Optimization method for energy transceiving and trajectory design of unmanned aerial vehicle in cellular-removed mMIMO system

Abstract

The invention discloses an optimization method for energy transceiving and trajectory design of an unmanned aerial vehicle (UAV) in a cellular-removed mMIMO system, which comprises the following steps of: establishing a CF-mMIMO system model, defining an optimization objective function to maximize the throughput of the system, ensuring the communication fairness among different mobile users while ensuring the endurance of the UVA, and improving the throughput of the system. A DQN algorithm model based on communication assistance is selected to solve the optimization problem, and an optimal UAV energy receiving and transmitting and trajectory distribution scheme is obtained; according to the optimization method for energy transceiving and trajectory design of the unmanned aerial vehicle in the honeycomb-removed mMIMO system, the throughput of the system is maximized under the condition that the energy limitation of the unmanned aerial vehicle (UAV) and the charge-discharge time slot constraint are met, the cruising ability of the UAV is improved, meanwhile, the communication fairness of mobile users is ensured, the system can adaptively cope with various complex scenes through an intelligent decision-making mechanism, and the system performance is improved. The robustness and flexibility of the system are improved, so that the endurance time is maximized, and the overall communication performance of the system is improved.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology, and in particular to an optimization method for obtaining energy transceiver and trajectory design of a drone in a decellularized MMIMO system using deep reinforcement learning. Background Art

[0002] In recent years, with the rapid development of portable devices, inter-device communications, and Internet of Things applications, the demand for wireless communication services and wireless data traffic has continued to increase. Cellular-free massive multiple-input multiple-output (CF-mMIMO) systems have shown great potential in supporting wireless networks to achieve higher data rates and wider connections due to their excellent spectrum utilization and spatial multiplexing capabilities. At the same time, drones, as mobile access points (APs), play an increasingly important role in communication and mission execution due to their high mobility and flexibility. However, when UAVs perform complex tasks, their endurance becomes a key factor restricting their widespread application due to energy constraints. Although there are temporary strategies such as battery replacement or intelligent battery capacity management, these methods are costly and may not be practical. To this end, researchers have begun to explore the use of wireless power transmission (WPT) technology to provide energy replenishment for UAVs to extend their endurance and ability to perform tasks.

[0003] During the flight of UAV, its trajectory optimization is crucial to improve communication quality and energy efficiency. Traditional methods usually have difficulty in handling complex nonlinear optimization problems, such as the joint optimization of UAV's flight trajectory, charging and discharging time slots, and beamforming configuration. Therefore, the introduction of deep reinforcement learning (DRL) technology, especially deep Q network (DQN), provides a new way to solve these problems. By combining the powerful fitting ability of deep neural networks and the decision-making ability of Q learning algorithms, DQN can effectively handle high-dimensional state space and action space problems and learn the optimal strategy in complex environments. However, the current DQN algorithm still has the limitation of low efficiency in handling complex multi-objective optimization problems, especially in the combination of UAV trajectory optimization and energy management, there is a lack of effective solutions. Summary of the invention

[0004] Purpose of the invention: To provide an optimization method for energy transmission and reception and trajectory design of UAVs in a de-cellular mMIMO system, so as to solve the limitation that the current DQN algorithm in the prior art still has the low efficiency when dealing with complex multi-objective optimization problems, especially in the combination of UAV trajectory optimization and energy management, and lacks an effective solution.

[0005] Technical solution:

[0006] An optimization method for UAV energy transmission and reception and trajectory design in a decellularized mMIMO system comprises the following steps:

[0007] S1: Establish a CF-mMIMO system model, wherein the CF-mMIMO system model includes L APs, an energy-collecting UAV, and K mobile users, wherein the UAV has the functions of communication, reception, and storage power supply;

[0008] S2: defining an optimization objective function to maximize the throughput of the system, ensuring the fairness of communication between different mobile users while ensuring the endurance of the UVA, the optimization objective function includes a joint function of a fairness index and a total user communication rate, the fairness index is used to measure the fairness of the UAV in providing communication services to different mobile users, and the total communication rate of all mobile users during the entire flight time of the UAV is considered to ensure the overall communication efficiency of the system;

[0009] S3: Select a communication-assisted DQN algorithm model to solve the optimization problem and obtain the optimal UAV energy transmission and reception and trajectory allocation solution. In the DQN algorithm model, the DQN includes two neural networks. By randomly initializing the weights and biases of the network and using experience replay to optimize the learning process, the neural network agent UAV runs through states, selects states and obtains rewards in each time slot, changes new positions and adjusts strategies to meet constraints.

[0010] In a further embodiment, in S1, the L APs include aerial APs and ground APs. In the CF-mMIMO system model, the UAV is equipped with a transceiver and an energy receiver. The UAV serves as the aerial AP to provide communication services for the mobile device, receives the radio frequency energy broadcast by the ground AP, and stores the radio frequency energy in a rechargeable battery. The rechargeable battery ensures the endurance of the UAV during the service process.

[0011] In a further embodiment, in S2, based on the constraint of the UAV's endurance, the energy collection time proportional factor of the UAV in each time slot, the proportion of the communication time between the UAV and each of the mobile users, the trajectory of the UAV and the collaborative beamforming vector of the UAV for data transmission are jointly optimized.

[0012] In a further embodiment, in S2, the fairness index used to measure the fairness of the UAV in providing communication services to different mobile users is: And considering that the total communication rate of all the mobile users during the entire flight time T of the UAV is To ensure the overall communication efficiency of the system, the optimization objective function of the present invention aims to simultaneously improve the system throughput and the communication fairness between mobile users, and achieve overall optimization of system performance under the constraint of ensuring the UAV service power;

[0013] Therefore, the optimization objective function F is defined as the fairness index Total communication rate with users The joint function is expressed as follows:

[0014]

[0015] Where μ is f n Scaling factor in the objective function.

[0016] In a further embodiment, in S2, by jointly optimizing the resource allocation variables, the UAV energy collection time scaling factor in each time slot is τ e [n], the proportion of the communication time between the UAV and each mobile user is τ uk [n], the trajectory of the UAV is q u [n], the UAV’s coordinated beamforming vector for data transmission is w uk [n], so the optimization objective problem is expressed as follows:

[0017]

[0018] Among them, V max is the maximum rate of the UAV, q0 represents the initial starting point of the UAV, constraint C1 indicates that the cumulative energy consumption of the UAV in each time slot cannot exceed the existing energy of the UAV, C2 indicates that the sum of the energy collection time and the communication time of the UAV supporting the mobile user in each time slot cannot exceed the duration of a time slot, and C3 and C4 give the maximum flight rate V max The constraints on the UAV position change are as follows.

[0019] In a further embodiment, in S3, the two neural networks are a policy network and a target network, respectively, the policy network is used to select actions according to the environment and serve as the agent's decision, and the target network is used to calculate the ratio of policy values ​​before and after policy update.

[0020] In a further embodiment, in S3, the UAV is regarded as an agent interacting with the system environment in a series of discrete time instances. At each time slot n, the UAV observes the state s u (n) take action u (n), for the current state s u (n) Output action a u (n), the feedback from the environment is regarded as a reward function, that is, the reward r u (n), and so on, in the n+1th time slot, it transfers to the new state s u (n+1), the corresponding elements are set as follows:

[0021] State space s u (n): In time slot n, the policy network collects state information, which includes the location of the UAV, the UAV battery energy, and the current time slot, that is, the description of the state space of the UAV is expressed as follows:

[0022] s u [n]=(q u [n]; e u [n]; n)

[0023] Among them, e u [n] represents the energy state of the UAV at time slot n, which is determined by the residual energy of the UAV at time slot n-1;

[0024] Action Space a u (n): For the agent UAV, the action includes the adjusted flight trajectory, the transmit beamforming vector, the charging time ratio, and the time ratio for serving the user. That is, the action space description of the UAV is expressed as follows:

[0025] a n =(q u [n]; w u1 [n],...,w uK [n]; τ e [n]; τ u1 [n],...,τ uK [n])

[0026] Reward u (n): In the nth time slot, the reward function is calculated as:

[0027]

[0028] In a further embodiment, the UAV utilizes a playback memory D having a certain capacity. u To store <s u (t), a u (t), r u (t), s u (t+1)>samples.

[0029] The beneficial effects of the present invention are as follows: the throughput of the system is maximized while satisfying the UAV energy limitation and the charging and discharging time slot constraints, the endurance of the UAV is improved, and the communication fairness of mobile users is ensured. Specifically:

[0030] (1) Aiming at the energy transmission and reception and trajectory optimization problems of UAVs in CF-mMIMO systems, this paper proposes an optimization method for UAV energy transmission and reception and trajectory design based on WPT, which maximizes the total rate of all mobile users during the entire flight time of the UAV and ensures the communication performance of each user through reasonable resource allocation and scheduling strategies;

[0031] (2) Comprehensively considering the energy consumption, collection and communication needs of UAVs, the communication fairness of mobile users and the endurance of UAVs are improved by jointly optimizing the UAV flight trajectory, charging and discharging time slots and beamforming configuration. By jointly optimizing the energy collection time proportional factor, communication time proportion, flight trajectory and coordinated beamforming vector of data transmission of UAVs in each time slot, the present invention can efficiently manage the energy use of UAVs. When the UAV is low on power, the ground access point starts the WPT mechanism to provide the required energy for the UAV flying at low altitude, which significantly improves the endurance of UAVs. The communication fairness between different mobile users is considered, and reasonable resource allocation and scheduling strategies are used to ensure that each user can obtain relatively fair service quality.

[0032] (3) In order to reduce sample complexity and improve learning efficiency, the present invention adopts a communication-assisted DQN algorithm, which enables the UAV to act as an intelligent agent, realizes intelligent optimization of the UAV flight trajectory and energy management, makes autonomous decisions in a dynamically changing environment, and flexibly adjusts strategies. This intelligent decision-making mechanism enables the system to adaptively respond to various complex scenarios, improves the robustness and flexibility of the system, thereby maximizing its endurance and improving the overall communication performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is an overall flow chart of the optimization method of UAV energy transmission and reception and trajectory design based on the DQN algorithm of the present invention.

[0034] Figure 2 This is a model diagram of the UAV-assisted CF-mMIMO system based on WPT of the present invention. DETAILED DESCRIPTION

[0035] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features well known in the art are not described.

[0036] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0037] Reference Figure 1-2, is a method for optimizing energy transmission and reception and trajectory design of a UAV in a decellularized mMIMO system disclosed in the present invention, comprising the following steps:

[0038] S1: Establish a CF-mMIMO system model, which includes L APs, an energy-collecting UAV and K mobile users. The UAV has the functions of communication, reception and storage of power supply; the L APs include aerial APs and ground APs. In the CF-mMIMO system model, the UAV is equipped with a transceiver and an energy receiver. The UAV serves as the aerial AP to provide communication services for the mobile users, receives the radio frequency energy broadcast by the ground AP, and stores the radio frequency energy in a rechargeable battery. The rechargeable battery ensures the endurance of the UAV during the service process. The construction of the CF-mMIMO system model is based on the UAV assistance of WPT, represented by u, and a group of K single-antenna mobile users, represented by k∈{1,2,…,K}, where each AP is equipped with M antennas and the UAV is equipped with A antennas.

[0039] Assuming that all users are moving on the ground at an altitude of 0, the UAV starts from a fixed position and moves at a fixed height H∈[H min ,H max ]flight.

[0040] Mobile users use the time division multiple access protocol to communicate with UAVs. The flight time of the UAV is represented by T, which is discretized into N time slots, so the width of each time slot is τ = T / N. The duration of the time slot is much smaller than the coherence time of the channel, so the position of the UAV and the user and the channel gain in the time slot remain almost unchanged. Assuming that N is large enough, the position of the UAV and the user and the channel gain in the time slot τ remain almost unchanged.

[0041] In each time slot, each mobile user has the possibility to communicate with the UAV. The specific communication situation is determined by the horizontal distance between the UAV and the user. uk [n] is the proportion of the time that the kth user communicates with the UAV in time slot n, τ e [n] represents the proportion of time that the UAV performs wireless energy transmission from the AP in time slot n.

[0042] Using two-dimensional Euclidean coordinates, let the fixed coordinates of the lth AP point be denoted as q l =(x l ,y l ), the two-dimensional coordinates of the kth user in the nth time slot are q k [n]=(x k [n],y k [n]), the plane coordinates of the UAV q u [n]=(xu [n],y u [n]).

[0043] At time slot n, the distance between the UAV and the kth user is denoted as D uk [n]:

[0044]

[0045] Similarly, at time slot n, the distance between the UAV and the lth ground AP is expressed as D ul [n]:

[0046]

[0047] Considering the influence of buildings and obstacles, the channel gain vector between the UAV and the kth user is expressed as:

[0048]

[0049] Where a0 represents the channel power at a reference distance of 1m, β uk represents the path loss index between the UAV and the kth user, g uk [n] represents the channel gain caused by path loss and multipath fading at time slot n.

[0050] Due to ground reflection and scattering, the data signal transmitted by the UAV to the kth user experiences both line-of-sight (LoS) and non-line-of-sight (NLoS) propagation. The probability of LoS propagation is Affected by the elevation angles of the UAV and the user, the LoS transmission probability between the UAV and the kth user is:

[0051]

[0052] Where a and b are environmental parameters, and the elevation angle between the kth user and the UAV is θ ukn =(180 / π)arctan(H / ||q k [n]-q u [n]|| 2 ).

[0053] Therefore, large-scale fading is modeled as:

[0054]

[0055] Among them, a1 and a2 are the path loss exponents of LoS and NLoS respectively. represents the probability of NLoS link, and

[0056] In addition, the Rician distribution of small-scale fading between the UAV and the k-th user is expressed as:

[0057]

[0058] in, is the Rician factor, is the UAV transmitting array LoS response, specifically expressed as:

[0059]

[0060] Among them, θ kn represents the deviation angle of the transmitted signal to the kth user, Δd represents the antenna spacing of the UAV, and λ is the carrier wavelength.

[0061] NLoS Component is modeled as a complex Gaussian distribution with zero mean and unit variance, i.e.

[0062] S2: Define an optimization objective function to maximize the throughput of the system and ensure the fairness of communication between different mobile users while ensuring the endurance of the UVA. The optimization objective function includes a joint function of a fairness index and a total user communication rate. The fairness index is used to measure the fairness of the UAV in providing communication services to different mobile users. At the same time, the total communication rate of all mobile users during the entire flight time of the UAV is considered to ensure the overall communication efficiency of the system.

[0063] In S2, based on the constraint of the UAV's endurance, the energy collection time proportional factor of the UAV in each time slot, the proportion of the UAV's communication time with each mobile user, the trajectory of the UAV, and the coordinated beamforming vector of the UAV for data transmission are jointly optimized. The fairness index used to measure the fairness of the UAV in providing communication services to different mobile users is: And considering that the total communication rate of all the mobile users during the entire flight time T of the UAV is To ensure the overall communication efficiency of the system, the optimization objective function of the present invention aims to simultaneously improve the system throughput and the communication fairness between mobile users, and achieve overall optimization of system performance under the constraint of ensuring the UAV service power;

[0064] In order to ensure the communication performance of each user, the optimization goal of the present invention is to maximize the total communication rate of all mobile users during the entire flight time of the UAV. At the same time, the fairness of computing bits between different users is guaranteed.

[0065] Since the antenna arrays of AP and UAV are both elevated, their forward transmission channels are dominated by LoS propagation. The channel matrix of the lth AP and UAV in time slot n is It is expressed as follows:

[0066]

[0067] Where a0 represents the channel power when the reference distance is 1 m, and b0 is the path loss index of the channel between the UAV and the lth AP. represents the channel gain matrix between the lth AP and UAV at time slot n, approximately in

[0068] Assume that all ground APs simultaneously use a fixed power p a Send energy signal to UAV, w ul [n] is the beamforming vector sent by the lth AP to the UAV. The energy signal received by the UAV at time slot n is:

[0069]

[0070] Among them, e m [n]~CN(0,I N ) represents the energy signal sent by the lth AP to the UAV.

[0071] However, due to hardware damage and noise, the energy actually received by the UAV at time slot n can be expressed as:

[0072]

[0073] Where ξ represents the hardware damage factor.

[0074] To maximize the system throughput under the constraint of ensuring the endurance of the UAV and ensure the fairness of communication between different users,

[0075] Based on Jain's fairness index, the fairness index is defined as:

[0076]

[0077] To measure the fairness of the cumulative communication rate between users in time slot n. Obviously, the fairness index f in time slot n n The ultimate fairness between users can be measured, f n The larger it is, the more fair it is.

[0078] Therefore, the optimization objective function F is defined as the fairness index Total communication rate with users The joint function is expressed as follows:

[0079]

[0080] Where μ is f n Scaling factor in the objective function.

[0081] Specifically, in S2, by jointly optimizing the resource allocation variables, the UAV energy collection time proportional factor in each time slot is τ e [n], the proportion of the communication time between the UAV and each mobile user is τ uk [n], the trajectory of the UAV is q u [n], the UAV’s coordinated beamforming vector for data transmission is w uk [n], at time slot n, the channel gain vector between the UAV and the kth user is denoted as h uk [n], let The UAV sends a signal s in time slot n k The beamforming vector of . Therefore, at time slot n, the received signal at the kth user is:

[0082]

[0083] in, is the additive white Gaussian noise received at user k, Represents the noise power of the user.

[0084] Therefore, at time slot n, the rate of user k is:

[0085]

[0086] Wherein, B represents bandwidth.

[0087] In order to maximize the system throughput under the constraint of UAV endurance while ensuring communication fairness between different users, the optimization objective problem is expressed as follows:

[0088]

[0089] Among them, V max is the maximum rate of the UAV, q0 represents the initial starting point of the UAV, constraint C1 indicates that the cumulative energy consumption of the UAV in each time slot cannot exceed the existing energy of the UAV, C2 indicates that the sum of the energy collection time and the communication time of the UAV supporting the mobile user in each time slot cannot exceed the duration of a time slot, and C3 and C4 give the maximum flight rate V max The constraints on the UAV position change are as follows.

[0090] In terms of energy consumption, in addition to communication energy consumption, flight energy consumption of UAV cannot be ignored. is a function of the UAV flight distance and is expressed as follows:

[0091]

[0092] Among them, η1 and η2 are the power required to keep the UAV in the air and the power consumed to move per unit distance, respectively.

[0093] Let P 0 Represents the initial energy of the UAV to perform the task. Since the total energy consumption of the UAV at time slot n cannot exceed its remaining energy, the specific total energy consumption constraint of the UAV is expressed as follows:

[0094]

[0095] Among them, P 0 Indicates the initial energy of the UAV to perform the mission.

[0096] S3: Select the communication-assisted DQN algorithm model to solve the optimization problem and obtain the optimal UAV energy transmission and reception and trajectory allocation solution. In the DQN algorithm model, the DQN includes two neural networks. By randomly initializing the weights and biases of the network and using experience replay to optimize the learning process, the neural network agent UAV runs through the state in each time slot, selects the state and obtains the reward, changes the new position and adjusts the strategy to meet the constraints, and finally obtains the optimal resource management and UAV trajectory planning strategy based on the DQN-assisted communication algorithm.

[0097] Specifically, in S3, the two neural networks are a policy network and a target network. The policy network is used to select actions according to the environment and is used as the agent's decision. The target network is used to calculate the ratio of the policy values ​​before and after the policy update. The two networks are updated using the network update rule. For optimization problems, Q-learning is more suitable for problems with limited action states. DQN is an extension of the Q-learning algorithm. It combines deep neural networks with a reinforcement learning framework to solve the problem that the Q-learning algorithm is limited in problems with large or continuous action state spaces. In the DQN algorithm, the UAV acts as an intelligent agent and interacts with the environment.

[0098] In S3, the UAV is regarded as an agent interacting with the system environment in a series of discrete time instances. At each time slot n, the UAV observes the state s u (n) take action u (n), for the current state s u (n) Output action a u (n), the feedback from the environment is regarded as a reward function, that is, the reward r u(n), and so on, in the n+1th time slot, it transfers to the new state s u (n+1), the corresponding elements are set as follows:

[0099] State space s u (n): In time slot n, the policy network collects state information. In the DQN algorithm, the state information includes the position of the UAV, the battery energy of the UAV, and the current time slot. That is, the description of the state space of the UAV is expressed as follows:

[0100] s u [n]=(q u [n]; e u [n]; n)

[0101] Among them, e u [n] represents the energy state of the UAV at time slot n, which is determined by the residual energy of the UAV at time slot n-1;

[0102] Action Space a u (n): For the proxy UAV, the action includes the adjusted flight trajectory, the transmit beamforming vector, the charging time ratio, and the time ratio for serving the user. That is, the action space description of the UAV is expressed as follows:

[0103] a n =(q u [n]; w u1 [n],...,w uK [n]; τ e [n]; τ u1 [n],...,τ uK [n])

[0104] Reward u (n): The goal of the present invention is to maximize the communication rate while ensuring fairness between users. Therefore, in the nth time slot, the reward function is calculated as:

[0105]

[0106] The UAV uses a playback memory D having a certain capacity. u To store <s u (t), a u (t), r u (t), s u(t+1)> samples, UAV uses two deep neural networks as policy and target networks, these networks have the same fully connected layer structure, but they differ in weight and bias parameters. The neural network consists of three parts: input layer, hidden layer and output layer. The hidden layer uses a 4-layer structure with 256, 128, 64 and 32 nodes respectively. In addition, to improve the performance and stability of the neural network, after each layer except the output layer, layer normalization is applied and exited with a probability of 0.2.

[0107] Policy Network in s u (n) Execute a in state u (n) The action value function is obtained as Q is used as the agent's decision, and the output value of the target network is It is used to update the network by calculating the loss function, which is used to compare the output of the policy and target networks. Specifically, the two networks operate separately. In order to select action a(n), at time slot n, the state s(n) is input to the weight In the neural network, a(n) maximizes the current probability with ε The output of or a random choice. After performing action a(n), the agent receives reward r(n) and moves to the next state s(n+1). Then, by minimizing The loss function To train DQN. Considering the robustness to outliers in the neural network noise estimation, the Huber loss is used to minimize the loss, that is, when the loss is small, the error value is the mean square error, and when the loss is large, the loss is the mean absolute error:

[0108]

[0109] Among them, δ represents the loss threshold and y represents the policy network Output of . Target value It is based on the reward r(n) obtained and the predicted reward By weight Target network The policy network continuously updates its weights based on the observed rewards and actions, while the target network's weights are not iteratively updated, but are only updated periodically by copying the policy network's weights after a specified time interval.

[0110] After training and updating the network, the agent can eventually derive the optimal resource management and UAV trajectory planning strategy based on the DQN-assisted communication algorithm.

[0111] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0112] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings; however, the present invention is not limited to the specific details in the above embodiments; within the technical concept of the present invention, various equivalent transformations can be made to the technical scheme of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. An optimization method for UAV energy transmission and reception and trajectory design in a decellularized mMIMO system, characterized by: The steps include: S1: Establish a CF-mMIMO system model, wherein the CF-mMIMO system model includes L APs, an energy-collecting UAV, and K mobile users, wherein the UAV has the functions of communication, reception, and storage power supply; S2: defining an optimization objective function to maximize the throughput of the system, ensuring the fairness of communication between different mobile users while ensuring the endurance of the UAV, the optimization objective function including a joint function of a fairness index and a total user communication rate, the fairness index being used to measure the fairness of the UAV in providing communication services to different mobile users, while considering the total communication rate of all mobile users during the entire flight time of the UAV; S3: Select a communication-assisted DQN algorithm model to solve the optimization problem and obtain the optimal UAV energy transmission and reception and trajectory allocation solution. In the DQN algorithm model, the DQN includes two neural networks. The neural network agent of the UAV runs through states, selects states and obtains rewards in each time slot, and changes new positions and adjusts strategies to meet constraints.

2. The optimization method for energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 1, characterized in that: In S1, the L APs include aerial APs and ground APs. In the CF-mMIMO system model, the UAV is equipped with a transceiver and an energy receiver. The UAV serves as the aerial AP to provide communication services for the mobile device, receives the radio frequency energy broadcast by the ground AP, and stores the radio frequency energy in a rechargeable battery.

3. The optimization method for energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 1, characterized in that: In S2, based on the constraints of the UAV's endurance, the energy collection time proportional factor of the UAV in each time slot, the proportion of the communication time between the UAV and each mobile user, the trajectory of the UAV and the collaborative beamforming vector of the UAV for data transmission are jointly optimized.

4. The method for optimizing energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 3, characterized in that: In S2, the fairness index used to measure the fairness of the UAV in providing communication services to different mobile users is: And considering that the total communication rate of all the mobile users during the entire flight time T of the UAV is To ensure the overall communication efficiency of the system, the optimization objective function of the present invention aims to simultaneously improve the system throughput and the communication fairness between mobile users, and achieve overall optimization of system performance under the constraint of ensuring the UAV service power; Therefore, the optimization objective function F is defined as the fairness index Total communication rate with users The joint function is expressed as follows: Where μ is f n Scaling factor in the objective function.

5. The method for optimizing energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 4, characterized in that: In S2, by jointly optimizing the resource allocation variables, the UAV energy collection time scaling factor in each time slot is τ e [n], the proportion of the communication time between the UAV and each mobile user is τ uk [n], the trajectory of the UAV is q u [n], the UAV’s coordinated beamforming vector for data transmission is w uk [n], so the optimization objective problem is expressed as follows: C4:q u [1]=q0. Among them, V max is the maximum rate of the UAV, q0 represents the initial starting point of the UAV, constraint C1 indicates that the cumulative energy consumption of the UAV in each time slot cannot exceed the existing energy of the UAV, C2 indicates that the sum of the energy collection time and the communication time of the UAV supporting the mobile user in each time slot cannot exceed the duration of a time slot, and C3 and C4 give the maximum flight rate V max The constraints on the UAV position change are as follows.

6. The method for optimizing energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 1, characterized in that: In S3, the two neural networks are a policy network and a target network, respectively. The policy network is used to select actions according to the environment and is used as the agent's decision-making. The target network is used to calculate the ratio of policy values ​​before and after policy update.

7. The method for optimizing energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 5, characterized in that: In S3, the UAV is regarded as an agent interacting with the system environment in a series of discrete time instances. At each time slot n, the UAV observes the state s u (n) take action u (n), for the current state s u (n) Output action a u (n), the feedback from the environment is regarded as a reward function, that is, the reward r u (n), and so on, in the n+1th time slot, it transfers to the new state s u (n+1), the corresponding elements are set as follows: State space s u (n): In time slot n, the policy network collects state information, which includes the location of the UAV, the battery energy of the UAV, and the current time slot, that is, the description of the state space of the UAV is expressed as follows: s u [n]=(q u [n];e u [n];n) Among them, e u [n] represents the energy state of the UAV at time slot n, which is determined by the residual energy of the UAV at time slot n-1; Action Space a u (n): For the agent UAV, the action includes the adjusted flight trajectory, the transmit beamforming vector, the charging time ratio, and the time ratio for serving the user. That is, the action space description of the UAV is expressed as follows: a n =(q u [n];w u1 [n],...,w uK [n];t e [n];t u1 [n],...,t uK [n]) Reward u (n): In the nth time slot, the reward function is calculated as:

8. The method for optimizing energy transmission and reception and trajectory design of drones in a decellularized mMIMO system according to claim 7, characterized in that: The UAV uses a playback memory D having a certain capacity. u To store <s u (t), a u (t), r u (t), s u (t+1)>samples.

Images