communication quality. For example, when the UAV is responsible for ground monitoring and detection work, it may be in an area with poor signal coverage, and it is difficult for the UAV to upload acquired video information to the eNB in time. In addition, when search and rescue actions are performed in remote areas, the UAVs may be in an off-line state, and communication between the UAVs is not possible, so that it is difficult for the UAV network to perform search and rescue activities normally. The U2U technology can reduce the load of the eNB and increase the capacity of the system, and thus becomes one of the key technologies for improving the UAV communication quality.

Due to the limited functionality of UAVs, it is difficult to support the above-mentioned complex and burdensome business requirements with only a single UAV. Compared with a traditional single UAV, the UAV cluster can efficiently and quickly complete various complex tasks. For example, in military scenarios, UAVs often perform patrol and monitoring tasks in a cluster. Therefore, in order to cope with the trend of UAV system clustering, clustering UAVs become one of the important application scenarios considered by the U2U communication technology.

Another key issue in supporting U2U communications is the limited onboard energy of the UAV. In order to realize reliable communication of U2U and prolong flight time of UAV cluster, the energy resource of UAV cluster needs to be effectively distributed. Unlike ground based devices, UAVs generate additional propulsion energy costs to maintain flight and support their movement. Therefore, the optimization of the propulsion energy consumption also becomes a problem to be considered in the optimization design of the U2U communication system.

Many studies both at home and abroad take into account the communication and energy consumption problems of the UAVs. Among them, qia et al [1] C.Qiu, Z.Wei, Z.Feng, and P.Zhang, "Backhau-aware projection optimization of fixed-wire UAV-mounted base station for coherent available wireless Service," IEEE Access, vol.8, pp.60940-60950,2020, doi. However, this study only focuses on the case where the UAV hovers around a ground eNB, and the propulsion model is difficult to apply directly to a rotorcraft UAV. Although rotorcraft UAVs have a completely different propulsion principle than fixed wing UAVs, they are more flexible. Aiming at the application scene of the rotor UAV, zeng et al [2] Y.Zeng and R.Zhang, "Energy-efficiency UAV communication with project optimization," IEEE Transactions on Wireless Communications, vol.16, no.6, pp.3747-3760, jun.2017, doi. Zhan et al [3] C.Zhan and H.Lai, "Energy minimization in internet-of-ings system based on personnel-wing UAV," IEEE Wireless Communications Letters, vol.8, no.5, pp.1341-1344, oct.2019, doi. Ahmed et al [4] s.ahmed, m.z.chowdour, and y.m.jang, "Energy-influencing UAV-to-user scheduling to maximum throughput in wireless networks," IEEE Access, vol.8, pp.21215-21225,2020, doi. It is noted that documents [2] to [4] both consider only the problem of position deployment or trajectory optimization of a single UAV, thus assuming that the UAV does not suffer interference from the rest of the equipment. However, in practical scenarios, the interference strength of the UAV varies as its location varies. In addition, in a cluster scene, if co-frequency networking is considered, the UAV cannot ignore interference from other UAVs in the same cluster. Therefore, in optimizing the UAV trajectory, it is more realistic to consider the impact of the interference. Based on this, liu et al [5] A.Liu and V.K.N.lau, "Optimization of multi-UAV-aided Wireless network a ray-routing channel model," IEEE Transactions on Wireless Communications, vol.18, no.9, pp.4518-4530, sep.2019, doi. The algorithm may select a suitable location to deploy the UAV based on the interference of the ground user. However, this solution only optimizes the hover position of the UAV, does not optimize the trajectory of the UAV over a certain period of time, and ignores the energy consumption issue of the UAV during flight. It is noted that, most of the above schemes only focus on UAV assisted wireless communication scenarios, and are difficult to be directly applied to cluster U2U communication scenarios.

Currently, only a small amount of research is directly focused on U2U communication scenarios. Among them, bellido-mangannell et al analyzed the potential application scenario of the U2U technology and discussed the applicability of different media access schemes, forward error correction coding and modulation schemes to U2U data link design. Ranjha et al studied the problem of multi-hop UAV relay links delivering short ultra-reliable and low-latency (URLLC) instruction packets between Internet of Things (IoT) devices on the ground, and used a non-linear optimization method to minimize the probability of erroneous decoding. However, such schemes do not involve the resource allocation problem of the Media Access Control (MAC) layer. Du et al [6] B.Du, R.Xue, L.ZHao, and V.C.M.Leung, "cosmetic mapping for air-to-air and air-to-ground cognitive mapping," IEEE Transactions on Audio and Electronic Systems, doi:10.1109/TAES.2019.2958162. The problem of spectrum sensing and sharing between U2U pairs and cell users was studied assuming that U2U pairs share the same frequency resources with cell users as the secondary users of the cell. However, in a cluster UAV scene, multiple UAVs compete for the same channel in a spectrum sensing and sharing manner, so that good fairness is not easily ensured, and a higher communication service requirement of a cluster UAV system is difficult to meet. Zhang et al [7] S.Zhang, H.Zhang, B.Di, and L.Song, "Cellular UAV-to-X Communications: design and optimization for multi-UAV networks," IEEE Transactions on Wireless Communications, vol.18, no.2, pp.1346-1359, feb.2019, doi. However, this study only optimizes pairing between UAVs and does not involve optimization of trajectory and launch power of the UAVs. In practical scenarios, optimizing for the trajectory of the UAV may further improve the performance of the system. Wang et al [8] H.Wang, J.Wang, G.Ding, J.Chen, and J.Yang, "Completion time minimization for turning angle-constrained UAV-to-UAV communications," IEEE Transactions on Vehicular Technology, vol.69, no.4, pp.4569-4574, april 2020, doi. However, this solution only considers trajectory optimization for one U2U pair, neglecting interference from the rest of the UAVs, and does not analyze the propulsion energy consumption problem during flight. It should be noted that the documents [6] to [8] strongly depend on the centralized processing mechanism of the network node such as eNB, but it is difficult to find a centralized node to support the above scheme when the cluster is in the off-line state. Therefore, in a cluster U2U communication scenario, a distributed algorithm needs to be designed to deal with the U2U communication problem in an off-line state.

Disclosure of Invention

The invention provides a U2U distributed dynamic resource allocation method facing to intra-cluster communication, aiming at overcoming the technical defect that when a cluster is in an off-line state in the conventional cluster U2U communication scene, a centralized node is difficult to find to support communication.

In order to solve the technical problems, the technical scheme of the invention is as follows:

a U2U distributed dynamic resource allocation method facing to intra-cluster communication comprises the following steps:

s1: constructing an IaS-U2U communication scene model and a communication flow, wherein three kinds of UAVs are considered, including a central UAV (user equipment), namely a C-UAV, in a cluster central position, an assistant UAV (user equipment), namely an H-UAV, for actively providing U2U communication service and a requester UAV, namely an R-UAV, requiring U2U service;

s2: constructing an objective function of IaS-U2U communication to obtain a multi-objective optimization problem of the IaS-U2U communication;

s3: decomposing the multi-objective optimization problem, and carrying out a UAV matching process by using the H-UAV and the R-UAV by adopting a classical GS algorithm to complete UAV pairing;

s4: according to the UAV pairing result, completing the position optimization of the R-UAV by adopting an SPO algorithm, and completing the solution of a multi-objective optimization problem;

s5: and according to the solving result, completing the distribution of the U2U distributed dynamic resources.

In the scheme, the invention constructs a multi-objective optimization problem of transmission rate and propulsion power, and jointly optimizes parameters such as UAV matching and position. In order to solve the multi-objective optimization problem, the design provides a distributed communication process of the UAVs in the cluster, and the optimization problem is decomposed into a UAV pairing and position optimization process:

firstly, forming effective U2U pairing based on similarity of content characteristics among UAVs;

secondly, in order to solve the position optimization problem, the design provides a distributed algorithm based on Sequential Convex Approximation (SCA). The algorithm is divided into two stages, wherein the first stage constructs the problem of the maximum transmission rate, and a feasible solution meeting the constraint of the original problem is obtained by adopting an SCA algorithm according to the initial position of the UAV as a starting point; and in the second stage, the feasible solution obtained in the first stage is used as a starting point, and the local optimal solution of the original problem is obtained by adopting an SCA algorithm.

In the above solution, the present invention provides an SCA-based Position Optimization (SPO) solution in a cluster; the contributions of this scheme mainly include: aiming at an IaS-U2U communication scene which is considered less, different from a mechanism that the existing U2U document mainly depends on a centralized node, a distributed U2U communication flow is provided, and the distributed U2U communication flow is suitable for IaS-U2U transmission in an off-of-Coverage (OOC) state; the defects that actual conditions such as UAV propelling power, interference among UAVs, qoS constraint and the like are ignored in the existing U2U literature are overcome, the influence of UAV mobility inside a cluster on performance is concerned, the transmission rate and the propelling power of the cluster UAV are considered in a combined mode, a multi-objective optimization problem with QoS constraint is constructed, and UAV matching and position are optimized in a combined mode; the maximum transmission rate problem is constructed and solved, and a starting point is provided for iterative solution of the multi-objective optimization problem by judging whether the multi-objective optimization problem has a feasible solution, so that a feasible starting point can be provided for an SCA algorithm; simulation results show that compared with other existing schemes, the scheme can obtain remarkable performance gain.

Compared with the prior art, the technical scheme of the invention has the beneficial effects that:

the invention provides a U2U distributed dynamic resource allocation method facing intra-cluster communication, provides a distributed U2U communication process, is suitable for IaS-U2U transmission in an off-line state, and overcomes the defects that actual conditions such as UAV (unmanned aerial vehicle) propulsion power neglect, UAV interference neglect and QoS (quality of service) constraint neglect in the conventional U2U document.

Drawings

FIG. 1 is a schematic flow diagram of the process of the present invention;

fig. 2 is a schematic view of an IaS-U2U communication scenario in an embodiment of the present invention;

fig. 3 is a schematic diagram of an IaS-U2U communication flow in an embodiment of the present invention;

fig. 4 is a schematic view of design idea of an IaS-U2U communication scheme in an embodiment of the present invention;

FIG. 5 is a schematic diagram showing a comparison of positions before and after IaS-U2U communication optimization;

FIG. 6 is a schematic diagram of the impact of different cluster speeds and UAV protection distances on system performance;

fig. 7 is a graph comparing energy efficiency performance for different schemes.

Detailed Description

The drawings are for illustrative purposes only and are not to be construed as limiting the patent;

for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;

it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.

The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.

Example 1

As shown in fig. 1, a method for allocating U2U distributed dynamic resources for intra-cluster communication is provided, which includes the following steps:

s1: constructing an IaS-U2U communication scene model and a communication flow, wherein three kinds of UAVs are considered, including a central UAV (user equipment), namely a C-UAV, in a cluster central position, an assistant UAV (user equipment), namely an H-UAV, for actively providing U2U communication service and a requester UAV, namely an R-UAV, requiring U2U service;

s2: constructing an objective function of IaS-U2U communication to obtain a multi-objective optimization problem of the IaS-U2U communication;

s3: decomposing the multi-objective optimization problem, and carrying out a UAV matching process by using the H-UAV and the R-UAV by adopting a classical GS algorithm to complete UAV pairing;

s4: according to the UAV pairing result, completing the position optimization of the R-UAV by adopting an SPO algorithm, and completing the solution of a multi-objective optimization problem;

s5: and according to the solving result, completing the distribution of the U2U distributed dynamic resources.

In the specific implementation process, the invention focuses on the resource allocation scheme design of a cluster U2U communication (Swarm U2U, S-U2U) system, and particularly takes air communication as a main application scenario. Specifically, the invention focuses on a U2U (Intra-Swarm U2U, iaS-U2U) communication scene in a cluster, and designs a corresponding U2U communication resource scheduling algorithm by jointly considering the UAV cluster propulsion power.

In the specific implementation process, the method constructs a multi-objective optimization problem of transmission rate and propulsion power, and jointly optimizes parameters such as UAV matching and position. In order to solve the multi-objective optimization problem, the design provides a distributed communication process of the UAVs in the cluster, and the optimization problem is decomposed into a UAV pairing and position optimization process:

firstly, forming effective U2U pairing based on similarity of content characteristics among UAVs;

secondly, in order to solve the position optimization problem, the design provides a distributed algorithm based on Sequential Convex Approximation (SCA). The algorithm is divided into two stages, wherein the first stage constructs the problem of the maximum transmission rate, and a feasible solution meeting the constraint of the original problem is obtained by adopting an SCA algorithm according to the initial position of the UAV as a starting point; and in the second stage, the feasible solution obtained in the first stage is used as a starting point, and the local optimal solution of the original problem is obtained by adopting an SCA algorithm.

In a specific implementation process, the invention provides an SCA-based Position Optimization (SPO) scheme based on SCA in a cluster; the contributions of this scheme mainly include: aiming at an IaS-U2U communication scene which is considered less, different from a mechanism that the existing U2U document mainly depends on a centralized node, a distributed U2U communication flow is provided, and the distributed U2U communication flow is suitable for IaS-U2U transmission in an off-of-Coverage (OOC) state; the defects that actual conditions such as UAV propelling power, interference among UAVs, qoS constraint and the like are ignored in the existing U2U literature are overcome, the influence of UAV mobility inside a cluster on performance is concerned, the transmission rate and the propelling power of the cluster UAV are considered in a combined mode, a multi-objective optimization problem with QoS constraint is constructed, and UAV matching and position are optimized in a combined mode; the maximum transmission rate problem is constructed and solved, and a starting point is provided for iterative solution of the multi-objective optimization problem by judging whether the multi-objective optimization problem has a feasible solution, so that a feasible starting point can be provided for an SCA algorithm; simulation results show that compared with other existing schemes, the scheme can obtain remarkable performance gain.

Example 2

More specifically, on the basis of embodiment 1, as shown in fig. 2, the present invention studies the resource allocation problem of IaS-U2U communication, and in the IaS-U2U communication scenario model of step S1, three types of UAVs are considered, namely, a central UAV (Center-UAV, C-UAV) at a cluster Center position, an assistor UAV (Helper-UAV, H-UAV) that actively provides U2U communication service, and a Requester UAV (Requester-UAV, R-UAV) that requires U2U service. Wherein the C-UAV is always at a radius of r _sw For coordinating and governing movement of the remaining UAVs within the cluster. The rest UAVs are randomly distributed in the cluster range at the initial time, and the distance between the UAVs is not less than the preset protection distance

Considering the general case in an aviation scenario, we assume that the UAV cluster is operating in the OOC state, but when necessary, the C-UAV may access the satellites to acquire the routing information. In this scenario, the R-UAV needs to obtain the data packet from the H-UAV buffer space, and in order to obtain better transmission conditions, it needs to complete the optimization of the relative position and movement trajectory within the cluster.

It is assumed that the distance between clusters is long and thus the interference between clusters is small and negligible. The UAVs in the same cluster reuse the same frequency band, and mutual interference exists. Furthermore, due to the close distance between the UAVs within the cluster, it is assumed that the R-UAVs can each be in one slot δ _intra The movement of the inner completion position. Assume that there are M H-UAVs and N R-UAVs in the UAV cluster, with their ordinal numbers forming the sets M = {1,2, ·, M.,..., M } and Ν = {1,2,.., N.,. N }, respectively. Furthermore, let w _i ＝[x _i ,y _i ] ^T Denotes the coordinate vector of UAVi relative to C-UAV, and let W _H-UAV ＝[w _m ] _m∈M And W _R-UAV ＝[w _n ] _n∈N Representing matrices formed by coordinate vectors of the H-UAV and R-UAV, respectively. Assume that all H-UAVs and R-UAVs within the cluster can acquire δ from the intended flight path or C-UAV's temporary notification message _intra Absolute coordinates of C-UAV after slot completion

Then any UAV i can be based on its relative coordinates w within the cluster _i To calculate the corresponding absolute coordinates

Namely, the method comprises the following steps:

for simplicity of derivation, the present solution assumes that all UAVs inside the cluster are at the same flight altitude. Therefore, for

And i ≠ j, available d _i,j ＝||w _i -w _j I tableAnd (3) indicating the distance from the UAVi to the UAVj, wherein | | · | | is a two-norm operator.

In the IaS-U2U communication scenario, it is assumed that LOS channels are used among UAVs, doppler frequency shift can be accurately estimated, and the influence caused by Doppler can be completely offset [6]][9]M.c. erturk, j.haque, w.a.moreno, and h.arslan, "Doppler assignment in OFDM-based airborne communications," IEEE trans.aerosp.electron.syst., vol.50, no.1, pp.120-129,2014. Therefore, the channel model of the design adopts a free space path loss model [6]][7][8]. Then H-UAvm and R-UAVn are in slot delta _intra The inner channel gain can be expressed as:

where β and α represent the channel coefficient and the path loss index, respectively.

Furthermore, as previously mentioned, rotorcraft UAVs are more flexible than fixed-wing UAVs [2]Hence, the optimization is presented herein using a model of propulsion power for a rotorcraft UAV as an example. Propulsive power P of rotorcraft UAV _p Is a function of the magnitude of the velocity V of the UAV, denoted as [2]：

Wherein: p _bla Representing blade power; u shape _tip Representing tip speed; d _rat Representing fuselage drag ratio; ρ represents an air density; s _rot Represents the rotor volume; a represents the rotor disk area; p _ind Represents the induced power; v. of _rot Mean rotor induced velocity is indicated. According to [2]]It can be seen that the propulsive power of the rotorcraft UAV consists of three parts: blade power, induced power, and parasitic power. The blade power and the induced power need to overcome the blade resistance and the fuselage resistance and are respectively in direct proportion to the square and the third power of the speed V; induced power is the power required to overcome the blade induced drag, which decreases as the speed V increases. Thus, the propulsive power of the rotorcraft UAV appears to decrease first with increasing speed VAnd then increases with the increase of the speed V. Thus, there is a speed value, denoted V, of the rotorcraft UAV that minimizes the propulsive power _MP . Since the propulsion power of a single UAV is typically on the order of hundreds of watts [2]]It can be assumed that the transmit power is much less than the propulsion power.

More specifically, in the step S1, as shown in fig. 3, the IaS-U2U communication flow is briefly described as follows:

first, each R-UAV broadcasts a U2U request [10]3GPP TS 24.334Proximaty-services (ProSe) User Equipment (UE) to ProSe function protocol transactions to each H-UAV; stage 3 (Release 16). Https:// www.3gpp.org/ftp/Specs/archive/24. Servers/24.334/, 2020. And its own content feature information. Each H-UAV broadcasts its own content characteristic information upon receiving the request. Each R-UAV then performs a distributed pairing with each H-UAV based on a default H-UAV transmit power (e.g., which may be assumed to be full power) and content characteristics published by the H-UAV;

subsequently, each R-UAV successfully paired initiates a position optimization application to the C-UAV, and the C-UAV starts a position optimization process for the R-UAV in a polling mode, specifically:

when polling to a particular R-UAV, the C-UAV sends a list of location information to the R-UAV, the list containing location information for all UAVs in the cluster;

the R-UAV executes an SPO algorithm according to the position information list to carry out position optimization;

then, if the optimization is successful, the R-UAV feeds back an optimization result to the H-UAV and the C-UAV which are matched with the R-UAV, and then a position moving process is executed; if the optimization fails, feeding back an optimization failure identifier to the H-UAV, and waiting for the next round of matching by the R-UAV and the H-UAV;

the R-UAV calculates the absolute coordinates thereof according to the optimization result and carries out position movement;

the R-UAV informs the H-UAV paired with it that the position movement process is over.

Finally, each R-UAV establishes a U2U connection with its respective paired H-UAV and performs U2U communications.

More specifically, based on the IaS-U2U communication scene,

more specifically, in step S2, according to the communication process in step S1, considering that joint optimization is performed on two aspects of the transmission rate and the propulsion power of the trunking system, an objective function of IaS-U2U communication is constructed, specifically:

the pairing of the R-UAV and the H-UAV needs to be done before the R-UAV changes its own position and establishes a connection with the H-UAV. Defining a binary matrix X _M×N Element x thereof _m,n =1 denotes H-UAVm paired with R-UAVn, x _m,n And =0 represents unpaired, where m ∈ m and n ∈ Ν. Assuming that each H-UAV serves at most 1R-UAV, each R-UAV is served by at most one H-UAV, i.e.:

in addition, let

Denotes the initial coordinate of R-UAVn relative to C-UAV, w _n Indicating the coordinates of the next slot of the R-UAVn relative to the C-UAV. Since the present solution only optimizes the position of the R-UAV, it can be assumed that the velocities of the C-UAV and H-UAV are the same and defined as the velocity of the cluster, expressed as a vector

Wherein

Representing the velocity components of the cluster in the x-direction and y-direction, respectively. In summary, the velocity magnitude of R-UAV n can be represented by its two-norm velocity vector as:

wherein,

representing the relative speed of R-UAV n with respect to the cluster. Based on equation (2), the propulsive work of R-UAV n can be determinedRate w _n Expressed as:

suppose each H-UAV transmits at full power, and the transmit power of H-UAV m is noted as

The transmission rate of H-UAV m to R-UAV n may then be expressed as:

wherein B and ε ² Representing the channel bandwidth and noise power spectral density, respectively.

Based on the analysis, the formula (5) and the formula (6) can be combined to construct an objective function of IaS-U2U communication, namely:

wherein: theta _n ∈(0,1]Representing a preference factor of the R-UAVn for propulsion power, used for adjusting the proportion of the propulsion power in the multi-objective problem, theta _n The higher the ratio is, the more important the R-UAVn is to the propulsion power problem; positive real number

Has the unit of Mbit s ^-1 ·W ^-1 For imparting propulsion power P _p (w _n ) Is adjusted to be equal to the transmission rate R _intra (w _n ) The same order of magnitude and keeping the adjusted terms and transmission rate the same dimension.

Further, for the convenience of expression, the following variables are defined based on equations (5) and (6):

·

represents the sum of all R-UAV propulsion powers in the cluster:

·R ^tot represents the sum of all H-UAV transmission rates in the cluster:

·μ _intra representing the energy efficiency of an IaS-U2U communication System [11]A.Zappone and E.Jorswieck,"Energy efficiency in wireless networks via fractional programming theory,"Found Trends Commun.Inf.Theory,vol.11,nos.3-4,pp.185-396,2015：

To sum up, the multi-objective optimization problem of IaS-U2U communication can be expressed as:

P1:

limited by:

wherein:

constraints C1 and C2 are pairing constraints for H-UAV and R-UAV;

constraint C3 indicates that the flight speed of the R-UAV should not be higher than the UAV maximum flight speed V _max ；

Constraint C4 is the anti-separation cluster constraint, which means that the optimized relative coordinates of any R-UAV are still at radius R _sw Within the cluster formation range of (a);

constraint C5 is a QoS constraint, representing an arbitrary pairTransmission rate between H-UAV and R-UAV of work not lower than given

Constraints C6 and C7 are collision avoidance constraints, indicating that the distance between any R-UAV and any H-UAV and C-UAV (whose location is the origin of coordinates) should be greater than

In the specific implementation process, it is noted that the concavity and the convexity of the objective function of P1 are difficult to determine, the optimization variable X is a discrete variable, and the constraint conditions C1-C2 and C5-C7 are non-convex constraints. It can be seen that P1 is a Mixed-Integer Non-Linear Programming (MINLP) problem, and the calculation difficulty for directly solving the problem is large. Therefore, the original problem needs to be decomposed, and a distributed algorithm is designed to solve the problem.

More specifically, in order to effectively solve P1, the solution process of P1 is decomposed into the following two steps: UAV matching and position optimization. The UAV matching process is realized by an H-UAV and an R-UAV by adopting a classic GS algorithm [12] D.Gusfield and R.Irving.the stable margin protocol.

Before the pairing process is started, the H-UAV and the R-UAV acquire content classification information of each other in a broadcasting mode. Subsequently, each UAV calculates content similarity [13 ] from Jaccard coefficients]Niwattanaakul, j.singthongchai, e.naenudorn, and s.wanapu, "Using of Jaccard coefficients for keywords similarity," in proc.int.multi conf.eng.com.sci., mar.2013, vol.i, pp.380-384, and generate respective preference lists. In order to enable the position optimization result after pairing to meet the constraint condition C5, candidate pairings need to be screened. The specific method comprises the following steps: the R-UAV n measures the received Signal-to-Noise-and-Interference Ratio (SINR) for each H-UAV, which is reported as

If it is

Then R-UAV n deletes H-UAV m from its list of preferences, where

Representing the SINR threshold of R-UAV n, the expression is as follows:

after each R-UAV modifies the preference list according to the SINR threshold, the classical GS algorithm [18 ]]Namely, the one-to-one matching between the R-UAV and the H-UAV can be realized, and a matching matrix X is obtained ^* . The serial numbers of the H-UAV and R-UAV which are successfully paired are respectively formed into new sets M 'and N'. And after the matching is finished, optimizing the position of the R-UAV.

More specifically, the step S4 specifically includes the following steps:

s41: converting P1 into an optimization problem P2 according to the pairing result;

s42: equating the optimization problem P2 to be an optimization problem P3 according to the property 1, and relaxing and approximating the P3 to obtain a convex problem P5;

s43: constructing a maximum transmission rate problem P6 according to the optimization problem P2, and obtaining an optimization problem P7 by adopting the same relaxation and approximation methods of P3-P5;

s44: and solving P7 and P5 by stages according to the SPO algorithm to complete the position optimization of the R-UAV, and solving the multi-objective optimization problem.

In the specific implementation, in the step S41, for convenience of description, the transmission power of each H-UAV in m' is formed into a vector

Each R-UAV that is successfully paired will be location optimized based on the power value. Based on equation (6), for a successfully paired R-UAV n, the achievable transmission rate is:

where m represents the H-UAV Serial number paired with R-UAV n, i.e., x _m,n =1. Further, when UAV pairing is complete, the objective function in equation (7) will degenerate as:

wherein,

representing the objective function of R-UAV n. Therefore, for each R-UAVn, the following optimization problem needs to be solved:

P2:

limited by:

wherein C8-C12 are constrained C3-C7 of P1 and w _n The associated constraints.

In the specific implementation process, by paying attention to step S41, the objective function of the optimization problem P2 is non-concave, and the constraints C10 to C12 are non-convex constraints, which makes it difficult to solve the original problem; therefore, in step S42, the objective function and the constraint condition of P2 need to be relaxed, and then the SCA algorithm is adopted to solve the problem; the SCA algorithm solves the original problem in an iterative mode, and the optimal solution obtained by the last iteration is needed in each iterative solution; definition K =0,1,2 _max -1 represents the number of iterations of the SCA algorithm, where K _max Representing the maximum number of iterations allowed by the algorithm; particularly, when the first iteration solution is carried out, a starting point needs to be provided for an SCA algorithm; the starting point is any feasible solution which meets the constraint condition of the original problem, and the index k =0;

to this end, a relaxation variable [2] is first introduced:

using equation (18), equation (5) can be approximated as follows:

it is noted that | | v _n || ² And v _n || ³ Are all convex functions, tau is an affine function, thus

Is a convex function [20 ]]. Will be provided with

After substituting P2 and introducing the relaxation variable τ, P2 can be transformed into the following optimization problem:

P3:

limited by: C8-C13, wherein:

based on the above analysis, the following properties are present:

properties 1: the optimization problem P3 is equivalent to P2.

And (3) proving that: when in use

In order to maximize the objective function of P3, the value of τ can be reduced, i.e. the propulsion power in equation (21) is reduced

Thereby increasing the objective function P3And (4) taking values. When τ continues to decrease to C13, i.e., when the equal sign of equation (18) holds, equation (19) degenerates to equation (5), i.e., there is at this time

At this time, the optimization problem P3 is equivalent to P2, and is verified.

Since C13 in equation (18) is still a non-convex constraint, further operation is required. After squaring both sides of equation (18) and taking the reciprocal, we can get:

the two equations in equation (22) add, and C13 can be equivalently constrained as follows:

C14：

since C14 is still a non-convex constraint, the left end of C14 may be relaxed. For C14, i.e. τ in formula (23) ² And

at τ = τ, respectively ^(k) And

is subjected to a first order Taylor expansion, wherein (·) ^(k) Representing the value obtained for the k iteration of a certain variable, there are:

where · is the vector inner product operator. The two equations in equation (24) are added to obtain:

according to [2]]The relaxation variable value formula (4) is used to give a given taskIntention is that

After, τ ^(k) And

can be calculated from the following formula:

based on the above processing, the non-convex constraint C14 may relax into a convex constraint C15:

C15:

further, by using | | | | a | | | b | | | > or more than a · b, the left sides of the inequalities of C11 and C12 can be relaxed as follows:

recouple (28), C11 and C12, two of equations (28) can be relaxed into the following convex constraints, respectively:

C16:

C17:

thus, the optimization problem P3 can be transformed into the optimization problem P4:

P4:

limited by: C8-C10 and C15-C17. Note that the objective function of P4

And the constraint C10 includes a non-concave function R _intra (w _n ). To solve P4 effectively, R needs to be solved _intra (w _n ) An approximation is made. Similar to a first order Taylor expansion for log (a + y) at y =0, R _intra (w _n ) Can be approximated by [5]]：

Wherein,

due to the function g in equation (32) _l (w _n ) And g _j (w _n ) The unevenness of (A) is difficult to determine, is described in [5]]Inspiring, can be right

One-step approximation is performed.

Specifically, according to formula (1), first, g is treated _l (w _n ) In that

Performing a first-order Taylor expansion to obtain:

in the formula (33), | w _n -w _l Is about w _n And in the normal case, the road loss index of the line-of-sight channel α > 1[ 1], [14]]Thus- | | w _n -w _l || ^α Is about w _n Concave function of [14]]S.Boyd and L.Vandenberghe, convex optimization. Cambridge, U.K.: cambridge Univ.Press,2004, i.e., 2004

Is about w _n A concave function of (a). In the same way, pair

The first-order Taylor expansion is processed to obtain:

it is clear that,

is about w _n Affine function of, therefore

Is about w _n Concave function of [14]]。

Due to the fact that

And

are all about w _n So that the non-negatively weighted sum of the two is still with respect to w _n Concave function of [14]]. Thus, in the formula (32)

Can be approximated by the following concave function:

in summary, the formula (21)

Can be approximated by the expression:

further, according to the formula (19)

As a convex function, i.e. in equation (36)

Is a concave function, equation (36) is a non-negatively weighted sum of two concave functions, i.e.

Is a concave function [14]。

Similarly, since C10 also contains a non-concave function R _intra (w _n ) According to equation (32) -equation (35), C10 may be approximated as the following convex constraint:

C18:

therefore, according to equations (36) and (37), the optimization problem P4 can be converted into the following convex problem P5:

P5:

limited by: C8-C9, C15-C18. So far, the optimization problem P2 can be converted into a convex problem P5 after being approximated and relaxed by P3-P5, so that the CVX tool can be utilized to iteratively solve through an SCA algorithm [14].

More specifically, when P5 is solved for the first time, the starting point of P5 needs to be found

Since the original problem of P5 is P3, the starting point needs to satisfy the constraint condition [14] of the optimization problem P3 according to the nature of SCA algorithm]. Further, according to property 1, P2 is equivalent to P3, and thus P2 and P3 are equivalent to each other

The constraint of P2 needs to be satisfied. However, due to the existence of the constraint condition C10 of P2, it cannot be determined whether P2 has a feasible solution, so the starting point

Is difficult to straightenAnd (6) obtaining. To determine whether a feasible solution exists for P2, the maximum transmission rate of R-UAV n needs to be determined

Whether QoS constraints can be met. Therefore, in said step S43, to obtain the maximum transmission rate of R-UAV n

The push power in the P2 objective function can be temporarily ignored and the QoS constraint, i.e., C10, temporarily ignored, so that P2 degrades into a single target problem P6 that maximizes the transmission rate:

P6：

limited by: C8-C9, C11-C12. Since P6 is a degenerate result of P2, P6 can be converted into P5 by a similar method as P2:

first, an objective function of P6 is approximated according to equation (32) to equation (35);

next, the non-convex constraints C11 and C12 are relaxed to C16 and C17, respectively, according to equations (29) and (30).

In summary, P6 can be transformed into the following convex problem:

P7:

limited by: C8-C9 and C16-C17.

Initial position of R-UAV n

As a starting point for the optimization problem P7, i.e.

Subsequently, by iteratively solving P7 through the SCA algorithm, the maximum transmission rate of H-UAV m to R-UAV n can be obtained

And R-Optimized position of UAV

If it is not

It indicates that a feasible solution exists for P2. Since P2 and P3 are equivalent, P3 has a feasible solution, that is, P5 has a starting point, which can be solved by using the SCA algorithm. Next, can order

And calculating τ from equation (26) ⁽⁰⁾ Then is followed by

And as the starting point of P5, iteratively solving P5 by adopting the SCA algorithm again.

In a specific implementation process, the process of solving P7 and P5 by the SPO algorithm in stages in step S44 is shown in table 1:

TABLE 1 SPO Algorithm flow

Wherein,

indicating the convergence accuracy of the algorithm. The algorithm needs to be performed for each R-UAV that is successfully paired. When R-UAV n obtains an optimized position

Later, the absolute coordinates of the C-UAV may be based

Calculating its own absolute coordinates, i.e.

Wait for R-UAV n to move to

After the position, the U2U communication is performed again. So far, the processing flow of the SPO algorithm proposed by the present invention has been introduced. Finally, the design idea of the IaS-U2U communication system designed by the present invention is summarized in fig. 4, which is briefly described as follows:

first, P1 is the MINLP problem and is difficult to solve directly. Therefore, the solution process of P1 is divided into two processes of UAV pairing and R-UAV position optimization. Wherein, the position optimization process needs R-UAV distributed solving P2.

Since the P2 objective function is complex and non-concave, and the constraint condition is non-convex, P2 can be equivalent to P3 according to property 1, and P3 is relaxed and approximated to obtain convex problem P5. When the P5 is solved by iteration through the SCA algorithm, a starting point is needed. Since the original problem of P5 is P2, it is necessary to find the starting point of P5 from the feasible solution of P2. Due to the existence of the QoS constraint C10, it is difficult to determine whether P2 has a feasible solution. Therefore, we neglect the propulsion power in the P2 objective function for the moment and neglect C10 for the moment, construct the maximum transmission rate problem P6, and use the same relaxation and approximation method of P3-P5 to get P7.

And finally, solving P7 and P5 in stages according to the SPO algorithm to complete the position optimization of the R-UAV.

Example 3

More specifically, in addition to examples 1 and 2, the following description will explain a specific embodiment of the present invention by taking software simulation as an example.

In this embodiment, a simulation platform developed based on MATLAB 2014a is used to perform simulation performance evaluation on the novel cluster U2U communication resource scheduling algorithm of the present invention, and a plurality of schemes are constructed and compared. In particular, to minimize the R-UAV overall propulsive power

To this end, based on the prior document [2]]The scheme modifies the channel model and the transmission rate modelThe model adopted by the design is replaced, qoS constraints are newly added, and a Propulsion Power Minimization (PPM) scheme is constructed, so that the method can be applied to the application scenarios described herein. Furthermore, in order to evaluate the gain due to considering UAV propulsive power, due to the lack of existing solutions that can be directly contrasted to maximize the overall H-UAV transmission rate R ^tot To this end, a Transmission Rate Maximization (TRM) scheme is constructed. For the sake of fairness comparison, the UAV pairing process proposed by the design is adopted for all three schemes, namely SPO, PPM and TRM, and the schemes all go through the processing flow of the first stage of the SPO algorithm. It should be noted that, after going through the stage one of the SPO algorithm, the TRM scheme only reserves U2U pairs that satisfy the QoS constraint, that is, only statistically satisfies the transmission rate of the U2U pair of C5 in equation (12); and the PPM scheme optimizes the propulsion power of the R-UAV after undergoing the SPO algorithm stage one. In summary, both the TRM scheme and the PPM scheme can be regarded as simplified versions of the SPO scheme.

The simulation parameters of the present example are shown in table 2, unless otherwise specified below. Wherein, the parameters related to the UAV propulsion power in the formula (2) are all related to [2]Is consistent with (1). UAV maximum flight velocity V _max According to the provisions in 3GPP TR36.777, 45m/s is set. The simulation takes video service as an example, and the minimum transmission quantity in the cluster is obtained

With a setting of 4Mbit/s, 720p high-resolution video applications can be supported. Reference [15]Zhu, Y.Liang and M.Yan, "A flexible compliance assembly trajectory for the format of multiple unamended orthogonal vehicles," IEEE Access, vol.7, pp.140743-140754,2019, doi

2m, cluster UAV control with an accuracy of 2m can be achieved at a cluster speed of 45m/s.

Table 2: simulation parameter list

In the specific implementation process, fig. 5 simulates UAV position comparison before and after IaS-U2U communication optimization. As shown, under the UAV position constraints, the R-UAV will not collide with the H-UAV and the C-UAV, nor will it depart from the cluster scope. Notably, partial pairing after optimization through stage 1 of algorithm 1, the maximum achievable rate

QoS requirements are not met resulting in failure of the final location optimization. Thus, not all H-UAVs are able to provide U2U transmission services in the same sub-optimal process. UAVs that fail matching or fail optimization may proceed to the next round of optimization.

Different UAV protection distances are simulated by adopting a Monte Carlo method

Next, R-UAV Overall propulsive Power for the SPO scheme

H-UAV Overall Transmission Rate R ^tot And energy efficiency mu _intra V with cluster speed | | v ^sw The variation relationship of | is shown in fig. 6 (a), 6 (b), and 6 (c), respectively. It should be noted that the propulsion power preference factors for each R-UAV are set to the same value and are all 0.5, i.e., θ ₁ ＝θ ₂ ＝...＝θ _N ＝0.5。

In the specific implementation, as can be seen from fig. 6 (a), when the cluster speed changes, it is different

The lower propelling power change trend is that the propelling power is decreased gradually and then increased gradually, and when the cluster speed | | v ^sw | | ≧ 10m/s, their propulsive power followsThe rate of increase of the cluster speed gradually increases. This is because, on the one hand, the present solution optimizes the relative position of the R-UAV, and as can be seen from equation (4), the position optimization only affects the relative speed of the R-UAV

According to the formulas (4) and (5), the propulsion power of the R-UAV is also equal to the cluster speed | | v | ^sw And | | is related. Therefore, the temperature of the molten metal is controlled,

the variation trend of (C) is also influenced by the cluster speed | | | v ^sw The effect of | l. On the other hand, according to equation (2), the rotorcraft UAV presents a flight speed V corresponding to the lowest propulsive power _MP The trend of the propulsion power is therefore first decreasing and then increasing, this trend being associated with [2]]The conclusions are consistent.

In addition, when the cluster speed | | | v ^sw When the value of | is small,

with following

Is increased and decreased. This is because, as shown in FIG. 6 (a), when

And | | | v ^sw When the value of | is small,

smaller, R ^tot Is relatively large. In combination with equation (7) -equation (9), the proportion of propulsion power in the multi-objective problem is small, and the optimization algorithm is based on the transmission rate objective. However, with

The combination of the formula (1) and the formula (6) shows that the transmission rate of the R-UAV is gradually reduced, so that the occupation ratio of the propulsion power in the multi-objective problem is gradually increased, at the moment, the optimization algorithm gradually emphasizes the propulsion power target, and the combination of the formula (7) and the formula (9) shows that,

will be reduced. Overall analysis, H-UAV Overall Transmission Rate R ^tot And R-UAV Total propulsive Power

There is a trade-off relationship.

However, when the cluster speed | | | v ^sw As l continues to increase in size,

enter the growth interval and the growth trend follows

Is increased and accelerated. This is because, on the one hand, when

When the growth interval is entered, the data is transmitted,

with | | | v ^sw The increase in | increases. According to R ^tot And

the trade-off relationship of (1) shows that the optimization algorithm tends to decrease

To maximize the objective function in equation (7). On the other hand, however, according to C6 and C7 in the formula (12), as follows

With the increase in the number of possible solutions for the R-UAV, the set of feasible solutions for the R-UAV will decrease, and the distance between the H-UAV and which the R-UAV is optimized to mate with will increase, thus the H-UAV transmission rate will decrease, resulting in the number of R-UAVs meeting the QoS condition, i.e., meeting C5 in equation (12), after the SPO algorithm stage one, will decrease, and the failed R-UAV will remain in place. This accounts for R-UAV movement limitsThe enhancement of (c) may result in a reduction in the number of R-UAVs that may ultimately perform position movement. In combination with equation (8), the optimization algorithm is implemented by changing the position of the R-UAV

The effect of the decrease will also gradually diminish. Combining the above two factors, when

When the size of the pipe is increased, the pipe is enlarged,

with | | | v ^sw The growth of | will increase rapidly.

From the H-UAV overall transmission rate performance curve given in FIG. 6 (b), it can be observed that the H-UAV overall transmission rate R of the present scheme ^tot With cluster speed | | | v ^sw The increase in | decreases. According to C3 in the formula (12), | | v ^sw An increase in | may increase the movement limit of the R-UAV; referring again to the conclusion of FIG. 6 (a), it can be seen that the number of R-UAVs that eventually undergo position movement is reduced. Thus, according to formula (9), R ^tot V with cluster speed | | v ^sw The increase in | decreases. Furthermore, R ^tot With following

Is increased and decreased. This is because when

When the distance between the R-UAV and the H-UAV matched with the R-UAV is increased after optimization, according to the formula (1) and the formula (6), the transmission rate of the H-UAV is reduced, so that the R-UAV is caused to have a reduced transmission rate ^tot And decreases. Finally, FIG. 6 (b) shows the results when

When increasing, with | | v ^sw Variation of | R ^tot The variation range of (a) is reduced and gradually becomes gentle. According to C3 in the formula (12), when | | | v ^sw When | l changes within a certain range, the optimization algorithm can be based on | v | ^sw I adjusting R-UAV optimizationRear position w _n I.e. w _n Will vary within a feasible solution. And in combination with formula (9), R ^tot Is about w _n A function of (1), thus R ^tot Will also vary within certain limits. However, with

From C6 and C7 in equation (12), the mobile location limit of the R-UAV increases, resulting in a decreasing set of feasible solutions. This means that w _n The adjustable range is also reduced, as shown in formula (9), R ^tot Will also gradually narrow. At this time, on one hand, the limitation on the mobile position of the R-UAV is increased, which increases the difficulty of meeting the QoS constraint after the R-UAV is optimized, so that the number of U2U pairs for which the optimization succeeds is reduced; on the other hand, the interference term in equation (6)

And the QoS requirement can be more easily met after the R-UAV is optimized. In summary, the number of successfully optimized U2U pairs tends to be stable after being reduced to a certain degree, and R is stable ^tot The attenuation does not continue but gradually approaches a certain minimum value, as shown in fig. 6 (b).

FIG. 6 (c) shows the difference

Energy efficiency of the system, μ _intra Performance comparison of (2). It can be seen that the energy efficiency all shows a trend of increasing first and then decreasing with the cluster speed. As can be seen from FIG. 6 (a), the energy efficiencies are all

The lowest reaches the peak point. In addition, referring back to FIGS. 6 (a) and 6 (b), when

At increasing overall transmission rate R ^tot Will be greatly reduced and the total propulsion power will be reduced

Without significant drop, and thus energy efficiency mu _intra To R ^tot Influence following

Increases and decreases as shown in fig. 6 (c). However, as future UAV technology develops, the control accuracy is expected to be improved continuously, and the protection distance of the UAV may be kept at a smaller value, which may provide better implementation conditions and basis for implementation of the present solution.

More specifically, the energy efficiencies of the SPO, TRM and PPM schemes were compared. Combining the definition of the TRM scheme and the PPM scheme, the SPO scheme is the weighted sum of the target functions of the TRM scheme and the PPM scheme, and the weight distribution is subject to the preference factor theta according to the formula (7) _n The influence of (c). Thus, θ _n Will be applied to the objective function U of the IaS-U2U system _intra And R ^tot And

all bring influence, and as can be seen from equation (10), this will further influence the energy efficiency μ of the system _intra As shown in fig. 7. This phenomenon illustrates that at different cluster speeds | | | v ^sw Under | |, the preference factor theta of the R-UAV can be adjusted by adjusting _n To obtain energy efficiency mu _intra Gain in performance.

In the specific implementation process, as shown in fig. 7, when the cluster speed is in the middle-low speed interval, i.e., | | v ^sw When | | | is less than or equal to 32.5m/s, the energy efficiency performance of the SPO scheme is superior to that of the TRM scheme and the PPM scheme, and the preference factor theta can be increased _n To improve the gain in energy efficiency. When the cluster speed is in a high-speed interval, namely | | | v ^sw When | | > 32.5m/s, although the SPO scheme is at θ _n The performance is lower than that of the TRM scheme when the preference factor theta is larger than or equal to 0.5, but the preference factor theta can be reduced _n To gain in energy efficiency performance.

In table 3, the average number of iterations of the SPO algorithm at different convergence accuracies is simulated. The simulated hardware environment is a Windows 7.0 operating system loaded with Intel i7-6700 CPU. Here, the preference factor for each R-UAV is set to 0.5. Because the complexity of the position optimization algorithm is higher than that of other functional modules, the complexity of the whole algorithm is mainly determined by the complexity of the position optimization algorithm. The position optimization algorithm adopts a CVX tool interior point method to solve, and the complexity of the CVX tool interior point method can be approximate to O (a) ^3.5 )log ₂ (1/κ)[8]. Where a and κ represent the number of variables and the interior point method convergence accuracy, respectively. The complexity and the convergence precision k of the interior point method can be understood as the precision used by the SPO algorithm when the CVX tool is called each time to solve the convex problem, and this parameter is determined by the CVX tool. Thus, the complexity of stage 1 and stage 2 of the SPO algorithm are O (n), respectively _var1 ) ^3.5 K ₁ log ₂ (1/κ) and O (n) _var2 ) ^3.5 K ₂ log ₂ (1/κ) wherein n is _var1 And n _var2 The numbers of variables in stage 1 and stage 2 are shown, respectively. For phase 1, the optimization variable is w _n ＝[x _n ,y _n ] ^T Thus n is _var1 =2; for stage 2, the optimization variable is w _n ＝[x _n ,y _n ] ^T And τ, thus n _var2 =3. Furthermore, K ₁ And K ₂ The number of iterations for phase 1 and phase 2 are shown separately. K ₁ 、K ₂ The values of (c) can be referred to in table 3. As can be seen from Table 3, this value is consistent with the convergence accuracy of the SPO algorithm

In connection with, among others, convergence accuracy

Parameters to determine when phases 1,2 converge. With following

The number of iterations of both stage 1 and stage 2 of the SPO algorithm increases.

Table 3: average iteration number required by SPO algorithm under different convergence precision targets

It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims (10)

1. A U2U distributed dynamic resource allocation method facing to intra-cluster communication is characterized by comprising the following steps:

s1: constructing an IaS-U2U communication scene model and a communication flow, wherein three kinds of UAVs are considered, including a central UAV (user equipment), namely a C-UAV, in a cluster central position, an assistant UAV (user equipment), namely an H-UAV, for actively providing U2U communication service and a requester UAV, namely an R-UAV, requiring U2U service;

s2: constructing an objective function of IaS-U2U communication to obtain a multi-objective optimization problem of the IaS-U2U communication;

s3: decomposing the multi-objective optimization problem, and carrying out a UAV matching process by using the H-UAV and the R-UAV by adopting a classical GS algorithm to complete UAV pairing;

s4: according to the UAV pairing result, completing the position optimization of the R-UAV by adopting an SPO algorithm, and completing the solution of a multi-objective optimization problem;

s5: and according to the solving result, completing the distribution of the U2U distributed dynamic resources.

2. The method as claimed in claim 1, wherein in the IaS-U2U communication scenario model of step S1, the C-UAV is always in a state with radius r _sw The circular cluster area center is used for coordinating and controlling the movement of the rest UAVs in the cluster; the remaining UAVs are initially randomly distributed within the cluster, eachThe distance between the UAVs is not less than the preset protection distance

Considering the general situation in an aviation scene, it is assumed that the UAV cluster works in an OOC state, but when necessary, the C-UAV may access a satellite to acquire route planning information; in this scenario, the R-UAV needs to obtain a data packet from the H-UAV buffer space, and in order to obtain better transmission conditions, it needs to complete optimization of the relative position and movement trajectory of the R-UAV in the cluster;

the distance between clusters is assumed to be long, so that the interference between the clusters is small and can be ignored; the UAVs in the same cluster reuse the same frequency band, and mutual interference exists; furthermore, due to the close distance between the UAVs within the cluster, it is assumed that the R-UAVs can each be in one slot δ _intra Movement of the inner completion location; assuming that there are M H-UAVs and N R-UAVs in the UAV cluster, their ordinal numbers constitute the sets M = {1,2, ·, M.,..., M } and Ν = {1,2,.., N.,. N }, respectively; furthermore, let w _i ＝[x _i ,y _i ] ^T Denotes the coordinate vector of UAVi relative to C-UAV, and let W _H-UAV ＝[w _m ] _m∈M And W _R-UAV ＝[w _n ] _n∈N Respectively representing matrixes formed by coordinate vectors of the H-UAV and the R-UAV; assume that all H-UAVs and R-UAVs within the cluster can acquire δ from the intended flight path or C-UAV's temporary notification message _intra Absolute coordinates of C-UAV after time slot is over

Then any UAV i can be based on its relative coordinates w within the cluster _i To calculate the corresponding absolute coordinates

Namely, the method comprises the following steps:

assuming all UAVs inside the cluster are at the same flight altitude, then for

And i ≠ j, d _i,j ＝||w _i -w _j | l represents the distance from the UAV i to the UAVj, wherein | · | | is a two-norm operator;

in an IaS-U2U communication scene, the UAVs are assumed to be LOS channels, the Doppler frequency shift can be accurately estimated, and the influence caused by Doppler can be completely offset; therefore, the channel model of the model adopts a free space path loss model, and then H-UAVm and R-UAVn are in the time slot delta _intra The channel gain in is expressed as:

wherein, beta and alpha respectively represent a channel coefficient and a path loss index; furthermore, the rotorcraft UAV is more flexible than a fixed-wing UAV, and is therefore optimized with a model of the propulsion power P of the rotorcraft UAV _p Is a function of the velocity magnitude V of the UAV, written as:

wherein: p _bla Representing blade power; u shape _tip Representing tip speed; d _rat Representing fuselage drag ratio; ρ represents an air density; s _rot Represents the rotor volume; a represents the rotor disk area; p _ind Represents the induced power; v. of _rot Represents the average rotor induction speed; wherein, the propulsive power of rotor UAV comprises three parts: blade power, induced power, and parasitic power; the blade power and the induced power need to overcome the blade resistance and the fuselage resistance and are respectively in direct proportion to the quadratic power and the cubic power of the speed V; induced power is the power required to overcome the blade induced drag, which decreases as the speed V increases; therefore, the propulsion power of the rotorcraft UAV is reduced along with the increase of the speed V and then increased along with the increase of the speed V, and the rotorcraft UAV has the characteristics ofA speed value, denoted V, which minimizes propulsion power _MP (ii) a Since the propulsion power of a single UAV is typically in the order of hundreds of watts, in the IaS-U2U communication scenario model, the transmit power is assumed to be much smaller than the propulsion power.

3. The method for allocating U2U distributed dynamic resources for intra-cluster communication according to claim 2, wherein in step S1, the communication flow of the IaS-U2U communication scenario model specifically includes:

s11: each R-UAV broadcasts a U2U request and content characteristic information of the R-UAV to each H-UAV; after each H-UAV receives the request, broadcasting the content characteristic information of the H-UAV; then, each R-UAV performs distributed pairing with each H-UAV based on the default H-UAV transmitting power and the content characteristics issued by the H-UAV;

s12: each R-UAV successfully paired initiates a position optimization application to the C-UAV, and the C-UAV starts a position optimization process to the R-UAV in a polling mode, wherein the position optimization process specifically comprises the following steps:

i. when polling a particular R-UAV, the C-UAV sends a list of location information to the R-UAV, the list containing location information for all UAVs in the cluster;

the R-UAV executes an SPO algorithm according to the position information list for position optimization;

if the optimization is successful, the R-UAV feeds back an optimization result to the H-UAV and the C-UAV which are matched with the R-UAV, and then a position moving process is executed; if the optimization fails, feeding back an optimization failure identifier to the H-UAV, and waiting for the next round of matching by the R-UAV and the H-UAV;

calculating the absolute coordinates of the R-UAV according to the optimization result, and implementing position movement;

v. the R-UAV telling it that the H-UAV position movement process paired with it is finished;

s13: each R-UAV establishes a U2U connection with its respective paired H-UAV and performs U2U communications.

4. The method according to claim 3, wherein in step S2, according to the communication process in step S1, joint optimization is performed in consideration of two aspects of transmission rate and propulsion power of the cluster system, and an objective function of IaS-U2U communication is constructed, specifically:

pairing the R-UAV and the H-UAV before the R-UAV changes its own position and establishes a connection with the H-UAV; defining a binary matrix X _M×N Element x thereof _m,n =1 denotes H-UAVm paired with R-UAVn, x _m,n =0 then denotes unpaired, where m ∈ m, n ∈ Ν; assuming that each H-UAV serves at most 1R-UAV, each R-UAV is served by at most one H-UAV, i.e.:

in addition, let

Denotes the initial coordinates, w, of R-UAV n relative to C-UAV _n Representing coordinates of the R-UAV n next time slot relative to the C-UAV; since only the position of the R-UAV is optimized, i.e., assuming that the velocities of the C-UAV and H-UAV are the same, and defining the velocity as the velocity of the cluster, expressed as a vector

Wherein

Representing the velocity components of the cluster in the x-direction and the y-direction, respectively; in summary, the velocity magnitude of R-UAV n is represented by the two-norm of its velocity vector as:

wherein,

representing the relative velocity of R-UAV n with respect to the cluster; based on the formula (2)Using w as the propulsion power of R-UAV n _n Expressed as:

suppose each H-UAV transmits at full power, and the transmit power of H-UAV m is noted as

The transmission rate of H-UAV m to R-UAV n is then expressed as:

wherein B and ε ² Respectively representing the channel bandwidth and the noise power spectral density; based on the analysis, the objective function of the IaS-U2U communication is constructed by combining the formula (5) and the formula (6), namely:

wherein: theta _n ∈(0,1]Representing a preference factor of the R-UAVn for propulsion power, used for adjusting the proportion of the propulsion power in the multi-objective problem, theta _n The higher the ratio is, the more important the R-UAVn is to the propulsion power problem; positive real number

Has the unit of Mbit s ^-1 ·W ^-1 For applying propulsive power P _p (w _n ) Is adjusted to be equal to the transmission rate R _intra (w _n ) The same order of magnitude and keeping the adjusted terms and transmission rate the same dimension.

5. The intra-cluster communication-oriented U2U distributed dynamic resource allocation method according to claim 4, wherein in the step S2, for convenience of expression, the following variables are defined based on equations (5) and (6):

·

represents the sum of all R-UAV propulsion powers in the cluster:

·R ^tot represents the sum of all H-UAV transmission rates in the cluster:

·μ _intra representing the energy efficiency of the IaS-U2U communication system:

to sum up, the multi-objective optimization problem of IaS-U2U communication is expressed as:

limited by:

wherein:

constraints C1 and C2 are pairing constraints for H-UAV and R-UAV;

constraint C3 indicates that the flight speed of the R-UAV should not be higher than the UAV maximum flight speed V _max ；

Constraint C4 is the anti-separation cluster constraint and represents the optimized relative coordinates of any R-UAVStill at a radius r _sw Within the cluster formation range of (1);

constraint C5 is a QoS constraint indicating that the transmission rate between the H-UAV and the R-UAV for any successful pairing is not lower than a given

Constraints C6 and C7 are collision avoidance constraints, indicating that the distance between any R-UAV and any H-UAV and C-UAV (whose location is the origin of coordinates) should be greater than

6. The method for allocating U2U distributed dynamic resources to intra-cluster communication according to claim 5, wherein in the step S3, it is noticed that the concave-convex property of the objective function of P1 is difficult to determine, the optimized variable X is a discrete variable, and the constraint conditions C1-C2 and C5-C7 are non-convex constraints, so that P1 is a mixed integer nonlinear problem, and the calculation difficulty of directly solving the problem is large, and therefore, the solving process of P1 needs to be decomposed into the following two steps: UAV matching and position optimization; wherein, the UAV matching process specifically comprises:

before a pairing process is started, acquiring content classification information of an opposite party by an H-UAV and an R-UAV in a broadcasting mode; then, each UAV calculates content similarity according to the Jaccard coefficient and generates a respective preference list; in order to enable the paired position optimization result to satisfy the constraint condition C5, candidate pairs need to be screened, specifically: R-UAV n measures the received Signal to noise ratio, SINR, for each H-UAV, noted

If it is

Then R-UAV n deletes H-UAV m from its list of preferences, where

The SINR threshold of R-UAV n is expressed as follows:

after each R-UAV modifies the preference list according to the SINR threshold, the classical GS algorithm is used for realizing one-to-one matching between the R-UAV and the H-UAV, and a pairing matrix X is obtained ^* (ii) a And meanwhile, the serial numbers of the H-UAV and the R-UAV which are successfully paired are respectively formed into a new set of M 'and N', and the UAV pairing is completed.

7. The method according to claim 6, wherein the step S4 specifically includes the following steps:

s41: converting P1 into an optimization problem P2 according to the pairing result;

s42: equating the optimization problem P2 to be an optimization problem P3 according to the property 1, and relaxing and approximating the P3 to obtain a convex problem P5;

s43: constructing a maximum transmission rate problem P6 according to the optimization problem P2, and obtaining an optimization problem P7 by adopting the same relaxation and approximation methods of P3-P5;

s44: and (4) solving P7 and P5 in stages according to the SPO algorithm to complete the position optimization of the R-UAV and solve the multi-objective optimization problem.

8. The method according to claim 7, wherein the step S41 specifically includes:

forming the transmitting power of each H-UAV in the M' into a vector

Each R-UAV successfully paired will be position optimized based on the power value; based on equation (6), for the successfully paired R-UAV n, the transmission rate it obtainsComprises the following steps:

where m represents the H-UAV Serial number paired with R-UAV n, i.e., x _m,n =1; further, when UAV pairing is complete, the objective function in equation (7) will degenerate as:

wherein,

representing an objective function of R-UAV n; therefore, for each R-UAVn, the following optimization problem needs to be solved:

limited by:

wherein C8-C12 are constrained C3-C7 of P1 and w _n The associated constraints.

9. The method according to claim 8, wherein the step S42 specifically includes:

it is noted in step S41 that the objective function of the optimization problem P2 is non-concave, and the constraints C10-C12 are non-convex constraints, which makes it difficult to solve the original problem; therefore, the objective function and the constraint condition of the P2 need to be relaxed, and then the SCA algorithm is adopted for solving; the SCA algorithm solves the original problem in an iterative mode, and the original problem needs to be obtained by the last iteration in each iterative solutionThe optimal solution of (2); definition K =0,1,2 _max -1 represents the number of iterations of the SCA algorithm, where K _max Representing the maximum number of iterations allowed by the algorithm; particularly, when the first iteration solution is carried out, a starting point needs to be provided for an SCA algorithm; the starting point is any feasible solution which meets the constraint condition of the original problem, and the index k =0;

to this end, a relaxation variable is first introduced:

using equation (18), equation (5) is approximated as follows:

it is noted that | | v _n || ² And v _n || ³ Are all convex functions, tau is an affine function, thus

Is a convex function;

will be provided with

After substituting P2 and introducing the relaxation variable τ, P2 is transformed into the following optimization problem:

limited by: C8-C13, wherein:

based on the above analysis, the following properties were present:

properties 1: the optimization problem P3 is equivalent to P2:

when in use

In order to maximize the objective function of P3, the propulsion power in equation (21) is reduced by reducing the value of τ

Thereby improving the value of the target function P3; when τ continues to decrease to C13, i.e., when the equal sign of equation (18) holds, equation (19) degenerates to equation (5), i.e., there is

At this time, the optimization problem P3 is equivalent to P2;

since C13 in equation (18) is still a non-convex constraint, further operation is required; squaring both sides of the formula (18) and taking the reciprocal to obtain:

the two equations in equation (22) add, equating C13 to the constraint:

since C14 is still a non-convex constraint, the left end of C14 is relaxed; for C14, i.e. τ in formula (23) ² And

at τ = τ respectively ^(k) And

is subjected to a first order Taylor expansion, wherein (·) ^(k) Representing the value obtained for the k iteration of a certain variable, there are:

where · is the vector inner product operator; the two equations in equation (24) are added to obtain:

wherein, tau ^(k) And

calculated from the following formula:

after the above processing, the non-convex constraint C14 is relaxed to a convex constraint C15:

further, by using | | a | | | b | | | | > or more than a · b, the left sides of the inequalities of C11 and C12 are relaxed as follows:

and then connecting the two formulas (28), (C11) and (C12) to relax the two formulas in the formula (28) into the following convex constraints respectively:

thus, the optimization problem P3 is transformed into the optimization problem P4:

limited by: C8-C10, C15-C17; note that the objective function of P4

And the constraint C10 includes a non-concave function R _intra (w _n ) (ii) a To solve P4 effectively, R needs to be solved _intra (w _n ) Making an approximation; similar to a first order Taylor expansion for log (a + y) at y =0, R _intra (w _n ) The approximation is:

wherein,

due to the function g in equation (32) _l (w _n ) And g _j (w _n ) Is difficult to determine, therefore, it is suitable for

Performing one-step approximation;

specifically, according to formula (1), first, g is treated _l (w _n ) In that

Performing first-order Taylor expansion to obtain:

in the formula (33), | w _n -w _l I is about w _n And in general the road loss index alpha of the line-of-sight channel is > 1, thus- | w _n -w _l || ^α Is about w _n Concave function of, i.e.

Is about w _n A concave function of (a); in the same way, for g _j (w _n ) In that

Performing a first-order Taylor expansion:

it is clear that,

is about w _n Affine function of, therefore

Is about w _n A concave function of (d); due to the fact that

And

are all about w _n So that the non-negatively weighted sum of the two is still with respect to w _n A concave function of (a); thus, in the formula (32)

Approximated by the concave function:

in summary, the formula (21)

Approximated by the expression:

further, according to the formula (19)

As a convex function, i.e. in equation (36)

Is a concave function, equation (36) is a non-negatively weighted sum of two concave functions, i.e.

Is a concave function;

similarly, since C10 also contains a non-concave function R _intra (w _n ) According to equation (32) -equation (35), C10 is approximated as the following convex constraint:

therefore, according to equations (36) and (37), the optimization problem P4 can be converted into the following convex problem P5:

limited by: C8-C9, C15-C18; so far, the optimization problem P2 is converted into a convex problem P5 after the approximation and relaxation of P3-P5, and then the iterative solution is carried out through an SCA algorithm by utilizing a CVX tool.

10. According to claim 9The U2U distributed dynamic resource allocation method facing to intra-cluster communication is characterized in that when P5 is solved, a starting point of P5 needs to be found

Because the original problem of P5 is P3, according to the property of the SCA algorithm, the starting point needs to meet the constraint condition of the optimization problem P3; further, according to property 1, P2 is equivalent to P3, and thus P2 and P3 are equivalent to each other

The constraint condition of P2 needs to be satisfied; however, due to the existence of the constraint condition C10 of P2, it cannot be determined whether P2 has a feasible solution, so the starting point

Difficult to obtain directly; to determine whether P2 has a feasible solution, the maximum transmission rate of R-UAV n needs to be determined

Whether QoS constraints can be met; therefore, in said step S43, to obtain the maximum transmission rate of R-UAV n

Ignoring the push power in the P2 objective function for the moment and ignoring the QoS constraint, i.e., C10, for the moment, so that P2 degrades into a single objective problem P6 that maximizes the transmission rate:

limited by: C8-C9, C11-C12; since P6 is a degenerate result of P2, P6 can be converted into P5 by a similar method as P2:

first, the objective function of P6 is approximated according to equation (32) -equation (35);

secondly, relaxing the non-convex constraints C11 and C12 to C16 and C17, respectively, according to equations (29) and (30);

in summary, P6 is transformed into the following convex problem:

limited by: C8-C9, C16-C17;

initial position of R-UAV n

As a starting point for the optimization problem P7, i.e.

And then, iteratively solving P7 through an SCA algorithm to obtain the maximum transmission rate from H-UAV m to R-UAV n

And optimized position of R-UAV

If it is not

Indicating that a feasible solution exists for P2; because P2 and P3 are equivalent, P3 has a feasible solution, namely P5 has a starting point, and an SCA algorithm is adopted for solving; next, let

And calculating τ from equation (26) ⁽⁰⁾ Then is followed by

And as the starting point of P5, iteratively solving P5 by adopting the SCA algorithm again.

Images