阅读说明：本技术 无人机蜂群多微蜂窝频谱资源管理方法 (Unmanned aerial vehicle swarm multi-microcellular frequency spectrum resource management method ) 是由 张周 谢佳 赵维维 王彤彤 陈小庆 赵润森 于 2021-07-30 设计创作，主要内容包括：本发明公开了一种无人机蜂群多微蜂窝频谱资源管理方法,该方法为：构建无人机蜂群多微蜂窝网络模型,并针对无人机蜂窝网中的下行资源调度问题进行建模；采用分布迭代算法,首先在固定功率分配的条件下将各个RU分配给子节点,然后在信道调度固定的情况下更新功率分配,有限次的反复迭代后,即可达到上述资源管理优化问题的相对最优解。本发明方法可实现无人机微蜂窝集群间的频谱共享和干扰协调,使得同一频率载波能够在相邻的不同无人机微蜂窝集群间多次复用,解决对抗环境下频谱资源短缺问题,进一步提升无人蜂群的效能,应对复杂多变的组网环境。(The invention discloses a multi-microcellular frequency spectrum resource management method for an unmanned aerial vehicle swarm, which comprises the following steps: constructing an unmanned aerial vehicle swarm multi-microcellular network model, and modeling aiming at a downlink resource scheduling problem in an unmanned aerial vehicle cellular network; and (3) adopting a distributed iteration algorithm, firstly allocating each RU to a child node under the condition of fixed power allocation, then updating the power allocation under the condition of fixed channel scheduling, and after repeated iteration for a limited time, achieving the relatively optimal solution of the resource management optimization problem. The method can realize the spectrum sharing and interference coordination among the micro-cellular clusters of the unmanned aerial vehicle, so that the carrier wave with the same frequency can be reused among the adjacent micro-cellular clusters of different unmanned aerial vehicles for multiple times, the problem of spectrum resource shortage under the confrontation environment is solved, the efficiency of the unmanned bee colony is further improved, and the unmanned aerial vehicle can cope with the complex and changeable networking environment.)

1. An unmanned aerial vehicle swarm multi-microcellular frequency spectrum resource management method is characterized in that an unmanned aerial vehicle swarm multi-microcellular network model is built, a frequency spectrum sharing problem is described as two subproblems of channel scheduling and power distribution, the optimal solution of one variable is solved while the other variable is fixed, the channel scheduling problem is an unconstrained optimization problem, the power distribution problem is a constrained convex optimization problem, and the optimal solution is resolved in each power distribution solution, and the method comprises the following steps:

step 1: in the initial network establishment stage, for the double-layer cellular network of the unmanned aerial vehicle, when the initial network establishment is finished, namely a cluster sub-node and a cluster head exist in a cluster, the cluster head selectively starts a resource scheduling module;

step 2: the cluster subnode sends a resource request packet to a cluster head, and the cluster head performs initial parameter configuration on a resource management module, wherein the initial parameter configuration comprises a working frequency band, the number of distributable RUs and the maximum transmitter power;

and step 3: first-stage information collection: the cluster head judges nodes and RUs which need information collection according to the uplink and downlink demand information of each node, and sends corresponding training sequences to estimate channel CSI;

and 4, step 4: preliminary RU allocation: the cluster head distributes RUs according to a channel scheduling rule based on the collected channel CSI transmitted by each node on the RUs;

and 5: the cluster head judges whether the cluster head receives channel scheduling information of other cluster heads, if so, the initial RU distribution result is combined with the channel distribution results of other cluster heads, and the node which is most likely to be interfered by the cluster head is judged and is called a main interference node; then, the cluster head sends a training sequence to estimate the interference path CSI of the main interference node;

step 6: based on the preliminary RU distribution result, the cluster head calculates the corresponding optimal power on each RU according to the obtained channel information; when the calculated power result is less than zero, taking the power of 0, and correcting the preliminary RU distribution result;

and 7: when the resource scheduling is finished, the resource management module of the cluster head outputs a scheduling result to the MAC layer, and broadcasts the scheduling result of the current round to other cluster heads through the macro cell;

and 8: the next scheduling cycle begins and returns to step 3.

2. The method for spectrum resource management of the drone swarm multi-microcellular, according to claim 1, wherein the model for the drone swarm multi-microcellular network is constructed as follows:

the unmanned aerial vehicle swarm network is divided into two layers, the upper layer is a macro cell, a centerless ad-hoc network form is adopted, the downlink is a micro cell, a centralized network form taking a cluster head as the center is adopted, the lower layer micro cell is a cluster, the center node is the cluster head and is responsible for inter-cluster interaction and intra-cluster resource scheduling, the bottom layer communication node is a cluster sub-node, and the cluster sub-node performs channel access and data transmission according to the resource scheduling information of the cluster head to which the cluster sub-node belongs; the channel link in the cluster is divided into an uplink and a downlink, the uplink is from a cluster sub node to a cluster head, the downlink is from the cluster head to the cluster sub node, the uplink and the downlink use different frequency bands, and the resource scheduling is mutually independent;

based on the network model, the downlink resource scheduling problem in the unmanned aerial vehicle cellular network is modeled into the following optimized model

Wherein m represents the index number of the microcell, i.e. the cluster head; k represents the index number of the cluster child node in the microcell, and if the child node K belongs to the microcell m, K belongs to K_m，K_mRepresenting a set of cluster child nodes within microcell m; n denotes the RU index number, assuming the network-wide RU sequence number is the same asStep, i.e. one n is corresponding to the RU of one frequency, and it is assumed that all the micro cells share the same large frequency band;represents the long-term average of the channel rate of node k over all RUs in microcell m, and is formulated as Indicating the allocation of RUs ifIn the micro cell m, the RU n is allocated to the node k for data transmission, and in the same cell, one RU can only serve one node, but one node can occupy multiple RUs at the same time;means that the sum of the powers of the signals transmitted in microcell m for data transmission on RU with node k should not exceed the total power of the cluster head transmitter

Receiving signal-to-interference-and-noise ratio (SINR) of downlink channel from cluster head m to cluster child node k on RU nIs calculated as follows

And (3) adopting a distributed iteration algorithm, firstly allocating each RU to a child node under the condition of fixed power allocation, then updating the power allocation under the condition of fixed channel scheduling, and after repeated iteration for a limited time, achieving the relatively optimal solution of the resource management optimization problem.

3. The method according to claim 2, wherein the spectrum sharing problem is described as two sub-problems of channel scheduling and power allocation, the optimal solution of one variable is solved while the other variable is fixed, the channel scheduling problem is an unconstrained optimization problem, the power allocation problem is a constrained convex optimization problem, and each power allocation solution is solved by solving a resolved optimal solution, specifically as follows:

the first stage is as follows: channel scheduling

Assuming that all RUs are unoccupied during initial network entry, all RUs are occupiedN is the total number of RUs, and for each RU N, the child nodes are allocated according to the following rule

If it isThen

WhereinThe estimated value of the channel instantaneous speed when the subnode k occupies RU n in the honeycomb m;

and a second stage: power distribution

In the case of fixed RU channel scheduling, the optimization problem in equation (1) is reduced to a convex optimization model as follows

Converting the constrained optimization problem in the formula (2) into an unconstrained optimization problem to solve, wherein

Using the KKT conditional solution to the unconstrained optimization problem in equation (3) to obtain a solution for power allocation

Wherein the content of the first and second substances,indicating varying transmission power of cluster sub-nodes within a cellObtaining an analytical expression by derivation on the influence of channel receiving performance of cluster sub-nodes in adjacent cells subjected to same frequency interference on the same RU;

more than one cell simultaneously interfered by same frequency cells can be used for all the interfered cluster subnodesSum, parameterThe gain of the downlink interference channel is represented, and the summation process is simplified into a multiple relation, specifically:

wherein c is more than or equal to 1.

4. The method according to claim 3, wherein in step 6, based on the preliminary RU allocation result, the cluster head calculates the corresponding optimal power on each RU according to the obtained channel information; when the calculated power result is less than zero, taking the power of 0, and correcting the preliminary RU distribution result at the same time, specifically:

step 6.1: for each cluster head m, and the cluster child nodes K in the cluster head m are belonged to K_mInitializing λ_mWhen the iteration step size is 0, the iteration step size is set to 0.01;

step 6.2: selecting a main interference node and a corresponding cluster head based on a nearest principle according to local channel CSI and adjacent cell CSI obtained through cross-cluster estimation, and calculating an influence value t of a power change alignment link in the cell;

step 6.3: according to the t parameter, calculating the channel occupation parameter for all the nodes k and RU numbers nTransmission Signal Power for RUs other than 0

Step 6.4: judging the corresponding transmitting signal power on each node k and Ru nWhether the value is greater than 0 or not, and if the value is less than zero, taking a zero value;

step 6.5: ascending: for each cluster child node k within cluster head m, the occupied RU is traversed, i.e.The corresponding power values are summed, and whether the corresponding power values meet the maximum power constraint of the transmitter of the cluster sub-node is judged;

descending: summing all power values of all cluster sub-nodes and RUs in the cluster head m, and judging whether the maximum power constraint of a transmitter of the cluster head node is met;

if the constraint is not satisfied, then λ is_mIncrease of λ_m＝λ_m+0.01, return to step 6.3; otherwise, entering step 6.6;

step 6.6: and finishing the power distribution calculation, and correcting the primary RU distribution, specifically comprising the following steps: if power value of RUnAnd channel occupation flag bitThe channel occupancy flag is set to zero,

Technical Field

The invention belongs to the technical field of wireless communication, and particularly relates to a multi-microcellular frequency spectrum resource management method for an unmanned aerial vehicle swarm.

Background

The existing cellular technology is mainly applied to a 5G network, a large number of small cellular base stations are densely deployed to improve the overall performance of the system, and the network is in a super-dense state. UDNs, as a typical feature of 5G cellular networks, can improve spectrum and energy efficiency, increase network capacity, and become one of the main features of future networks. Based on densely deployed small cell base stations, the UDN can meet the requirement of large-scale data transmission, and a foundation is provided for an mMTC scene. On the other hand, densely deployed small cell base stations, macro base stations, micro cell base stations and the like form a multilayer heterogeneous cellular network so as to improve multiplexing gain of a vertical space of the network. The heterogeneous network can improve the system data rate and improve the spectrum utilization efficiency by utilizing the existing network infrastructure and integrating different levels of frequency band resources, so that the heterogeneous network can provide service for high-capacity and high-rate data transmission services in the eMB scene. With the 5G technology enabling various industries, a plurality of emerging communication scenes such as Internet of vehicles, unmanned aerial vehicle communication, Internet of things and the like are developed vigorously, so that a wireless network presents high dynamic characteristics, and the requirements of control related operations in the high dynamic network on the reliability of communication and transmission delay are high.

The spectrum sharing is one of key technologies for improving the spectrum utilization efficiency and relieving the spectrum supply and demand contradiction, plays an important role in a 5G/5G mobile communication system, is a resource guarantee for enabling the high-performance demand diversification application of 5G different scenes, and is a power support unmanned plane swarm network for promoting the long-term development of wireless communication, and the form of a honeycomb is adopted in most cases, but the spectrum resource is more limited than that of a 5G network, so that the spectrum sharing technology of a heterogeneous network is combined with the unmanned plane swarm, and the spectrum supply and demand contradiction is relieved, and the spectrum sharing has a wide application prospect.

In recent years, research on spectrum sharing technology has made remarkable progress. However, considering the complex dynamics of 5G networks, the following deficiencies still exist in the existing related research on spectrum sharing:

(1) for efficient spectrum utilization and sharing, although research has been conducted to expand single time domain sharing of a spectrum to space-time two-dimensional sharing, a spatial spectrum sharing scheme based on geographical location separation is proposed. However, the above scheme only considers coarse-grained location separation among users in the network (i.e., co-channel interference decoupling of two users is based on their distance apart). By considering the scene characteristics of the unmanned plane swarm, users in the network are densely distributed, and the spatial distance between the users is short. Due to the ultra-dense and fine deployment, the existing spatial multiplexing based on coarse-grained geographical location separation is difficult to apply, so that the performance gain of spatial dimension spectrum sharing cannot be obtained. The single time domain dimension shared spectrum is inefficient, and cannot support the wireless access requirements of dense users. In addition, the existing centralized optimization algorithm for learning and updating according to global information often causes huge signaling overhead and intolerable time delay. Thus, the currently existing distributed efficient spectrum sharing techniques still have deficiencies.

(2) For reliable access sharing of frequency spectrum, most of the existing research aims at a static network topology structure, that is, the position of a user node in a network is fixed and invariable, and a future complex network not only contains mobile nodes with more diversified static user nodes, so that the wireless cellular network topology presents dynamics, and the dynamics brings challenges to the robust access of frequency spectrum sharing. Although the prior literature considers the influence of the dynamics of the network environment on the spectrum sharing, the research mostly depends on the ideal channel state information assumption, i.e. the channel state is assumed to be static or quasi-static in the user contention access time slot. Although easy to analyze based on this ideal assumption, there are problems in practical applications: in a wireless network with dynamic characteristics, a large amount of time and signaling resources are consumed for accurately estimating channel state information, and the limitation of competing access time slots of spectrum sharing users is considered, so that accurate channel information cannot be obtained in the limited time slots, and a learning algorithm based on ideal channel state information hypothesis cannot be applied. In summary, currently, research on a spectrum sharing scheme more conforming to reality in a dynamic cellular network is still deficient.

(3) For the design of optimization algorithms, although many iterative learning algorithms have been proposed in the existing research, these algorithms are mostly based on determining a game model, that is, the values of each element in the game are all clear, and in the execution of the algorithms, game participants perform learning of the surrounding environment and updating of strategy selection according to the effect values obtained by their direct observation. However, in an emerging communication scenario with high dynamics, relevant information such as user location, channel state, etc. in the network presents an obvious uncertain characteristic. In this case, since the traditional determined game model cannot describe uncertain information, it is difficult to accurately model the optimization problem under the dynamic uncertain condition, and more importantly, the time-varying effect value of the user caused by the dynamic environment makes it difficult for the existing learning algorithm to give consideration to performance indexes such as the optimization performance, access robustness and implementation complexity, and even leads to the failure of convergence to the stable optimization state. Therefore, there is a significant gap in the current research regarding optimizing learning algorithms in terms of spectrum sharing in emerging high dynamic wireless networks.

Based on the analysis of the existing research, it can be seen that how to implement robust and efficient spectrum sharing in various complex network scenarios such as an ultra-dense network, a heterogeneous cellular network, a high dynamic network, and the like in a swarm is still a difficult point of the current technology, and there is a great theoretical challenge.

Disclosure of Invention

The invention aims to provide a multi-microcellular frequency spectrum resource management method for an unmanned aerial vehicle swarm, which can realize stable and efficient frequency spectrum sharing in various complex network scenes such as an ultra-dense network, a heterogeneous cellular network, a high-dynamic network and the like in the swarm.

The technical solution for realizing the purpose of the invention is as follows: an unmanned aerial vehicle bee colony multi-microcellular frequency spectrum resource management method comprises the steps of constructing an unmanned aerial vehicle bee colony multi-microcellular frequency spectrum network model, describing a frequency spectrum sharing problem into two subproblems of channel scheduling and power distribution, solving an optimal solution of one variable while fixing the other variable, solving the channel scheduling problem into an unconstrained optimization problem, solving the power distribution problem into a constrained convex optimization problem, and solving and analyzing the optimal solution every time of power distribution, wherein the method comprises the following steps:

step 1: in the initial network establishment stage, for the double-layer cellular network of the unmanned aerial vehicle, when the initial network establishment is finished, namely a cluster sub-node and a cluster head exist in a cluster, the cluster head selectively starts a resource scheduling module;

step 2: the cluster subnode sends a resource request packet to a cluster head, and the cluster head performs initial parameter configuration on a resource management module, wherein the initial parameter configuration comprises a working frequency band, the number of distributable RUs and the maximum transmitter power;

and step 3: first-stage information collection: the cluster head judges nodes and RUs which need information collection according to the uplink and downlink demand information of each node, and sends corresponding training sequences to estimate channel CSI;

and 4, step 4: preliminary RU allocation: the cluster head distributes RUs according to a channel scheduling rule based on the collected channel CSI transmitted by each node on the RUs;

and 5: the cluster head judges whether the cluster head receives channel scheduling information of other cluster heads, if so, the initial RU distribution result is combined with the channel distribution results of other cluster heads, and the node which is most likely to be interfered by the cluster head is judged and is called a main interference node; then, the cluster head sends a training sequence to estimate the interference path CSI of the main interference node;

step 6: based on the preliminary RU distribution result, the cluster head calculates the corresponding optimal power on each RU according to the obtained channel information; when the calculated power result is less than zero, taking the power of 0, and correcting the preliminary RU distribution result;

and 7: when the resource scheduling is finished, the resource management module of the cluster head outputs a scheduling result to the MAC layer, and broadcasts the scheduling result of the current round to other cluster heads through the macro cell;

and 8: the next scheduling cycle begins and returns to step 3.

Further, the unmanned aerial vehicle swarm multi-microcellular network model is established as follows:

the unmanned aerial vehicle swarm network is divided into two layers, the upper layer is a macro cell, a centerless ad-hoc network form is adopted, the downlink is a micro cell, a centralized network form taking a cluster head as the center is adopted, the lower layer micro cell is a cluster, the center node is the cluster head and is responsible for inter-cluster interaction and intra-cluster resource scheduling, the bottom layer communication node is a cluster sub-node, and the cluster sub-node performs channel access and data transmission according to the resource scheduling information of the cluster head to which the cluster sub-node belongs; the channel link in the cluster is divided into an uplink and a downlink, the uplink is from a cluster sub node to a cluster head, the downlink is from the cluster head to the cluster sub node, the uplink and the downlink use different frequency bands, and the resource scheduling is mutually independent;

based on the network model, the downlink resource scheduling problem in the unmanned aerial vehicle cellular network is modeled into the following optimized model

Wherein m represents the index number of the microcell, i.e. the cluster head; k represents the index number of the cluster child node in the microcell, and if the child node K belongs to the microcell m, K belongs to K_m，K_mRepresenting a set of cluster child nodes within microcell m; n represents the index number of RUs, and if the serial numbers of RUs in the whole network are synchronous, namely, one n is only corresponding to the RU of one frequency, and if all micro cells share the same large frequency band;represents the long-term average of the channel rate of node k over all RUs in microcell m, and is formulated asIndicating the allocation of RUs ifIn microcell m, RUn is allocated to node k for data transmission, and within the same cell, one RU can serve only one node, but one node can occupy multiple RUs at the same time;means that the sum of the powers of the signals transmitted in microcell m for data transmission on RU with node k should not exceed the total power of the cluster head transmitter

Receiving signal-to-interference-and-noise ratio (SINR) of downlink channel from cluster head m to cluster child node k at RUnIs calculated as follows

And (3) adopting a distributed iteration algorithm, firstly allocating each RU to a child node under the condition of fixed power allocation, then updating the power allocation under the condition of fixed channel scheduling, and after repeated iteration for a limited time, achieving the relatively optimal solution of the resource management optimization problem.

Further, the spectrum sharing problem is described as two sub-problems of channel scheduling and power allocation, the optimal solution of one variable is solved while the other variable is fixed, the channel scheduling problem is an unconstrained optimization problem, the power allocation problem is a constrained convex optimization problem, and each power allocation solution is solved to resolve the optimal solution, specifically as follows:

the first stage is as follows: channel scheduling

Assuming that all RUs are unoccupied during initial network entry, all RUs are occupiedN is the total number of RUs, for each RUn, the child nodes are assigned according to the following rule

If it isThen

WhereinIs the estimated value of the channel instantaneous speed when the sub-node k occupies RUn in the cell m;

and a second stage: power distribution

In the case of fixed RU channel scheduling, the optimization problem in equation (1) is reduced to a convex optimization model as follows

Converting the constrained optimization problem in the formula (2) into an unconstrained optimization problem to solve, wherein

Using the KKT conditional solution to the unconstrained optimization problem in equation (3) to obtain a solution for power allocation

Wherein the content of the first and second substances,indicating varying transmission power of cluster sub-nodes within a cellObtaining an analytical expression by derivation on the influence of channel receiving performance of cluster sub-nodes in adjacent cells subjected to same frequency interference on the same RU;

cell simultaneously subject to inter-cell interference of the same frequencyCapable of more than one, for all cluster subnodes subject to interferenceSum, parameterThe gain of the downlink interference channel is represented, and the summation process is simplified into a multiple relation, specifically:

wherein c is more than or equal to 1.

Further, based on the preliminary RU allocation result, the cluster head calculates the corresponding optimal power for each RU according to the obtained channel information in step 6; when the calculated power result is less than zero, taking the power of 0, and correcting the preliminary RU distribution result at the same time, specifically:

step 6.1: for each cluster head m, and the cluster child nodes K in the cluster head m are belonged to K_mInitializing λ_mWhen the iteration step size is 0, the iteration step size is set to 0.01;

step 6.2: selecting a main interference node and a corresponding cluster head based on a nearest principle according to local channel CSI and adjacent cell CSI obtained through cross-cluster estimation, and calculating an influence value t of a power change alignment link in the cell;

step 6.3: according to the t parameter, calculating the channel occupation parameter for all the nodes k and RU numbers nTransmission Signal Power for RUs other than 0

Step 6.4: judging the corresponding transmitting signal power on each node k and RunWhether the value is greater than 0 or not, and if the value is less than zero, taking a zero value;

step 6.5: ascending: for each cluster child node k within cluster head m, the occupied RU is traversed, i.e.The corresponding power values are summed, and whether the corresponding power values meet the maximum power constraint of the transmitter of the cluster sub-node is judged;

descending: summing all power values of all cluster sub-nodes and RUs in the cluster head m, and judging whether the maximum power constraint of a transmitter of the cluster head node is met;

if the constraint is not satisfied, then λ is_mIncrease of λ_m＝λ_m+0.01, return to step 6.3; otherwise, entering step 6.6;

step 6.6: and finishing the power distribution calculation, and correcting the primary RU distribution, specifically comprising the following steps: if power value of RUnAnd channel occupation flag bitThe channel occupancy flag is set to zero,

compared with the prior art, the invention has the following remarkable advantages: (1) the dynamic heterogeneous network characteristics are considered, and the algorithm is suitable for dynamic networks, such as unmanned aerial vehicle cellular networks, vehicle-mounted communication networks and other channel dynamic change networks, and has good applicability and robustness; (2) aiming at the frequency bandwidth of the completely overlapped ground, the algorithm has higher frequency reuse rate, less information required by the algorithm and low signaling overhead, can realize more reliable and efficient parallel communication in the unmanned aerial vehicle environment with resource shortage, and has good application prospect.

Drawings

Fig. 1 is a schematic diagram of an unmanned swarm network architecture.

Fig. 2 is a flow chart of a continuous resource scheduling algorithm.

Fig. 3 is a flow chart of power calculation in the continuous resource scheduling algorithm.

Fig. 4 is a schematic diagram of a multiplexed RU number of a resource scheduling algorithm based on a continuous rate channel.

Detailed Description

The invention considers a 10km multiplied by 10km network coverage scene, and an unmanned swarm system consisting of 100 unmanned aerial vehicle platforms operates in the network scene. The unmanned swarm system comprises rich heterogeneous main bodies, each unmanned aerial vehicle platform carries different task loads, the functions are different, rapid deployment and allocation can be carried out according to task needs, and various collaborative tasks are efficiently completed. Therefore, the ad hoc network performance of the unmanned bee colony directly determines whether the cooperative task goal can be efficiently realized, especially in a special scene with limited spectrum resources. The invention provides a spectrum resource scheduling algorithm based on a continuous ideal rate channel in an unmanned aerial vehicle swarm network and an implementation method thereof, aiming at dynamic channel access under the condition of strong environmental spectrum restriction. For convenience of description, the network architecture model shown in fig. 1 is specifically described as follows:

considering a 5km multiplied by 5km network coverage scene, 50 unmanned bee colonies formed by unmanned aerial vehicles are distributed in the scene. The unmanned aerial vehicle swarm network is divided into two layers, the upper layer is a macro cell, a centerless ad-hoc network form is adopted, the downlink is a micro cell, a centralized network form taking a cluster head as a center is adopted, a lower micro cell is called a cluster for short, a center node is the cluster head and is responsible for inter-cluster interaction and in-cluster resource scheduling, a bottom communication node is a cluster sub-node, and the cluster sub-node performs channel access and data transmission according to resource scheduling information of the cluster head to which the cluster sub-node belongs. The channel link in the cluster is divided into an uplink and a downlink, the uplink is from a cluster sub node to a cluster head, the downlink is from the cluster head to the cluster sub node, the uplink and the downlink use different frequency bands, and the resource scheduling is mutually independent.

Based on the network model, the downlink resource scheduling problem in the unmanned aerial vehicle cellular network can be modeled as the following optimization model

Where m represents the index number of the microcell (clusterhead); k represents the index number of the cluster child node in the microcell, and if the child node K belongs to the microcell m, K belongs to K_m，K_mRepresenting a set of cluster child nodes within microcell m; n represents the index number of RUs, and if the serial numbers of RUs in the whole network are synchronous, namely, one n corresponds to one RU with one frequency, in order to solve the problem of complex resource sharing in a dense space, all micro cells are assumed to share the same large frequency band;represents the long-term average of the channel rate of node k over all RUs in microcell m, and is specifically formulated asIndicating the allocation of RUs ifIn microcell m, RUn is allocated to node k for data transmission, and within the same cell, one RU can serve only one node, but one node can occupy multiple RUs at the same time;means that in microcell m, the sum of the power of the signals transmitted for data transmission with node k on RU should not exceed the total power of the cluster head transmitterThe calculation formula of the received signal to interference plus noise ratio (SINR) of the downlink channel from the cluster head m to the cluster child node k at RUn is as follows

The key problem of resource scheduling is to select different sub-nodes for each RU in each cell, and different transmission signal powers are not allocated to the sub-nodes, and under the condition that the number of channels is sufficient, each microcell wants to schedule the sub-nodes which are not interfered with adjacent cells at all, and an enumeration method can be used for searching.

The first stage is as follows: channel scheduling

Assuming that all RUs are unoccupied during initial network entry, all RUs are occupiedN is the total number of RUs, for each RUn, its child node is assigned according to the following rule

If it isThen

WhereinThe formula and R are calculated for the estimate of the instantaneous rate of the channel when the child node k occupies RUn in cell m_mkSimilarly, the reference sequence is acquired at the receiving end by transmitting it on the corresponding RU channel.

And a second stage: power distribution

In the case of fixed RU channel scheduling, we reduce the optimization problem in (1) to a convex optimization model as follows

The constrained optimization problem in (2) is considered to be converted into an unconstrained optimization problem to be solved, so that

Using the unconstrained optimization problem in KKT conditional solution (3) to obtain a solution for power allocation

Wherein the content of the first and second substances,indicating varying transmission power of cluster sub-nodes within a cellAnd in mathematical analysis, obtaining an analytical expression by derivation on the influence of the channel receiving performance of cluster sub-nodes in adjacent cells subjected to same frequency interference on the same RU. Since there may be more than one cell simultaneously subject to inter-cell interference, all of the sub-nodes in the cluster subject to interference need to be servedSumming, parameters involved thereinThe gain of the downlink interference channel is represented as a non-local value and is mainly obtained through signaling interaction, and in order to reduce signaling overhead, the summation process is simplified into a multiple relation, specifically represented as

Wherein c is more than or equal to 1, and the performance is better when c is 2 through simulation verification.

The downlink and uplink resource scheduling problem in the unmanned aerial vehicle cellular network is similar to the downlink resource scheduling principle, the working frequency bands are orthogonal, and uplink and downlink cross interference does not exist, so that on the basis of the downlink scheduling problem, the optimal solution of the corresponding uplink resource scheduling problem can be obtained only by considering the difference of interference paths and channel parameters, and redundant description is omitted.

Based on the above derivation, the spectrum sharing problem can be described as two sub-problems: channel scheduling and power allocation. Solving the optimal solution of one variable while fixing the other variable, wherein the channel scheduling problem is an unconstrained optimization problem, the power distribution problem is a constrained convex optimization problem, and the optimal solution is solved by solving the power distribution problem each time, and the algorithm flow is represented by a figure 2 and can be divided into 8 steps

Step 1: and in the initial network establishment stage, for the double-layer cellular network of the unmanned aerial vehicle, when the initial network establishment is completed, namely after cluster subnodes and cluster heads exist in a cluster, the cluster heads select and start the resource scheduling module conditionally.

Step 2: the cluster subnode sends a resource request packet to a cluster head, and the cluster head performs initial parameter configuration on a resource management module, such as working frequency band, number of distributable RUs, maximum transmitter power and the like.

And step 3: first-stage information collection: and the cluster head judges the nodes and RUs which need information collection according to the uplink and downlink requirement information of each node, and sends corresponding training sequences to estimate the channel CSI.

And 4, step 4: preliminary RU allocation: and the cluster head allocates the RUs according to the channel CSI collected and transmitted on the RUs by each node and the channel scheduling rule.

And 5: and the cluster head judges whether the cluster head receives the channel scheduling information of other cluster heads, if so, the initial RU distribution result is combined with the channel distribution results of other cluster heads, and the node which is most likely to be interfered by the cluster head is judged, and the node is called as a main interference node subsequently. And then the cluster head sends a training sequence to estimate the interference path CSI of the main interference node.

Step 6: based on the preliminary RU allocation result, the cluster head calculates the corresponding optimal power on each RU according to the obtained channel information by combining formulas (5) and (6). When the calculated power result is less than zero, the power is taken to be 0, and the preliminary RU allocation result is corrected, and the specific flow is shown in fig. 3 and can be described as the following sub-steps

Step 6.1: for each cluster head m, and the cluster child nodes K in the cluster head m are belonged to K_mInitializing λ_mWhen the iteration step size is 0, the iteration step size is set to 0.01;

step 6.2: selecting a main interference node and a corresponding cluster head thereof according to a local channel CSI and an adjacent cell CSI obtained through cross-cluster estimation and a nearest principle, and calculating an influence value t of a power change alignment link in the cell;

step 6.3: according to the t parameter, calculating the channel occupation parameter for all the nodes k and RU numbers nTransmission Signal Power for RUs other than 0

Step 6.4: judging the corresponding transmitting signal power on each node k and RunWhether the value is greater than 0 or not, and if the value is less than zero, taking a zero value;

step 6.5: up-going- -for each cluster child node k within cluster head m, traverse the RU it occupies, i.e., it isThe corresponding power values are summed, and whether the corresponding power values meet the maximum power constraint of the transmitter of the cluster sub-node is judged; and downlink, summing all power values of all cluster subnodes and RUs in the cluster head m, and judging whether the power values meet the maximum power constraint of the transmitter of the cluster head node. If the constraint is not satisfied, then λ is_mIncrease of λ_m＝λ_m+0.01, return to step 6.3; otherwise, entering step 6.6;

step 6.6: and after the power distribution calculation is completed, correcting the primary RU distribution, specifically describing as follows: if the power value at RUnAnd channel occupation flag bitThe channel occupancy flag is set to zero,the power allocation ends and step 7 is entered.

And 7: after the resource scheduling is finished, the resource management module of the cluster head outputs a scheduling result to the MAC layer, and broadcasts the scheduling result of the current round to other cluster heads through the macro cell, so that the other cluster heads can be dynamically allocated in real time.

And 8: the next scheduling cycle begins and returns to step 3.

The algorithm is preliminarily simulated in matlab. Averagely distributing 9 cluster heads within the range of 10 x 10km, randomly generating 150 cluster subnodes, clustering each node according to the nearest principle, wherein the working frequency is 2GHz, the bandwidth is 5M, considering the existence of isolation bandwidth in practical application, the RU number is 13, and the single RU bandwidth is 345 KHz. Different moving speeds are set for the cluster head and the cluster sub-nodes, in each algorithm period, the cluster head moves by 100m at most, the cluster sub-nodes move by 500m at most, and the moving directions are random.

Continuous channel rate is considered, namely single-channel signal-to-noise ratio constraint does not exist, throughput maximization of the whole network is only realized, simulation results are shown as solid lines in fig. 4, the number of optimized available RUs is more than 100, the multiplexing rate is about 7-8 times, efficient spectrum sharing and interference coordination among same-frequency microcells in a dense cellular networking environment are realized, and more reliable communication conditions are provided for network communication of zhhanchang unmanned aerial vehicles with short resources.

In summary, the invention provides a resource scheduling algorithm based on an ideal continuous rate channel based on an unmanned aerial vehicle swarm double-layer macro-micro cell network, which is used for coordinating frequency spectrum scheduling and power distribution among same-frequency micro cells and realizing frequency spectrum sharing and interference coordination among multiple cells in a dense environment. The method comprises the steps of designing an implementation flow of an algorithm in practical application, verifying the effectiveness of the algorithm based on matlab simulation software, further constructing a double-layer network framework in a system and the simulation software, and designing and implementing related signaling interaction flows including packet format definition, signaling interaction, state conversion and the like.