Energy harvesting method and system for UAV-assisted CEP-TML communication network

By establishing a system model and interference hypergraph model of the UAV assisted CEP-TML communication network, combining learning theory and Lagrangian dual theorem, the resource allocation conflict problem caused by overlapping interference is solved, and the UAV assisted CEP is realized - Effects of maximizing throughput and energy harvesting efficiency in the TML network.

CN118611806BActive Publication Date: 2024-12-03SHENGRISHENG (SHANXI) AIRCRAFT MANUFACTURING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410819757.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-24
Publication Date
2024-12-03
Estimated Expiration
2044-06-24

AI Technical Summary

Technical Problem

In the UAV-assisted CEP-TML communication network, overlapping interference leads to resource allocation conflicts, affecting network throughput and energy harvesting efficiency. How to minimize resource allocation conflicts while maximizing network throughput is a challenge.

Method used

Establish a system model and interference hypergraph model of the UAV-assisted CEP-TML communication network, build an optimization model to maximize network throughput, and decompose it into low-energy and high-energy collection models through learning theory and Lagrangian dual theorem. Applied to solve the gain of imperfect communication links.

Benefits of technology

It effectively reduces the impact of overlapping interference on resource allocation, improves network throughput and resource utilization, extends the service life of drones, and improves the reliability and efficiency of low-power Internet of Things.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118611806B_ABST
    Figure CN118611806B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of resource management of the Internet of Things, and specifically discloses an energy collection method and system for a UAV-assisted CEP-TML communication network. The method and system first establish a system model of the UAV-CEPC-TML network, and then establish an interference model based on a hypergraph based on the system model to achieve communication between CEPs and improve resource utilization under the premise of anti-interference of the UAV-CEPC-TML network. Then, in order to maximize the network throughput under the premise of ensuring the EH of the UAV, an incomplete CSI optimization model of the UAV-CEPC-TML network is established, which combines transmission power constraints, EH constraints, rate constraints and resource allocation conflict tolerance constraints. Finally, the problem is decomposed into two sub-problems, and an IH-EHRA strategy for the UAV-CEPC-TML network is proposed based on learning theory to reduce the impact of imperfect CSI and obtain the optimal solution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of resource management of the Internet of Things, and in particular to an energy collection method and system for a UAV-assisted CEP-TML communication network. Background Art

[0002] In natural disasters, emergencies and remote environments, if the target area lacks base station coverage, consumer electronics products (CEP) often cannot operate normally. To address this problem, a solution has been proposed to use unmanned aerial vehicles (UAVs) to assist in communication. UAVs are used as aerial base stations to collect data and transmit it to other CEPs, or directly to ground base stations to ensure the reliable operation of CEPs.

[0003] In addition, TinyML is an emerging technology that can analyze data and make real-time decisions without relying on cloud computing. Integrating TinyML into CEP (CEP-TML) has many advantages, including reduced latency, enhanced security and privacy, and improved scalability, which can revolutionize various CEPs. However, in UAV-assisted CEP-TML communication (UAV-CEPC-TML) networks, there are also some challenges in achieving low-power IoT: 1) spectrum scarcity due to the increase in the number of users; 2) surging demand for higher data rates; 3) extending the service life of drones.

[0004] In order to solve the problem of limited energy life of drones in low-power IoT networks and support more connections, more and more scholars and industries have begun to study this issue. Therefore, people have proposed the simultaneous wireless information and power transfer (SWIPT) technology to extend the device operation time and enhance wireless connections. This technology can enhance the effectiveness and reliability of wireless communication networks through energy harvesting (EH), reduce equipment maintenance costs, and further promote the development of networks in the direction of environmental protection, energy saving, efficiency, and safety.

[0005] In recent studies, scholars have focused on the EH problem of UAVs, which prolong their flight time and enhance their mission execution capabilities by harvesting energy from radio frequency (RF) signals in the environment. In the current study, researchers fully considered the impact of EH technology on resource management of UAV-assisted communication networks and took effective measures to improve the reliability and effectiveness of low-power IoT networks. However, the high density of mobile CEPs in UAV-CEPC-TML networks brings new challenges to resource allocation. If transmission and reception operations are performed simultaneously on the same device, it may cause interference to the device itself, thereby reducing the quality of the received signal. In addition, the overlapping interference generated in large-scale networks will also reduce the energy harvesting efficiency and lead to resource allocation conflicts. Therefore, in the case of overlapping interference, how to reduce resource allocation conflicts while achieving the maximum throughput of UAV-CEPC-TML networks is a new challenge. Summary of the invention

[0006] The present invention provides an energy collection method and system for a UAV-assisted CEP-TML communication network, and solves the technical problem of how to reduce resource allocation conflicts while achieving the maximum throughput of the UAV-CEPC-TML network.

[0007] In order to solve the above technical problems, the present invention provides an energy collection method for a UAV-assisted CEP-TML communication network, comprising the steps of:

[0008] S1. Establish a system model of UAV-CEPC-TML communication network, i.e., UAV-CEPC-TML network. CEP-TML refers to consumer electronic products integrated with TinyML technology.

[0009] S2. Establishing an interference hypergraph model of the UAV-CEPC-TML network based on the interference relationship in the system model of the UAV-CEPC-TML network;

[0010] S3, based on the system model of the UAV-CEPC-TML network and the interference hypergraph model, an optimization model is constructed to maximize the network throughput while ensuring the energy collection of the UAV;

[0011] S4, decomposing the optimization model range into a low energy collection model and a high energy collection model;

[0012] S5. Applying learning theory to estimate the gain of imperfect communication links The low energy collection model and the high energy collection model are solved using the Lagrange duality theorem.

[0013] Furthermore, in step S1, the system model of the UAV-CEPC-TML network includes B base stations, D cellular CEPs, N D2D pairs and J communication links, each of the D2D pairs consists of a CEP and a UAV, each of the CEPs integrates TinyML technology, the D2D pairs reuse the spectrum of the cellular CEP for communication, the cellular CEP communicates with the base station and the D2D pair through the communication link, and the D2D pair also transmits information to the UAV through the communication link; the UAV collects energy from the environment and transmits data to the base station and the cellular CEP at the same time, and the CEP receives control information from the cellular CEP on the same communication link.

[0014] Further, in the step S2, the interference relationship in the system model of the UAV-CEPC-TML network includes independent interference and cumulative interference;

[0015] Independent interference refers to interference affected by only one CEP-DPs, which is divided into strong interference and weak interference. A group of cellular CEP and D2D pairs constitutes a CEP-DPs. In the UAV-CEPC-TML network, when two CEP-DPs interfere with each other, strong interference occurs when the interference of one CEP-DP exceeds the maximum interference threshold of the other CEP-DPs, otherwise weak interference occurs.

[0016] In the case where multiple CEP-DPs interfere with each other, cumulative interference occurs when the total interference from other CEP-DPs exceeds the maximum interference threshold of any one CEP-DP.

[0017] Furthermore, in the step S2, CEP-DPs are used as nodes of the hypergraph, and the interference relationship of communication between CEP-DPs is used as the hyperedge of the hypergraph, and the constructed hypergraph is the interference hypergraph model.

[0018] Furthermore, in step S3, the optimization model is constructed as follows:

[0019]

[0020] stC1:R n,j ≥R min

[0021] C2:E H ≥E min

[0022] C3:AX≤I

[0023] C4:0≤p n,j ≤p max

[0024] Among them, max means maximization, st means satisfying, and x n,j is the element in the nth row and jth column of the resource matrix X, R n,j represents the user rate of the nth D2D pair on the jth communication link, E H represents the total power consumption of the nth D2D pair, A is the adjacency matrix representing the relationship between nodes in the interference hypergraph model, I is the interference threshold matrix, and p n,j represents the transmission power of the nth D2D pair on the jth communication link, R min The minimum rate threshold, E min represents the minimum energy threshold of each D2D pair, p max represents the maximum power threshold of each D2D pair, C1 represents the minimum rate constraint required by each D2D pair, C2 represents the minimum energy collection constraint of each D2D pair, C3 represents the interference tolerance constraint based on the interference hypergraph model, and C4 represents that the power of the D2D pair cannot exceed the power threshold;

[0025] x n,j Defined as:

[0026]

[0027] Among them, v n Represents the nth node in the interference hypergraph model.

[0028] Furthermore, the total power consumption E of the nth D2D pair H Calculated by the following formula:

[0029]

[0030] Among them, p i,j represents the transmission power of the i-th D2D pair on the j-th communication link, g i,j represents the channel gain of the i-th D2D pair on the j-th communication link, i≠n, P is the total transmit power of the base station, G represents the communication link gain from the base station to the D2D pair, and α is the energy conversion efficiency;

[0031] The user rate R of the nth D2D pair on the jth communication link n,j Calculated by the following formula:

[0032] R n,j =B 1 log 2 (1+γ n )

[0033] Among them, B 1 represents the bandwidth of each communication link, γ n represents the signal-to-noise ratio of the nth D2D pair.

[0034] Furthermore, in step S4, when E min ≤αγ min σ 2 When , the optimization model scope is decomposed into the low energy harvesting model:

[0035]

[0036] C2:E H ≥E min

[0037] C3:AX≤I

[0038] C4:0≤p n,j ≤p max

[0039] C5:E min ≤αγ min σ 2

[0040] Among them, σ 2 is the noise power, For customized parameters;

[0041] When E min >αγ min σ 2 When , the optimization model is decomposed into the high energy collection model:

[0042]

[0043] C3:AX≤I

[0044] C4:0≤p n,j ≤p max

[0045]

[0046] in, A custom auxiliary variable.

[0047] Furthermore, the step S5 specifically includes the steps of:

[0048] Collect multiple imperfect channel gain samples to form an original sample set;

[0049] The original sample set is subjected to data set division, shape learning and size calibration to obtain the communication link gain.

[0050] Reconstructing the low energy collection model and the high energy collection model into corresponding Lagrangian dual problems;

[0051] Utilizing Communication Link Gain And the Lagrangian dual problem of the low energy collection model and the high energy collection model is solved by Lagrangian KKT condition to obtain corresponding solutions.

[0052] Further, in obtaining the communication link gain During the learning process, each iteration point is updated by the following formula:

[0053]

[0054] in, is the transmission power p at the kth iteration n,j , is the transmission power p at the k+1th iteration n,j , is the step size of the kth iteration, express The Lagrangian function value under represents the gradient;

[0055] During the iteration process, when When , the iteration ends and the communication link gain is output Θ represents the convergence accuracy, express The Lagrangian function value under .

[0056] The present invention also provides an energy collection system for a UAV-assisted CEP-TML communication network, the key of which is to apply the energy collection method for the UAV-assisted CEP-TML communication network.

[0057] The energy collection method and system of the UAV-assisted CEP-TML communication network provided by the present invention, in order to solve the effective resource management problem of the UAV-CEPC-TML network in the overlapping interference scenario and realize the low-power Internet of Things, firstly establishes a system model of the UAV-CEPC-TML network, and then establishes an interference model based on a hypergraph based on the system model, so as to realize the communication between CEPs and improve resource utilization under the premise of anti-interference of the UAV-CEPC-TML network. Then, in order to maximize the network throughput under the premise of ensuring the EH of the UAV, this embodiment establishes an incomplete CSI optimization model of the UAV-CEPC-TML network, which combines the transmission power constraint, EH constraint, rate constraint and resource allocation conflict tolerance constraint. Finally, this embodiment decomposes the problem into two sub-problems, and proposes an IH-EHRA strategy for the UAV-CEPC-TML network based on learning theory to reduce the impact of imperfect CSI and obtain the optimal solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a system model diagram of a UAV-assisted CEP-TML communication network provided by an embodiment of the present invention;

[0059] Figure 2 is an example diagram of an interference model provided by an embodiment of the present invention;

[0060] Figure 3 is an example diagram of an interference hypergraph model provided by an embodiment of the present invention;

[0061] Figure 4 is the power threshold P provided in the embodiment of the present invention max Relationship diagram with the throughput of DDs;

[0062] Figure 5 is the power threshold P provided in the embodiment of the present invention max The relationship diagram with the interference efficiency of DDs;

[0063] Figure 6 is the rate threshold R provided in the embodiment of the present invention min Relationship diagram with UAV-CEPC-TML network performance;

[0064] Figure 7 is a relationship diagram between the number of DDs and interference efficiency provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The following specifically illustrates the implementation mode of the present invention in conjunction with the accompanying drawings. The embodiments are provided for illustrative purposes only and cannot be understood as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0066] The energy collection method of the UAV-assisted CEP-TML communication network provided in the embodiment of the present invention comprises the following steps:

[0067] S1. Establish a system model of UAV-CEPC-TML communication network, i.e., UAV-CEPC-TML network. CEP-TML refers to consumer electronic products integrated with TinyML technology.

[0068] S2, establishing an interference hypergraph model of the UAV-CEPC-TML network based on the interference relationship in the system model of the UAV-CEPC-TML network;

[0069] S3, based on the system model and interference hypergraph model of UAV-CEPC-TML network, an optimization model is constructed to maximize the network throughput while ensuring the energy collection of UAV;

[0070] S4, decomposing the optimization model range into a low energy collection model and a high energy collection model;

[0071] S5. Applying learning theory to estimate the gain of imperfect communication links The Lagrange duality theorem is used to solve the low energy harvesting model and the high energy harvesting model.

[0072] (1) Step S1: Establishing the system model of the UAV-CEPC-TML network

[0073] The scenario studied in this paper is the UAV-CEPC-TML scenario using D2D technology, which aims to realize a low-power IoT network, such as Figure 1 The UAV-CEPC-TML network includes B base stations (BS), D cellular CEPs, N D2D pairs and J communication links. BS Indicates that the bth base station is denoted as BS b , and there is Denotes any D cellular CEPs by set D CEP In other words, the dth CEP is represented by CEP d , and there is Each CEP integrates TinyML technology to achieve real-time data analysis and decision-making. In order to solve the challenge of spectrum scarcity in wireless communications, the UAV-CEPC-TML network is equipped with N D2D pairs (or DDs / D2D devices). Each D2D pair consists of a CEP and a drone. The D2D devices reuse the spectrum of the cellular CEP for communication. The N D2D pairs use a set of N DD Indicates that the nth D2D is represented by DD n ,

[0074] In the CEPC-TML scenario, the cellular CEP communicates with the BS and the D2D pair through the communication link, and the D2D pair also transmits information to the UAV through the communication link. Specifically, the UAV collects energy from the environment and transmits data to the BS and cellular CEP at the same time, while the CEP receives control information from the device on the same communication link. In the D2D network, each UAV is equipped with a SWIPT function. Since the trajectory optimization of the UAV is not considered, electromagnetic energy can be stably collected from the surrounding environment, effectively extending the service life of the UAV. This example defines the communication link set as J LK , where the jth communication link is represented by LK j , and there is

[0075] Since the sampled additive white Gaussian noise in the process of converting the RF band to the baseband signal is less than the sum of the interference power of the main base station and the DDs, the EH of the noise can be assumed to be zero. Therefore, the signal-to-noise ratio (SINR) of the dth cellular CEP and the nth DDs are expressed as:

[0076]

[0077] Among them, p d,j and p n,j are the transmission power of the d-th cellular CEP and the n-th D2D pair on the j-th communication link, respectively, and h d and h n are the communication link gains of the dth cellular CEP and the nth D2D pair respectively. i,j It refers to the transmission power of the i-th D2D pair on the j-th communication link, g n,j is the channel gain of the nth D2D pair in the jth communication, and g i,j It refers to the channel gain of the i-th D2D pair on the j-th communication link. P is the total transmit power of the BS, G is the communication link gain from the BS to the D2D pair, σ 2 is the noise power. At the same time, the time-varying nature of the wireless communication link between DDs and the non-real-time communication link estimation will lead to imperfect CSI.

[0078] (2) Step S2: Constructing an interference hypergraph model

[0079] In the UAV-CEPC-TML network, this example uniformly refers to the cellular CEP and D2D pairs as CEP-DPs. Interference is divided into two categories: independent interference and cumulative interference. Independent interference refers to interference affected by only one CEP-DPs, which is divided into strong interference and weak interference. In the UAV-CEPC-TML network, when two CEP-DPs interfere with each other, strong interference occurs when the interference of one CEP-DP exceeds the maximum interference threshold (mIT) of the other CEP-DPs, otherwise weak interference occurs. In the case of multiple CEP-DPs interfering with each other, cumulative interference occurs when the total interference from other CEP-DPs exceeds the mIT of any CEP-DPs. It is worth noting that strong interference and weak interference may exist simultaneously in cumulative interference.

[0080] In the UAV-CEPC-TML scenario, interference has the potential to compromise network stability. On the one hand, when independent interference occurs, it may result in packet loss or increased latency. On the other hand, cumulative interference can overwhelm the network and cause overload. It is critical to address the interference problem in network resource allocation to mitigate its impact. In the UAV-CEPC-TML network, strong interference can be minimized through proper resource management, thereby reducing its impact on the network. However, as the number of network devices increases, the cumulative interference caused by weak interference within the overlapping communication range is less considered. Therefore, the subsequent work of this embodiment will focus on solving the problem of cumulative interference caused by weak interference. Interference types such as Figure 2 shown.

[0081] The single connectivity of nodes and edges in traditional graphs limits the expressiveness of the graph and cannot fully reflect the relationship of cumulative interference. Compared with traditional graphs, hypergraphs provide a more flexible solution to better evaluate and optimize communication quality. In a hypergraph, the connection relationship between nodes can be many-to-many, and one edge can connect multiple nodes. This flexibility enables the hypergraph to more accurately represent the interactions in cumulative interference. Therefore, this embodiment uses hypergraph theory to analyze the interference types in the network and construct a hypergraph model, which helps to better understand and deal with interference phenomena in the network.

[0082] The hypergraph can be represented as G = {V, E}, where V = {ν 1 ,ν 2 ,...} is the set of nodes in the hypergraph, E={e 1 ,e 2 ,...} is a set of hyperedges. In this example, the nodes of the hypergraph represent CEP-DPs, and the edges represent the interference relationship of communication between CEP-DPs.

[0083] In order to further understand the occupation of channels by CEP-DPs in the hypergraph, this embodiment creates a resource matrix X, which consists of the following rules:

[0084]

[0085] where x n,j is the element in the nth row and jth column of the resource matrix X, v n Represents the nth node in the interference hypergraph model.

[0086] According to the definition of the hypergraph model, this embodiment regards the nodes in the hypergraph as CEP-DPs and connects multiple nodes in the hyperedge to indicate that there is an interference relationship between each CEP-DPs, thereby establishing an interference hypergraph model example. The connection of straight lines means that CEP-DPs exist in 2 or more hyperedges, that is, they are located in corresponding overlapping areas.

[0087] Figure 3 An example of an interference hypergraph model is given, which contains 5 hyperedges (e 1 ,e 2 ,...,e 5 ) and 12 nodes (P 1 ,P 2 ,...,P 12 ). This model describes the strong interference and cumulative interference relationship between multiple CEP-DPs, that is, CEP-DPs in the same super edge interfere with each other and cannot use the same resources. The CEP-DPs in the region can be divided into non-overlapping regions (P 1 ,P 2 ,P 3 ,P 8 ,P 10 ) and the overlapping area (P 4 ,P 5 ,P 6 ,P 7 ,P 9 ,P 11 ,P 12 )(CEP-DPs are associated with multiple hyperedges).

[0088] The hyperedge is built on the definition of independent interference and cumulative interference. For ease of understanding, Figure 3 As an example. When P 4 The interference generated is greater than P 6 When the mIT is greater than , a super edge will be formed between them, representing strong interference among independent interferences. 1 , P 5 and P 6 , P 11 In the communication relationship, one of the CEP-DPs is selected, and the interference generated by all other CEP-DPs is greater than the mIT of this CEP-DPs, and a super edge is formed between them, indicating the cumulative interference.

[0089] The relationship between nodes in the interference hypergraph model can usually be represented by an adjacency matrix A, which consists of the following rules:

[0090]

[0091] in is the nth adjacency matrix 1 Row and nth 2 The value of the column, n 1 、n 2 =1,2,…,N. Figure 3 The adjacency matrix of the interference hypergraph model can be expressed as A = {0,1} N×N , A is expanded as follows:

[0092]

[0093] When the value of a row or a column is 1, it indicates that there is interference between the two CEP-DPs, and when it is 0, it indicates that there is no interference.

[0094] In order to improve resource utilization, this embodiment assumes that under certain interference, the CEP-DP resources in the same hyperedge may be different, that is, the following conditions are met:

[0095] AX≤I (6)

[0096] Where I is the interference threshold matrix.

[0097] (3) Step S3: Constructing an optimization model

[0098] In the above interference model, this embodiment analyzes the interference relationship and resource conflict relationship between CEP-DPs and obtains the resource relationship matrix between devices. Through this matrix, the resource allocation between nodes can be observed. When the value in the matrix is ​​less than the value of the interference threshold matrix, it indicates that the nodes can communicate with each other under certain interference conditions; otherwise, communication is impossible. This method helps to resolve conflicts and optimize the configuration of communication resources. However, in this network, the life of the UAV is a problem worthy of attention, which requires EH algorithms and strategies to extend the use time of the UAV and maximize the throughput of the UAV-CEPC-TML network.

[0099] According to formula (2), the user rate expression of the nth UAV on the jth communication link is obtained as:

[0100] R n,j =B 1 log 2 (1+γ n ) (7)

[0101] In the formula, B 1 Indicates the bandwidth of each communication link.

[0102] This embodiment ignores the EH noise in the network model. Therefore, this embodiment models the EH of the drone as linear, expressed as:

[0103]

[0104] Where α is the energy conversion efficiency. Equation (8) represents the total power consumption of the nth DDs, that is, the captured energy should meet the actual power consumption of the DDs.

[0105] In the UAV-CEPC-TML network, the resource allocation problem is achieved by maximizing the network throughput under the premise of ensuring the EH of the UAV, and the maximization of the network throughput is based on the premise of anti-interference. Therefore, this embodiment comprehensively considers the constraints such as transmission power, energy collection, and resource allocation conflict tolerance, and establishes an optimization model with the goal of maximizing the UAV-CEPC-TML network throughput, as shown in formula (9):

[0106]

[0107] stC1:R n,j ≥R min

[0108] C2:E H ≥E min

[0109] C3:AX≤I

[0110] C4:0≤p n,j ≤p max

[0111] Among them, R min Indicates the minimum allowable rate, E min represents the minimum allowable energy, p max Represents the maximum allowable power. C1 represents the required minimum rate constraint, C2 represents the minimum EH constraint, C3 represents the interference tolerance constraint based on the interference hypergraph, and C4 represents that the power cannot exceed the power threshold. max represents maximization, and st represents satisfying.

[0112] (4) Step S4: Model transformation

[0113] From the optimization model, we can see that the communication link gain g of the UAV-CEPC-TML network under imperfect CSI is n,j is uncertain, making it impossible for the model to solve the global optimal solution using traditional optimization methods. Therefore, this example uses the idea of ​​learning theory to train random probability constraints, obtain a more accurate uncertainty set, and apply it to the solution of the robust optimization model. This example explores the EH problem by decomposing it into two different sub-problems. In order to solve this problem, this embodiment proposes an IH-EHRA strategy, which not only effectively solves the problem, but also improves the performance of the system. In this embodiment, this embodiment derives and analyzes two different EH problems in the UAV-CEPC-TML network respectively. In order to analyze these two models, this embodiment first proposes Lemma 1.

[0114] Lemma 1: E min With αγ min σ 2Comparing them, we get different results:

[0115]

[0116] in, For customized parameters.

[0117] Proof: For the convenience of expression, based on the definition of SINR, this embodiment can assume that Z n is an auxiliary variable, i.e. The rate constraint can be rewritten as:

[0118]

[0119] in It is a custom parameter. n >0, formula (11) can be rewritten as:

[0120] p n,j g n,j ≥γ min σ 2 (12)

[0121] After transformation, formula (11) can be rewritten as:

[0122] p n,j g n,j ≥G n (13)

[0123] According to formula (12), formula (13) is transformed into:

[0124]

[0125] In the formula is a custom parameter, and max(,) means taking the larger value in the brackets. Formula (14) can be further written as:

[0126]

[0127] Therefore, the proof is complete.

[0128] By converting the optimization model into a low-energy harvesting (LEH) problem, the energy utilization efficiency and resource allocation efficiency of the system can be further studied. This helps the present embodiment to design a more efficient energy harvesting system and optimize resource allocation in the UAV-CEPC-TML network, thereby improving the energy utilization and sustainable development of the UAV.

[0129] LEH models are better suited for applications with small or distributed energy needs, such as personal electronic devices, IoT sensors, and small home solar systems. These models are lower cost, simpler in design, and generally offer greater flexibility.

[0130] When E min ≤αγ min σ 2 When established, we get:

[0131] p n,j g n,j ≥γ min σ 2 (16)

[0132]

[0133] Since α∈[0,1], this embodiment can obtain:

[0134]

[0135] This means max(E min -Z n )≤γ min σ 2 .

[0136] From the above derivation, we can see that when E min ≤αγ min σ 2 When , the problem is transformed into the LEH problem. Formula (9) is reformulated as:

[0137]

[0138] In the UAV-CEPC-TML network, through the above analysis, this embodiment transforms the EH problem into the LEH problem. It is worth noting that in the LEH problem, there is another constraint C5 that limits the EH efficiency. Through the analysis of the LEH model, it can be seen that when the energy collection threshold is less than αγ min σ 2 When the EH model is changed to the LEH model, it provides more accurate information for the UAV-CEPC-TML network, making small and medium-sized devices in the network more energy-efficient and improving the sustainability of the devices.

[0139] According to the same method, we can know that when E min ≥αγ min σ 2When the UAV-CEPC-TML network is connected, the problem becomes a high-energy harvesting (HEH) problem. The HEH model can be applied to energy-saving base station construction, high-altitude UAV networks, wireless sensor networks and other scenarios, providing stable and powerful energy support for these fields.

[0140] First, based on the LEH model, this embodiment assumes that when E min ≥αγ min σ 2 When the equation is established.

[0141]

[0142] The detailed derivation is as follows. In order to prove that formula (20) is valid, this embodiment assumes that:

[0143]

[0144] Established.

[0145] Therefore, formula (21) can be further written as:

[0146]

[0147] Since formula (17) holds, formula (22) can be expressed as:

[0148]

[0149] Since the derivation of equation (23) is equivalent to SINR, the assumption holds.

[0150] Based on the assumption of (20), C1 and C2 can be integrated into a constraint, that is, Therefore, when E min ≥αγ min σ 2 When p n,j g n,j ≥γ min σ 2 -Z n In the UAV-CEPC-TML network, the optimization problem (9) becomes a HEH problem, which can be expressed as:

[0151]

[0152] The high energy collection model can provide stronger power support for energy-limited networks, enhance the autonomy of network nodes, expand the deployment range of the network, and extend the life cycle of equipment. This means that the system can more fully collect and utilize energy above this threshold. This helps to improve energy utilization efficiency, achieve longer continuous energy supply, and reduce dependence on external energy supply.

[0153] Based on equations (19) and (24), two models with different EH efficiencies are obtained, namely the low energy harvesting model and the high energy harvesting model. These models can be used to describe the EH process in the UAV-CEPC-TML network and evaluate the impact of different parameters on the EH efficiency. However, in these models, due to the imperfect CSI environment of the UAV-CEPC-TML network, the communication link gain g n,j The uncertainty of the network brings challenges to the system. Specifically, obstacles, CEP-TML mobility, weather changes and other factors lead to imperfect CSI in the UAV-CEPC-TML network environment, affecting the optimization of the system.

[0154] (5) Step S5: Model solution

[0155] In order to cope with the challenge of communication link gain uncertainty in the UAV-CEPC-TML network, in this example, the communication link gain can be modeled as a polyhedron model (19) and (24) to describe the communication link uncertainty. The specific expression is as follows:

[0156]

[0157] In the formula and T are parameters learned from the sample dataset.

[0158] Therefore, this embodiment should collect M imperfect samples The sample is g n,j Learning, R represents the set of real numbers. Then, the confidence method is used to construct the prior distribution of imperfect CSI, and equation (25) is transformed into:

[0159]

[0160] In the formula, 1-θ is the specified confidence level, represents a probability threshold, and Pr{} represents the probability of the event in the brackets {}. In order to obtain relevant parameters, this embodiment designs a learning method to determine these parameters, which includes data set division, shape learning and size calibration.

[0161] Data set division: In this embodiment, the sample set Divide into several independent sample sets, remove samples far from the center, and get a more accurate sample set.

[0162] Shape Learning: yes The center of can be selected as the sample mean, that is:

[0163]

[0164] Next, this embodiment needs to confirm the size of the polyhedron model.

[0165] Size calibration: The task of size calibration is to adjust the uncertainty set so that it is aligned with Eq. (26) with a confidence level of 1-θ. The main challenge lies in estimating the quantiles of the data sample transformation:

[0166]

[0167] Where D(0) represents the data set The first data, D(1) represents the data set The second data.

[0168] For calibration This example assumes that the sample data set The 1-θ quantile q of the basic distribution of F(D) 1-θ for:

[0169] Pr{F(D)≤q 1-θ}=1-θ (29)

[0170] Therefore, this embodiment can obtain The function values ​​F(D) of all samples in are sorted in ascending order. ( represents rounding up) can be used as an uncertain set The size of F(D*) is given by 1-θ. Based on this analysis, the size of F(D*) can be given as:

[0171] Pr{F(D*)≥q 1-θ}=1-θ (30)

[0172] Finally, this embodiment obtains T as:

[0173] T=F(D*) (31)

[0174] By substituting formulas (27) and (31) into formula (25), the relevant data of the communication link gain in the UAV-CEPC-TML network can be obtained.

[0175] Through the above learning, this embodiment obtains a more accurate channel gain. Afterwards, this embodiment will solve the solutions of two different energy collection models. In order to maximize the throughput of the UAV-CEPC-TML network under different EH models, this embodiment adopts the IH-EHRA strategy to solve the model. In addition, considering the random parameters involved in the UAV-CEPC-TML network, the optimization model belongs to an NP-hard optimization problem. Therefore, this embodiment uses the Lagrange duality theorem to transform the optimization model into a convex optimization problem.

[0176] The pseudo code of the IH-EHRA energy harvesting method is shown in Algorithm 1 in Table 1. The specific steps are as follows:

[0177] 1) Get a sample dataset

[0178] 2) Sample data set For feature extraction and parameter learning, including data set separation, shape learning and size calibration steps to learn channel gain parameters and obtain communication link gain

[0179] In each iteration, the parameters are updated as follows:

[0180]

[0181] in, are the transmission power p at the kth and k+1th iterations respectively. n,j , is the step size of the kth iteration, express The Lagrangian function value under represents the gradient;

[0182] when (Θ represents convergence accuracy), the iteration ends and the communication link gain is output express The Lagrangian function value under ;

[0183] 3) Finally, the two different EH models are reconstructed into corresponding Lagrangian dual problems, using the communication link gain And the corresponding solution is obtained through Lagrangian KKT conditions.

[0184] Table 1

[0185]

[0186] An embodiment of the present invention further provides an energy collection system for a UAV-assisted CEP-TML communication network, which applies an energy collection method for a UAV-assisted CEP-TML communication network.

[0187] (6) Simulation analysis

[0188] This example verifies the performance of the proposed IH-EHRA algorithm by providing numerical and simulation results. In addition, this example also compares the joint robust and non-robust algorithms. The effectiveness of the IH-EHRA strategy for the two sub-problems of the IH-EHRA algorithm for high energy harvesting and the IH-EHRA algorithm for the second energy harvesting is verified. The simulation scenario is a UAV-CEPC-TML network using D2D technology. The coverage of the BS is a circle with a radius of 500 meters, and the BS bandwidth B 1 The frequency is 5 MHz, the number of CEP-TML is 20, and the number of data sets is 1000. The other simulation parameter settings are shown in Table 2.

[0189] Table 2 Simulation parameter settings

[0190] parameter value The number of D2D pairs N 20 CEP-TML quantity D 20 <![CDATA[BS Bandwidth B 1 > 5MHz BS total transmission power P 10mW The communication link gain G from BS to device 3 Interference power σ 0.4Mw Confidence level θ 0.02 Energy conversion rate α 1 BS coverage radius 500m Rate probability threshold probability ζ 0.80 <![CDATA[Minimum rate R min > 5bps / Hz <![CDATA[Maximum power p max > 10mW

[0191] Figure 4 It represents the power threshold P max Relationship with the throughput of DDs. It can be seen that the throughput is ranked from high to low in order of the IH-EHRA algorithm with high energy collection, the joint robust comparison algorithm, the EHRA algorithm with low energy collection, and the non-robust algorithm. This shows that the proposed strategy can achieve excellent throughput performance in the high energy collection model. This shows that under the same transmission power, the IH-EHRA algorithm reduces the impact of imperfect CSI on the communication link, thereby improving the success rate and throughput of data transmission. At the same time, it can be seen that as the power threshold increases, the throughput of the IH-EHRA algorithm with high energy collection will also increase. This is because as the power threshold increases, DDs can transmit data faster, resulting in an increase in SINR in the system, thereby increasing the throughput of the UAV-CEPC-TML network.

[0192] Figure 5 Indicates the power threshold P max The relationship between the interference efficiency of DDs. Interference efficiency refers to the number of bits transmitted per unit of interference energy applied to DDs. Figure 5 It can be seen that the interference efficiency of DDs is from high to low, which is the joint robust algorithm, the IH-EHRA algorithm with high energy collection, the IH-EHRA algorithm with low energy collection, and the non-robust algorithm. This shows that the proposed IH-EHRA algorithm can effectively resist the influence of overlapping interference and improve the communication quality and security of DDs.

[0193] Figure 6 The rate threshold R is shown min Relationship with UAV-CEPC-TML network performance, Figure 6(a) is the rate threshold R min The relationship between the interference efficiency and Figure 6 (b) is the rate threshold R min The relationship between the throughput of DDs and the Figure 6 As can be seen from (a), the IH-EHRA algorithm has lower IE when dealing with these two sub-problems compared to the baseline algorithm. This indicates that the IH-EHRA algorithm can reduce interference more effectively, thus achieving better performance. Figure 6 As can be seen from (b), the IH-EHRA algorithm has a higher throughput. This is because the IH-EHRA algorithm can improve the reliability and throughput of data transmission by reducing interference, thereby improving the overall performance of the communication system.

[0194] Figure 7 The relationship between the number of DDs and the interference efficiency is shown. Figure 7 It can be seen that as the number of DDs increases, the IE of the joint robust algorithm increases, while the IE of the IH-EHRA algorithm decreases. This is because as the number of DDs increases, the amount of data transmitted in the network will also increase, resulting in increased interference and increased communication link uncertainty. This shows that the IH-EHRA algorithm can effectively reduce the impact of interference on CSI when the number of DDs in the UAV-CEPC-TML network increases.

[0195] Under different environments, the IH-EHRA algorithm outperforms the baseline algorithm in terms of interference efficiency and throughput. This shows that the proposed IH-EHRA algorithm outperforms the baseline algorithm and can effectively maximize the UAV-CEPC-TML network throughput in the UAV-CEPC-TML scenario under dense overlapping interference, ensuring the EH requirements of the UAV.

[0196] In summary, in order to solve the problem of effective resource management of UAV-CEPC-TML network in overlapping interference scenarios and realize low-power Internet of Things, the embodiment of the present invention proposes an energy collection method and system for UAV-assisted CEP-TML communication network. First, a system model of UAV-CEPC-TML network is established, and then a hypergraph-based interference model is established based on the system model to realize communication between CEPs and improve resource utilization under the premise of anti-interference of UAV-CEPC-TML network. Then, in order to maximize the network throughput under the premise of ensuring the EH of UAV, this embodiment establishes an incomplete CSI optimization model of UAV-CEPC-TML network, which combines transmission power constraints, EH constraints, rate constraints and resource allocation conflict tolerance constraints. Finally, this embodiment decomposes the problem into two sub-problems, and proposes an IH-EHRA strategy for UAV-CEPC-TML network based on learning theory to reduce the impact of imperfect CSI and obtain the optimal solution. The simulation results verify the effectiveness of the scheme in the UAV-CEPC-TML scenario, which can effectively improve the performance of low-power Internet of Things network.

[0197] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. The energy harvesting method of UAV-assisted CEP-TML communication network is characterized in that: Includes steps: S1. Establish a system model of UAV-CEPC-TML communication network, i.e., UAV-CEPC-TML network. CEP-TML refers to consumer electronic products integrated with TinyML technology. In step S1, the system model of the UAV-CEPC-TML network includes B base stations, D cellular CEPs, N D2D pairs and J communication links, each of the D2D pairs consists of a CEP and a UAV, each of the CEPs integrates the TinyML technology, the D2D pairs reuse the spectrum of the cellular CEP for communication, the cellular CEP communicates with the base station and the D2D pair through the communication link, and the D2D pair also transmits information to the UAV through the communication link; the UAV collects energy from the environment and transmits data to the base station and the cellular CEP at the same time, and the CEP receives control information from the cellular CEP on the same communication link; S2. Establishing an interference hypergraph model of the UAV-CEPC-TML network based on the interference relationship in the system model of the UAV-CEPC-TML network; In the step S2, the interference relationship in the system model of the UAV-CEPC-TML network includes independent interference and cumulative interference; Independent interference refers to interference affected by only one CEP-DPs, which is divided into strong interference and weak interference. A group of cellular CEP and D2D pairs constitutes a CEP-DPs. In the UAV-CEPC-TML network, when two CEP-DPs interfere with each other, strong interference occurs when the interference of one CEP-DP exceeds the maximum interference threshold of the other CEP-DPs, otherwise weak interference occurs. In the case of mutual interference between multiple CEP-DPs, cumulative interference occurs when the total interference from other CEP-DPs exceeds the maximum interference threshold of any one CEP-DP; In the step S2, the CEP-DPs are used as nodes of the hypergraph, and the interference relationship of the communication between the CEP-DPs is used as the hyperedge of the hypergraph, and the constructed hypergraph is the interference hypergraph model; S3, based on the system model of the UAV-CEPC-TML network and the interference hypergraph model, an optimization model is constructed to maximize the network throughput while ensuring the energy collection of the UAV; In step S3, the optimization model is constructed as follows: Among them, max means maximization, st means satisfying, and x n,j is the element in the nth row and jth column of the resource matrix X, R n,j represents the user rate of the nth D2D pair on the jth communication link, E H represents the total power consumption of the nth D2D pair, A represents the adjacency matrix of the relationship between nodes in the interference hypergraph model, I is the interference threshold matrix, and p n,j represents the transmission power of the nth D2D pair on the jth communication link, R min The minimum rate threshold, E min represents the minimum energy threshold of each D2D pair, p max represents the maximum power threshold of each D2D pair, C1 represents the minimum rate constraint required by each D2D pair, C2 represents the minimum energy collection constraint of each D2D pair, C3 represents the interference tolerance constraint based on the interference hypergraph model, and C4 represents that the power of the D2D pair cannot exceed the power threshold; x n,j Defined as: Among them, v n Represents the nth node in the interference hypergraph model; The total power consumption E of the nth D2D pair H Calculated by the following formula: Among them, p i,j represents the transmission power of the i-th D2D pair on the j-th communication link, g i,j represents the channel gain of the i-th D2D pair on the j-th communication link, i≠n, P is the total transmit power of the base station, G represents the communication link gain from the base station to the D2D pair, and α is the energy conversion efficiency; The user rate R of the nth D2D pair on the jth communication link n,j Calculated by the following formula: R n,j =B1 log2(1+γ n ) Among them, B1 represents the bandwidth of each communication link, γ n represents the signal-to-noise ratio of the nth D2D pair; S4, decomposing the optimization model into a low energy collection model and a high energy collection model; In step S4, when E min ≤αγ min σ 2 When , the optimization model is decomposed into the low energy harvesting model: Among them, σ 2 is the noise power, For customized parameters; When E min >αγ min σ 2 When , the optimization model is decomposed into the high energy collection model: in, It is a custom auxiliary variable; S5. Applying learning theory to estimate the gain of imperfect communication links and using Lagrange duality theorem to solve the low energy collection model and the high energy collection model; The step S5 specifically includes the following steps: Collect multiple imperfect channel gain samples to form an original sample set; The original sample set is subjected to data set division, shape learning and size calibration to obtain the communication link gain. Reconstructing the low energy collection model and the high energy collection model into corresponding Lagrangian dual problems; Utilizing Communication Link Gain And solving the Lagrangian dual problem of the low energy collection model and the high energy collection model through the Lagrangian KKT condition to obtain corresponding solutions; Gaining communication link gain During the learning process, each iteration point is updated by the following formula: in, is the transmission power p at the kth iteration n,j , is the transmission power p at the k+1th iteration n,j , is the step size of the kth iteration, express The Lagrangian function value under represents the gradient; During the iteration process, when When , the iteration ends and the communication link gain is output Θ represents the convergence accuracy, express The Lagrangian function value under .