Equipment control method and device and ground base station

Abstract

The invention provides an equipment control method and device and a ground base station, and relates to the technical field of low-altitude communication. The method comprises the steps of generating an unmanned aerial vehicle displacement instruction and an equipment transmitting power instruction of a next time slot through a strategy generation model based on first environment sensing state information and first feasibility constraint under a current time slot; determining a downlink channel state from the ground base station to the unmanned aerial vehicle according to a predicted position of the unmanned aerial vehicle in the next time slot indicated by the unmanned aerial vehicle displacement instruction; generating a communication pre-coding matrix under a second feasibility constraint based on the downlink channel state; and controlling the flight of the unmanned aerial vehicle, the uplink communication of the ground Internet of Things equipment and the downlink beam of the ground base station based on the unmanned aerial vehicle displacement instruction, the equipment transmission power instruction and the communication precoding matrix, thereby shortening the image capture waiting time of the monitoring point and the data acquisition delay of the Internet of Things equipment, and improving the monitoring accuracy. And physical constraints such as flight safety, energy consumption limitation and transmitting power are met.

Description

Technical Field

[0001] This disclosure relates to the field of low-altitude communication technology, and in particular to a device control method, apparatus and ground base station. Background Technology

[0002] With the rapid development of the low-altitude economy, drones are gradually becoming new types of aerial information nodes, widely used in smart agriculture, emergency communications, and remote area monitoring. By carrying communication and computing modules, drones can collect data from nearby ground-based IoT devices during low-altitude cruises and transmit the processed information back to ground base stations, effectively alleviating the pressure on traditional ground networks in terms of coverage, power consumption, and bandwidth, and building an "air-ground collaborative" intelligent IoT system.

[0003] Against this backdrop, ground base stations not only need to schedule UAVs to perform sensing tasks, but also need to efficiently transmit control commands back to them. As the number of UAVs in the air increases, base stations need to utilize multi-antenna technology to conduct downlink communication with multiple air users simultaneously, which poses new requirements for beamforming design. However, existing beamforming technologies are mainly geared towards static ground users and are difficult to directly apply to UAV swarms with three-dimensional dynamic characteristics. Therefore, for low-altitude, dense UAV scenarios, there is an urgent need for a device control method that can collaboratively optimize the UAV flight trajectory and the base station's downlink beam to support efficient and reliable integrated air-to-ground communication and mission execution. Summary of the Invention

[0004] This disclosure provides a device control method, apparatus, and ground base station.

[0005] In a first aspect, this disclosure provides a device control method applied to a ground base station, wherein the ground base station communicates with at least one unmanned aerial vehicle (UAV), the at least one UAV being configured to acquire images of multiple monitoring points and collect environmental perception data from multiple ground-based Internet of Things (IoT) devices, and the total service period for image acquisition and environmental perception data collection is divided into multiple consecutive time slots. The method includes: generating a UAV displacement command and a device transmit power command for the next time slot based on a first service perception state information and a first feasibility constraint in the current time slot of the multiple consecutive time slots, using a strategy generation model; determining the downlink channel state from the ground base station to the at least one UAV based on the predicted position of the at least one UAV in the next time slot indicated by the UAV displacement command; generating a communication precoding matrix for maximizing the total data rate from the ground base station to the at least one UAV based on the downlink channel state of the at least one UAV under a second feasibility constraint; and controlling the flight of the at least one UAV, the uplink communication of the multiple ground-based IoT devices, and the downlink beam of the ground base station in the next time slot based on the UAV displacement command, the device transmit power command, and the communication precoding matrix.

[0006] Secondly, this disclosure provides a device control apparatus applied to a ground base station, the ground base station communicating with at least one unmanned aerial vehicle (UAV), the at least one UAV being configured to acquire images of multiple monitoring points and collect environmental perception data from multiple ground IoT devices, the total service period for image acquisition and environmental perception data collection being divided into multiple consecutive time slots, the apparatus comprising: a first generation module, configured to generate UAV displacement commands and device transmit power commands for the next time slot based on a first service perception state information and a first feasibility constraint in the current time slot of the multiple consecutive time slots through a strategy generation model; a determination module, configured to determine the downlink channel state from the ground base station to the at least one UAV based on the predicted position of the at least one UAV in the next time slot indicated by the UAV displacement command; a second generation module, configured to generate a communication precoding matrix for maximizing the total data rate from the ground base station to the at least one UAV based on the downlink channel state of the at least one UAV under a second feasibility constraint; and a control module, configured to control the flight of the at least one UAV, the uplink communication of the multiple ground IoT devices, and the downlink beam of the ground base station in the next time slot based on the UAV displacement command, the device transmit power command, and the communication precoding matrix.

[0007] Thirdly, this disclosure provides a ground base station, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the device control method as described in the first aspect of this disclosure.

[0008] Fourthly, this disclosure provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the device control method as described in the first aspect of this disclosure.

[0009] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the device control method as described in the first aspect of this disclosure.

[0010] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects: In each time slot, a strategy-generated model, combined with the first service perception state information and the first feasibility constraint under the current time slot, synchronously outputs UAV displacement commands and ground IoT device transmit power commands, thereby achieving joint dynamic optimization of perception task scheduling and uplink communication resources. Furthermore, based on the predicted UAV's three-dimensional position in the next time slot, the air-to-ground channel state is estimated in real time, and a precoding matrix maximizing the total downlink data rate is constructed under the second feasibility constraint. This enables multi-antenna ground base stations to adaptively adjust beam direction and power allocation for high-speed moving and densely distributed air users, effectively overcoming the inapplicability of traditional beamforming methods for static ground users in low-altitude scenarios, and significantly improving the throughput, reliability, and anti-interference capability of control command feedback. Finally, by implementing closed-loop control of UAV flight trajectory, ground device uplink transmission, and base station downlink beam in time slots, not only is the image capture waiting time of monitoring points and the data acquisition delay of IoT devices significantly shortened, but physical constraints such as flight safety, energy consumption limits, and transmit power are also strictly met, thus providing an efficient, robust, and scalable air-to-ground integrated control framework for applications in low-altitude dense UAV scenarios. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0012] Figure 1 This is a schematic flowchart of the device control method shown in the first embodiment of this disclosure; Figure 2 This is a schematic flowchart of the device control method shown in the second embodiment of this disclosure; Figure 3 This is a schematic flowchart of the device control method shown in the third embodiment of this disclosure; Figure 4 This is a schematic flowchart of the device control method shown in the fourth embodiment of this disclosure; Figure 5 This is a schematic diagram of the initial scene shown in an embodiment of the present disclosure; Figure 6 This is a schematic diagram illustrating the relationship between the reward function and the algorithm algebra, as shown in the embodiments of this disclosure; Figure 7 This is a schematic diagram illustrating the relationship between average successful data collection and the number of algorithm iterations in an embodiment of this disclosure; Figure 8 This is a schematic diagram illustrating the relationship between image capture waiting time and algorithm iteration count in an embodiment of this disclosure; Figure 9 This is a schematic diagram of a drone trajectory shown in an embodiment of this disclosure; Figure 10This is another schematic diagram of a drone trajectory shown in an embodiment of this disclosure; Figure 11 This is a schematic diagram illustrating the relationship between the average total rate and noise variance at all time points shown in the embodiments of this disclosure; Figure 12 This is a schematic diagram illustrating the variation of average total rate with system power as shown in the embodiments of this disclosure; Figure 13 This is a schematic diagram of the structure of the device control apparatus shown in the fifth embodiment of this disclosure; Figure 14 This is a schematic diagram of the structure of a ground base station according to an exemplary embodiment.

[0013] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0014] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0015] In related technologies, a base station equipped with a uniform planar array was used to achieve communication between two UAVs. This research constructed a joint optimization framework with the goal of maximizing the system's achievable communication rate, and significantly improved spectrum utilization efficiency through intelligent resource allocation. Some studies have further proposed waveform design schemes for multi-base station cooperation, enabling the system to maintain reliable communication with authorized UAVs.

[0016] However, the relevant technologies still have certain limitations. They mainly focus on drone trajectory optimization aimed at improving the communication performance indicators of ground base stations, but neglect the most critical task-oriented drone trajectory planning needs in low-altitude economic applications.

[0017] The technical solutions of this disclosure and how they solve any of the aforementioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this disclosure will now be described with reference to the accompanying drawings.

[0018] First, combine Figure 1The present disclosure provides an exemplary description of the device control method provided in the embodiments.

[0019] Figure 1 This is a schematic flowchart of the device control method shown in the first embodiment of this disclosure.

[0020] It should be noted that the device control method provided in this disclosure can be used with a ground base station, which communicates with at least one UAV. The at least one UAV is configured to acquire images of multiple monitoring points and collect environmental perception data from multiple ground IoT devices. The total service period for performing image acquisition and environmental perception data collection is divided into multiple consecutive time slots.

[0021] like Figure 1 As shown, the device control method includes the following steps: Step 101: Based on the first service perception state information and the first feasibility constraint in the current time slot among multiple consecutive time slots, generate the UAV displacement command and equipment transmission power command for the next time slot through the strategy generation model.

[0022] It is important to understand that, in order to achieve efficient perception in low-altitude economic scenarios, this disclosure divides the entire service process into multiple continuous time slots. The ground base station acts as a centralized decision-making center to uniformly dispatch one or more drones. In each time slot, the drone flies to the monitoring point to complete image capture and simultaneously approaches the ground IoT device to receive environmental perception data.

[0023] It should be noted that monitoring points can be configured according to actual application scenarios. Monitoring points may include, but are not limited to, farmland monitoring points, forest farm monitoring points, fishery monitoring points, pasture monitoring points, orchard monitoring points, water conservancy and soil monitoring points, etc. Correspondingly, environmental perception data is collected and uploaded by ground-based IoT devices, which may include, but are not limited to, crop perception data, forest farm perception data, fishery perception data, livestock perception data, soil and meteorological perception data, and infrastructure status perception data. Crop perception data may include, for example, images of pest and disease spots, fruit quantity and maturity, etc. Forest farm perception data may include, for example, smoke concentration, distribution of dead trees, temperature and humidity, etc. Fishery perception data may include, for example, water pH value, turbidity, water temperature and images of floating objects on the water surface, etc. Livestock perception data may include, for example, livestock body temperature, activity trajectory, etc. Soil and meteorological perception data may include, for example, soil moisture content, rainfall, light intensity, atmospheric pressure and carbon dioxide concentration, etc. Infrastructure status data may include, for example, the opening and closing status of irrigation valves, fence integrity, power supply equipment voltage, etc.

[0024] To achieve intelligent joint decision-making regarding UAV flight trajectories and ground equipment communication behavior, one possible approach is to have the ground base station invoke a pre-trained policy generation model in each time slot. This model takes the first service perception state information of the current time slot as input and, combined with the first feasibility constraint, outputs control commands for the next moment, including UAV displacement commands and equipment transmission power commands. The UAV displacement command specifies the displacement of each UAV in three-dimensional space, and the equipment transmission power command allocates appropriate uplink transmission power to each ground IoT device. The first service perception state information includes: the three-dimensional position coordinates of at least one UAV in the current time slot, the remaining energy or energy consumption status of at least one UAV, the data acquisition waiting time of multiple ground IoT devices, the image capture waiting time of multiple monitoring points, the uplink or downlink channel gain between the ground base station and at least one UAV, the communication link status from multiple ground IoT devices to their corresponding UAVs, the environmental status or historical perception data confidence level of multiple monitoring points, and the identification information of ground IoT devices or monitoring points that have completed service in the current time slot. The first feasibility constraint includes: flight range, maximum displacement, power limit, and energy consumption limit.

[0025] Step 102: Determine the downlink channel state from the ground base station to at least one UAV based on the predicted position of at least one UAV in the next time slot indicated by the UAV displacement command. Since the drones will move to a new location in the next time slot, the ground base station needs to know their three-dimensional coordinates in advance, and then estimate the wireless propagation environment between the base station and each drone. For example, based on the predicted location and the geometric relationship between the base station's antenna array, the path loss (considering distance, carrier frequency and large-scale fading parameters) is calculated, and the downlink channel state from the ground base station to at least one drone is generated by combining the Ricean fading model (including line-of-sight and scattering components).

[0026] Step 103: Based on the downlink channel state of at least one UAV, under the second feasibility constraint, generate a communication precoding matrix to maximize the total data rate from the ground base station to at least one UAV.

[0027] To achieve efficient concurrent transmission of control commands in dense UAV scenarios, as a possible approach, the ground base station solves an optimization problem using the obtained downlink channel matrix: under the premise of satisfying the second type of feasibility constraints (including positive semidefiniteness of the precoding matrix and equal power allocation of antennas), the optimal beamforming (precoding) matrix (i.e., the communication precoding matrix) is designed to maximize the total downlink data rate received by all UAVs. Thus, by using spatial multiplexing and interference management in dense airspace, control decision results (such as trajectory updates and task allocation) can be efficiently broadcast or directed to multiple UAVs, solving the problem that traditional ground communication schemes are difficult to adapt to three-dimensional dynamic aerial users.

[0028] Step 104: Based on the UAV displacement command, device transmit power command and communication precoding matrix, control the flight of at least one UAV, the uplink communication of multiple ground IoT devices and the downlink beam of the ground base station in the next time slot.

[0029] At the start of the next time slot, the UAV adjusts its flight path based on the UAV displacement command to approach the target monitoring point or IoT device; the ground equipment uploads sensing data (uplink communication) according to the allocated transmission power; at the same time, the ground base station transmits downlink beams based on the communication precoding matrix to send new control information back to the UAV, forming a closed-loop control system.

[0030] As one possible implementation, the ground base station generates UAV displacement commands, device transmit power commands, and a communication precoding matrix to directly control the flight trajectory of at least one UAV, the uplink communication of multiple ground IoT devices, and the downlink beam of the ground base station in the next time slot. For example, the UAV displacement command is directly sent to the UAV's flight control system, driving the UAV to adjust its position, altitude, and heading in real time, thereby accurately flying to the preset monitoring point. The ground base station can send the device transmit power command to each ground IoT device and control the ground IoT device to use the specified power in the device transmit power command for uplink data transmission in the next time slot. The communication precoding matrix is ​​directly loaded into the multi-antenna RF front-end of the ground base station to configure the phase and amplitude weights of the downlink signal in real time, forming a high-gain directional beam for the UAV.

[0031] As another possible implementation, the ground base station indirectly controls the flight trajectory of at least one UAV, the uplink communication of multiple ground IoT devices, and the downlink beam of the ground base station in the next time slot by generating UAV displacement commands, device transmit power commands, and communication precoding matrices. For example, the ground base station guides ground IoT devices to autonomously adjust their uplink transmission power based on the transmit power commands while meeting their own energy consumption constraints by issuing transmit power commands. At the same time, the ground base station sends UAV displacement commands to the UAV, and the UAV flies to the new monitoring position according to the received displacement commands, thereby dynamically changing the wireless propagation environment between air and ground and indirectly optimizing the uplink quality. In addition, the ground base station preprocesses the base station's transmitted signals through the communication precoding matrix to make them more accurately point to the UAV. Thus, the ground base station can effectively improve the transmission efficiency of downlink signals, reduce signal interference, and enhance the reliability and stability of communication.

[0032] The device control method of this disclosure, by generating a model through a strategy in each time slot, and combining the first service perception state information and the first feasibility constraint under the current time slot, synchronously outputs the UAV displacement command and the ground IoT device transmit power command, thereby realizing the joint dynamic optimization of perception task scheduling and uplink communication resources. Furthermore, based on the predicted three-dimensional position of the UAV in the next time slot, the air-to-ground channel state is estimated in real time, and a communication precoding matrix that maximizes the total downlink data rate is constructed under the second feasibility constraint. This enables multi-antenna ground base stations to adaptively adjust beam direction and power allocation for high-speed moving and spatially densely distributed air users, effectively overcoming the inapplicability of traditional beamforming methods for static ground users in low-altitude scenarios, and significantly improving the throughput, reliability, and anti-interference capability of control command feedback. Finally, by implementing closed-loop control of the UAV flight trajectory, ground device uplink transmission, and base station downlink beam in time slots, not only is the image capture waiting time of monitoring points and the data acquisition delay of IoT devices significantly shortened, but physical constraints such as flight safety, energy consumption limits, and transmit power are also strictly met, thereby providing an efficient, robust, and scalable air-to-ground integrated control framework for applications in low-altitude dense UAV scenarios.

[0033] To clearly illustrate how the downlink channel state from the ground base station to at least one UAV is determined based on the predicted position of at least one UAV in the next time slot as indicated by the UAV displacement command in the above embodiments, this disclosure proposes another device control method.

[0034] Figure 2 This is a schematic flowchart of the device control method shown in the second embodiment of this disclosure.

[0035] like Figure 2 As shown, the device control method may include the following steps: Step 201: Based on the first environmental perception state information and the first feasibility constraint in the current time slot among multiple consecutive time slots, generate the UAV displacement command and equipment transmission power command for the next time slot through the strategy generation model.

[0036] Step 202: Based on the UAV displacement command, determine the predicted three-dimensional coordinates of at least one UAV in the next time slot.

[0037] To improve the real-time performance and reliability of control command transmission, as a possible approach, the UAV displacement command generated in the current time slot (i.e., the flight increment along the (x), (y), and (z) directions) is combined with the current three-dimensional position of the UAV to calculate the predicted position of each UAV in the next time slot.

[0038] Step 203: Determine the path loss between at least one UAV and the ground base station based on the predicted three-dimensional coordinates and the antenna array position of the ground base station.

[0039] It is important to understand that in wireless communication systems, path loss is a key indicator that measures the degree of energy attenuation of a signal as it travels from the transmitter (such as a ground base station) to the receiver (such as a drone). Its magnitude directly affects communication quality and data transmission rate.

[0040] To accurately assess the propagation attenuation of wireless signals in low-altitude environments, one possible approach is to utilize the predicted drone location and base station antenna location, calculate the distance between them, and estimate the path loss of the signal during propagation.

[0041] As an example, the distance between at least one UAV and the ground base station is determined based on the predicted three-dimensional coordinates and the antenna array position of the ground base station; based on the distance corresponding to at least one UAV, the carrier frequency used for communication between at least one UAV and the ground base station, and the setting of large-scale fading parameters, the path loss between at least one UAV and the ground base station is calculated; wherein, the large-scale fading parameters include the path loss exponent and the shadow fading standard deviation.

[0042] In other words, based on the predicted three-dimensional coordinates of the UAV and the location of the ground base station antenna array, the Euclidean distance between each UAV and the base station is first calculated. Then, combining the carrier frequency used by the system and preset large-scale fading parameters (including path loss exponent and shadow fading standard deviation), the path loss of each link is calculated using a standard wireless propagation model (such as the logarithmic distance path loss model). This modeling method not only considers the free space attenuation caused by distance, but also introduces environment-related statistical characteristics (such as the shadow effect caused by obstacle occlusion), making the path loss estimation closer to the real low-altitude communication scenario.

[0043] Step 204: Generate the complex baseband channel vector of the downlink from the ground base station to the at least one UAV based on the path loss between the at least one UAV and the ground base station.

[0044] After determining the path loss between each UAV and the ground base station, this path loss can be used as a fundamental parameter for signal attenuation. Combined with the characteristics of the low-altitude communication environment, a downlink complex baseband channel vector from the ground base station to each UAV can be generated. It should be noted that, due to the typically strong line-of-sight propagation path between UAVs and base stations in low-altitude scenarios, the channel modeling employs a Ricean fading model, representing the channel vector as a weighted combination of a deterministic line-of-sight component and a random scattering component. The direction of the line-of-sight component is determined by the relative position between the UAV and the base station antenna array, and the spatial phase relationship is reflected by the array steering vector; the scattering component is used to simulate weak multipath effects in the environment; and the path loss is used as an overall amplitude scaling factor, acting on both components. Ultimately, each UAV corresponds to a complex vector containing amplitude, phase, and spatial characteristics—the complex baseband channel vector—which fully describes the wireless channel response from the base station's multiple antennas to that UAV.

[0045] Step 205: Determine the downlink channel state from the ground base station to at least one UAV based on the complex baseband channel vector of the downlink from the ground base station to at least one UAV.

[0046] As one possible implementation, the ground base station can determine the complete downlink channel state by integrating the downlink complex baseband channel vectors corresponding to each UAV. It should be noted that each complex baseband channel vector reflects the wireless channel characteristics from the base station's multi-antenna array to a single UAV, including signal amplitude attenuation, phase changes, and spatial direction information during propagation. These vectors, when combined, form a unified channel matrix that comprehensively describes the communication link status between the base station and all UAVs.

[0047] Step 206: Based on the downlink channel state of at least one UAV, under the second feasibility constraint, generate a communication precoding matrix to maximize the total data rate from the ground base station to at least one UAV.

[0048] Step 207: Based on the UAV displacement command, device transmit power command and communication precoding matrix, control the flight of at least one UAV, the uplink communication of multiple ground IoT devices and the downlink beam of the ground base station in the next time slot.

[0049] The explanations of steps 201 and 206 to 207 can be found in the relevant descriptions in any embodiment of this disclosure, and will not be repeated here.

[0050] In summary, by pre-determining the predicted three-dimensional coordinates of the UAV in the next time slot based on the UAV's displacement command, the spatial distance between each UAV and the base station can be accurately calculated by combining the position of the ground base station antenna array. Based on this spatial distance, a path loss conforming to low-altitude propagation characteristics can be derived. Furthermore, this path loss is used to construct a complex baseband channel vector containing the line-of-sight dominant component and random scattering components, which truly reflects the channel amplitude and phase characteristics of airborne mobile users. Finally, by integrating the complex baseband channel vectors of all UAVs, complete downlink channel state information is formed, providing high-precision and real-time channel basis for multi-user beamforming.

[0051] To clearly illustrate how the strategy generation model is invoked in the above embodiments to generate the UAV displacement command and device transmit power command for the next time slot based on the first environmental perception state information and the first feasibility constraint in the current time slot among multiple consecutive time slots, this disclosure proposes another device control method.

[0052] Figure 3 This is a schematic flowchart of the device control method shown in the third embodiment of this disclosure.

[0053] like Figure 3 As shown, the device control method may include the following steps: Step 301: Input the first service-aware state information into the policy generation model to obtain the candidate action distribution output by the policy generation model based on the first service-aware state information.

[0054] The candidate action distribution includes the displacement probability distribution of at least one UAV in three-dimensional space and the transmission power probability distribution of multiple ground IoT devices.

[0055] To achieve adaptive decision-making in complex dynamic environments, one possible approach is to input the first service-aware state, which includes multi-dimensional information such as UAV position, energy state, mission latency, and channel conditions, into the policy generation model. The model then outputs a probability distribution of joint actions, whereby the distribution simultaneously encompasses the three-dimensional displacement possibilities of all UAVs and the transmit power selection of all ground IoT devices.

[0056] The policy generation model is obtained by performing multiple rounds of training in a multi-time-slot simulation environment. The i-th round of training includes: (1) Based on the current policy network, control decisions are executed in the simulation environment, and trajectory data including state sequence, action sequence and reward sequence are collected; wherein, the state sequence includes the second service perception state information under each time slot, the action sequence includes the UAV displacement command and equipment transmission power command output under each time slot, and the reward sequence includes the reward value under each time slot. The reward value is generated based on the data acquisition waiting time of multiple ground IoT devices under the corresponding time slot, the image capture waiting time of multiple monitoring points and the degree of violation of feasibility constraints; To automatically generate high-quality interactive data in a simulation environment that closely resembles real-world scenarios, avoiding reliance on manual annotation or prior strategies, a possible implementation approach is to use a strategy generation model comprising a strategy network and a value network. In the (i)th training round, the current strategy network is used to autonomously run in a multi-time-slot simulation environment, sequentially generating control actions for each time slot (including UAV 3D displacement and ground equipment transmission power), and recording the resulting complete interactive trajectory data. The trajectory data includes: second service perception state (such as UAV position, energy, latency, channel gain, etc. in the simulation environment), actions (displacement and power commands), and rewards. The rewards consist of the negative values ​​of the average successful data acquisition latency and the average successful image capture latency, plus constraint violation penalties. The rewards are directly related to the core task objective (reducing latency) and system feasibility, guiding the strategy towards efficiency, safety, and feasibility.

[0057] (2) Use the current value network to estimate the value of each state in the trajectory data and calculate the advantage measure of the corresponding action; To achieve accurate assessment of the direction of strategy improvement, one possible approach is to introduce a value network to predict the expected long-term reward (V(s_t)) for each state (s_t) in the trajectory, and combine this with the actual accumulated reward to calculate the advantage (i.e., advantage metric) of each action relative to the average performance. This advantage metric reflects the advantage of taking a specific action (such as a displacement + power combination) over randomly selecting an action.

[0058] (3) Update the parameters of the current policy network based on the advantage metric, and update the parameters of the current value network based on the error between the actual cumulative reward corresponding to each state and the predicted value of the current value network, so as to obtain the policy network and value network obtained in the i-th round of training.

[0059] When (i = 1), the current policy network and the value network are initialized neural networks; when (i > 1), the current policy network and the current value network are the networks obtained after training in the (i-1)th round, and the actual cumulative reward is the accumulated value of the discounted reward obtained from the corresponding time slot to the end of the simulation.

[0060] To achieve co-optimization of the policy and value functions, one possible approach is to employ a policy gradient method (such as Proximal Policy Optimization, PPO). On one hand, the current policy network parameters are adjusted based on the dominance metric, increasing the probability of high-dominance actions being selected in the future. On the other hand, by minimizing the mean squared error (TD error or Monte Carlo error) between the actual cumulative reward and the value network's predicted value, the current value network's ability to estimate long-term state rewards is continuously calibrated. After this round of optimization, an updated policy and value network is obtained for the next round of training. It should be noted that the actual cumulative reward is the sum of all rewards obtained from the time slot corresponding to the state until the end of the simulation task (episode), i.e., a weighted sum of the instantaneous reward values ​​of each future time slot using a discount factor, serving as the true label for supervised learning of the value network. For example, the total service period is divided into 4 time slots (i.e., episode length (N=4)), and the discount factor... The instantaneous reward values ​​for each time slot are (r_1 = -2), (r_2 = -1), (r_3 = -3), and (r_4 = +5). Therefore, the actual cumulative reward for time slot 1 (state (s_1)) is... .

[0061] It should be noted that, in order to achieve multi-round iterative learning and performance improvement, the embodiments of this disclosure adopt an incremental training mechanism, that is, the first round of training uses a randomly initialized neural network as the starting point; from the second round onwards, each round inherits the policy and value network obtained from the previous round of training as the initial model, thereby achieving continuous accumulation of knowledge.

[0062] Step 302: Based on the first feasibility constraint, sample the UAV displacement command and equipment transmission power command for the next time slot from the candidate action distribution.

[0063] Among them, the UAV displacement command and the equipment transmission power command in the next time slot satisfy the first feasibility constraint.

[0064] Furthermore, after obtaining the distribution of candidate actions, as a possible implementation method, the sampling results are validated or projected to correct the legality based on the first feasibility constraints (including the UAV flight range, maximum displacement limit, upper and lower limits of equipment power, and UAV energy consumption safety threshold, etc.) to ensure that the final output displacement command and power command are physically executable and meet the system safety requirements.

[0065] It should be noted that the first feasibility constraint includes at least one of the following: UAV position constraint, flight displacement constraint, transmission power constraint, and energy consumption constraint.

[0066] Among them, the UAV position constraint is used to indicate that the position of at least one UAV in each time slot is within a preset three-dimensional feasible area; the flight displacement constraint is used to indicate that the displacement of at least one UAV in each direction between adjacent time slots does not exceed the maximum allowable displacement; the transmission power constraint is used to indicate that the transmission power of multiple ground IoT devices in each time slot is between a first set transmission power and a second set transmission power, wherein the first set transmission power is less than the second set transmission power; and the energy consumption constraint is used to indicate that the remaining energy of at least one UAV in each time slot is not less than the minimum safety threshold and does not exceed the set energy capacity.

[0067] Step 303: Determine the downlink channel state from the ground base station to at least one UAV based on the predicted position of at least one UAV in the next time slot indicated by the UAV displacement command.

[0068] Step 304: Based on the downlink channel state of at least one UAV, under the second feasibility constraint, generate a communication precoding matrix to maximize the total data rate from the ground base station to at least one UAV.

[0069] Step 305: Based on the UAV displacement command, device transmit power command and communication precoding matrix, control the flight of at least one UAV, the uplink communication of multiple ground IoT devices and the downlink beam of the ground base station in the next time slot.

[0070] The explanation of steps 303 to 305 can be found in the relevant description in any embodiment of this disclosure, and will not be repeated here.

[0071] In summary, by inputting the first service perception state, which includes multi-dimensional information such as UAV location, remaining energy, mission waiting time, and channel status, into the policy generation model, the model can output a structured candidate action distribution based on the current global system situation. This distribution simultaneously covers the displacement probability of each UAV in three-dimensional space and the transmission power probability of each ground IoT device. Furthermore, by introducing a first feasibility constraint into this probability distribution, it is ensured that the generated displacement and power commands always meet the requirements of physical realizability and system security.

[0072] To clearly illustrate how, based on the downlink channel state of at least one UAV, a communication precoding matrix for maximizing the total data rate from the ground base station to at least one UAV is generated under a second feasibility constraint in the above embodiments, this disclosure proposes another device control method.

[0073] Figure 4 This is a schematic flowchart of the device control method shown in the fourth embodiment of this disclosure.

[0074] like Figure 4 As shown, the device control method may include the following steps: Step 401: Based on the first environmental perception state information and the first feasibility constraint in the current time slot among the multiple consecutive time slots, generate the UAV displacement command and equipment transmission power command for the next time slot through the strategy generation model.

[0075] Step 402: Determine the downlink channel state from the ground base station to at least one UAV based on the predicted position of at least one UAV in the next time slot indicated by the UAV displacement command.

[0076] Step 403: Based on the downlink channel state, construct the downlink channel matrix from the ground base station to at least one UAV.

[0077] To ensure accuracy in concurrent downlink communication for UAVs, one possible approach is to stack the complex baseband channel vectors of the downlink channel states for each UAV row-wise to construct a complete downlink channel matrix. This downlink channel matrix characterizes the spatial channel response from multiple antennas of the base station to all airborne users and may include key information such as amplitude, phase, and inter-user channel correlation.

[0078] Step 404: Under the second feasibility constraint, solve the objective function to maximize the total downlink data rate of at least one UAV to obtain the communication precoding matrix.

[0079] The second feasibility constraint includes at least one of the following: a positive definiteness constraint on the precoding matrix and an antenna power allocation constraint; wherein, the positive definiteness constraint on the precoding matrix is ​​used to indicate that the communication precoding matrix used by the ground base station in each time slot is a positive semi-definite matrix; the antenna power allocation constraint is used to indicate that each diagonal element of the communication precoding matrix is ​​equal to the power of a single antenna after the total transmit power of the ground base station is evenly distributed to the multiple transmit antennas of the ground base station.

[0080] To achieve efficient control command feedback for dense low-altitude UAV swarms under limited launch resources, as a possible approach, an optimization problem aimed at maximizing the total downlink data rate is solved, thereby generating the optimal communication precoding matrix, while satisfying actual hardware and communication specifications.

[0081] It should be noted that, in order to ensure that the designed precoding matrix is ​​physically realizable and numerically stable, a positive semidefinite constraint and an antenna power allocation constraint are introduced. That is, the downlink communication precoding matrix is ​​required to be a positive semidefinite matrix, and the antenna power allocation constraint is used to indicate that each diagonal element of the communication precoding matrix is ​​equal to the power of a single antenna after the total transmit power of the ground base station is evenly distributed to the multiple transmit antennas of the ground base station, that is, the total power is evenly distributed to each antenna.

[0082] Step 405: Based on the UAV displacement command, device transmit power command and communication precoding matrix, control the flight of at least one UAV, the uplink communication of multiple ground IoT devices and the downlink beam of the ground base station in the next time slot.

[0083] The explanations of steps 401 to 402 and 405 can be found in the relevant descriptions in any embodiment of this disclosure, and will not be repeated here.

[0084] In summary, by constructing a complete downlink channel matrix from the base station to the UAV based on the downlink channel state, the system can comprehensively perceive the channel characteristics and spatial relationships of airborne users. Furthermore, under the strict constraints of the second feasibility constraint (including the positive semidefiniteness of the precoding matrix and equal power allocation of the antenna), the optimization problem aimed at maximizing the total downlink data rate is solved, generating a physically realizable and hardware-friendly communication precoding matrix. As a result, not only is high throughput and low interference ensured when the UAV receives control commands concurrently, but the engineering constraints of the actual base station radio frequency link are also taken into account, significantly improving the communication efficiency, system stability, and practical deployment feasibility of air-to-ground control in low-altitude dense scenarios.

[0085] To illustrate the above embodiments more clearly, examples are given below.

[0086] For example, a ground base station B is deployed within a low-altitude economic service area (A). This base station is equipped with a uniform linear array (ULA) of M antennas. There are a total of K UAVs in the system, denoted as KUAV. It is responsible for performing two types of collaborative sensing tasks: (1) From I ground IoT devices Collect crop sensing data; (2) For P monitoring points in the farmland To take images; The entire service period T is divided into N consecutive time slots, denoted as , The ground base station, acting as the global information center, serves as a centralized decision-making unit, generating control commands and transmitting them to all drones via downlink communication waveforms. The location of base station B... Ground-based IoT devices Location Ground-based IoT devices are deployed on the ground, therefore Monitoring points Location drones In the time slot Location .

[0087] This paper takes a typical UAV (Unmanned Aerial Vehicle) task for environmental information detection in the low-altitude economy as an example for modeling: the UAV swarm needs to collect data from IoT devices and simultaneously capture images of specific monitoring points. To achieve efficient and equitable data collection and image capture, this invention proposes two task-oriented performance metrics: one is the average successful data acquisition latency. The other is the average successful image capture wait time. Its mathematical definition is:

[0088]

[0089] in, and They represent time slots respectively. Internal equipment Complete data collection and monitoring points Waiting time to complete image capture and The update rules are as follows

[0090] If data collection is successful ,otherwise ;

[0091] If image capture is successful ,otherwise ; Among them, the associated indicator variable and They represent medium equipment or monitoring point Is it in drones? Within the communication / shooting range (range radius is respectively) and ), aggregate variables and Used to determine whether the device / monitoring point is covered by any drone, indicating successful data acquisition. The following conditions must be met simultaneously:

[0092] in, Indicates the actual transmission time. The maximum allowed threshold is used to simplify the successful image capture state to: This means that as long as the monitoring point is within the shooting range, it is considered a success; Internet of Things (IoT) devices In time slots ( The uplink transmission rate of ) is:

[0093] Where k represents the drone index serving the device. Indicates system bandwidth. The power of additive white Gaussian noise, Characterization equipment exist Transmit power of the time slot. Channel coefficients. Indicates Internet of Things (IoT) devices To drones Channel gain between ,in, and These represent the gain coefficients of the transmitting and receiving antennas, respectively. The signal wavelength is [wavelength]. It should be noted that, to eliminate some inter-user interference, the system employs orthogonal frequency division multiple access (OFDMA) technology to coordinate communication between IoT devices within the same UAV service range. Furthermore, when associated indicator variables... At that time, the image captures state variables. Automatically set to 1, meaning that as long as the nearest drone is located at the monitoring point Within the effective capture range, the image of the monitoring point is considered successfully acquired.

[0094] Define the downlink communication precoding matrix from the base station to the UAV as follows: The channel between the base station and the drone adopts the Ricean fading model, and the channel matrix... row k It consists of a deterministic line-of-sight component and a random scattering component, wherein,

[0095] Where k is the Rice factor, For reference range channel gain, For array guide vector, Where M represents the number of array elements and d represents the spacing between array elements. Indicates the angle of incidence of the signal. Indicates wavelength; drones In time slots ( The SINR of ) is , in, The total downlink data rate of the system is: .

[0096] The ultimate optimization goal is to minimize task latency while maximizing downlink data rate, involving three core variables: Collection of drone trajectories: The three-dimensional position coordinates of each drone in all time slots were recorded, along with the set of transmission power of IoT devices. The transmission power configuration of each IoT node under different time slots is defined, and the communication precoder set is defined. The communication beamforming matrix, which includes each time slot of the base station, is mathematically described as follows: Objective function:

[0097] Constraints:

[0098]

[0099]

[0100]

[0101]

[0102]

[0103] The optimization variables include the drone trajectory, the transmission power of IoT devices, and communication precoding. Constraints mainly involve the drone's operating range, travel distance, device power limits, energy consumption, precoding matrix characteristics, and communication quality requirements. Furthermore, , and The maximum flight range of the UAV along the x-axis, y-axis, and z-axis in each time slot was defined respectively. and The minimum and maximum transmit power ranges for IoT devices are defined. Indicates drone In the time slot The actual energy consumption, its maximum allowable value is determined by... These parameters correspond to six system constraints: C1 ensures that the drone operates within a specified airspace, C2 limits the maximum displacement of the drone in a single time slot, C3 regulates the power transmission range of IoT devices, C4 controls the upper limit of the drone's energy consumption, C5 requires the precoding matrix to satisfy the positive semi-definite characteristic, and C6 specifies the equal power allocation principle for each antenna element of the base station.

[0104] Because this problem is highly nonconvex and the variables grow over time, this disclosure decouples it into two sub-problems: I. Subproblem of UAV trajectory and power optimization Constraints: C1–C4; This subproblem is modeled as a Markov decision-making (policy generation model) process, with triples. The definition is as follows: (1) Intelligent agent and environment: The centralized intelligent agent is the ground base station, and the environment is the smart agriculture scenario; (2) State space set S: in At any given moment, the state space set ; (3) Action space set A: in Moment, Action Space Set ;in, express exist Flight distance along the s-axis; (4) Reward R, in Moments, rewards ,in, and Here, r represents the weights and bias factors, and r is the penalty term. If the constraints in the subproblem are not satisfied, r takes a sufficiently large value; otherwise, r = 0.

[0105] Solve using the Proximal Policy Optimization (PPO) algorithm: (1) Initialize the policy network (output action distribution) and the value network (evaluate state value); (2) Collect multiple trajectory data (status, action, reward); (3) Calculate the advantage function (to measure the quality of the action); (4) Update the policy network and limit the update step size to ensure stability; (5) Update the value network to reduce prediction error; (6) Iterate until convergence.

[0106] II. Downlink Precoding Optimization Subproblem The objective function can be expressed as:

[0107] Constraints: C5-C6 Using a known trajectory Constructing the channel matrix The optimal solution is obtained by using the zero-forcing (ZF) or water-filling algorithm to maximize the total rate while satisfying the semi-definite and equal power constraints.

[0108] For example, zero-forced precoding can be used first to eliminate interference between multiple users. In this step, a precoding matrix is ​​searched. Make ,in I is the channel matrix, and I is the identity matrix. Ideally, this will eliminate all multi-user interference. In practical applications, an error metric, such as the Frobenius norm, is typically calculated to evaluate the goodness of this approximation. In specific implementations, the ZF precoding matrix... This is obtained by solving the following linear equation:

[0109] Secondly, a water-filling method is used for power allocation. This method, based on channel state information, allocates more power to users with better signal-to-noise ratios (SNR), thereby maximizing the system's total throughput. In its implementation, the equivalent channel is first... Singular value decomposition is performed, and then the optimal power distribution is calculated using the water-filling method. Specifically, the following equation is solved to find a parameter. :

[0110] in, It is noise power. It is the square of the singular value of the i-th channel. Then, the optimal power allocation is calculated as:

[0111] Finally, this power allocation matrix is ​​incorporated into the precoding matrix to obtain the final precoding matrix. ,in It is a diagonal matrix whose diagonal elements represent the optimal power allocation.

[0112] The system capacity using the water injection method is calculated using the following formula:

[0113] in, This refers to the power allocation for the i-th user.

[0114] To verify the effectiveness of the algorithm, a detection scenario of drones collecting information in a low-altitude economy is assumed, where there are I=18 IoT devices (ground IoT devices) and P=9 detection points distributed in the service area. K=3 drones are used to collect data from the IoT devices and capture images of the detection points. A central base station located at (0m, 0m) forms a waveform to transmit decisions to the drones. To comply with flight regulations and avoid collisions, the altitude of the three drones is fixed at... , ,and Other parameter settings are listed in the table below:

[0115] Consider the initial scenario, such as Figure 5As shown in Figure 6, the reward function of the UAV trajectory planning subproblem is optimized using a near-end policy optimization algorithm. The cumulative reward function increases with the number of iterations, as shown in Figure 6. This indicates that the algorithm is effective. Specifically, the relationship between the two metrics (average successful data collection / image capture wait time) and the number of iterations in the UAV trajectory planning subproblem is shown in Figure 6. Figure 7 and Figure 8 As shown, both metrics—average successful data collection / image capture wait time—decrease with increasing iteration count, demonstrating the algorithm's effectiveness. The smaller metric indicates more efficient data collection / image capture, and it also reflects fairness between IoT devices and probe points.

[0116] To further illustrate the effectiveness of the algorithm, the trajectory of the UAV after 10,000 generations of the near-end policy optimization algorithm is shown in the figure below. Figure 9 and Figure 10 As shown, by Figure 9 and Figure 10 It can be seen that the drones flew as far as possible across all IoT and detection points to provide data collection / image capture services. Furthermore, the drones covered almost all IoT and detection points, demonstrating the fairness of the optimization. In addition, the drones tended to fly within their designated areas of responsibility, which helps to provide continuous service.

[0117] To demonstrate the effectiveness of the zero-breaking and water-filling algorithms, the relationship between the average total rate of the target at all time points and the noise variance is optimized as follows: Figure 11 As shown in the figure, the average total rate increases as the noise variance decreases. This is because the zero-breaking algorithm performs well under high signal-to-noise ratios. It can be seen that even when the noise is relatively low, the average total rate can still achieve a relatively good value, demonstrating the effectiveness of the algorithm.

[0118] Furthermore, to further demonstrate the effectiveness of the algorithm, the average total rate varies with system power as follows: Figure 12 As shown, by Figure 12 As can be seen, the average total data rate increases with increasing transmit power. This is because with increased transmit power, the power of the focused beam formed by the base station can be higher, thus achieving better communication performance.

[0119] To implement the above embodiments, this disclosure also proposes a device control apparatus.

[0120] Figure 13This is a schematic diagram of the device control apparatus shown in the fifth embodiment of this disclosure. It should be noted that this device control apparatus is applied to a ground base station, which communicates with at least one unmanned aerial vehicle (UAV). The at least one UAV is configured to acquire images of multiple monitoring points and collect environmental perception data from multiple ground-based IoT devices. The total service period for performing image acquisition and environmental perception data collection is divided into multiple consecutive time slots.

[0121] like Figure 13 As shown, the device control unit 1300 includes: a first generation module 1310, a determination module 1320, a second generation module 1330, and a control module 1340.

[0122] The first generation module 1310 is used to generate UAV displacement commands and device transmit power commands for the next time slot based on the first environmental perception state information and the first feasibility constraint in the current time slot among multiple consecutive time slots, through a strategy generation model; the determination module 1320 is used to determine the downlink channel state from the ground base station to at least one UAV based on the predicted position of at least one UAV in the next time slot indicated by the UAV displacement command; the second generation module 1330 is used to generate a communication precoding matrix to maximize the total data rate from the ground base station to at least one UAV based on the downlink channel state of at least one UAV under the second feasibility constraint; and the control module 1340 is used to control the flight of at least one UAV, the uplink communication of multiple ground IoT devices, and the downlink beam of the ground base station in the next time slot based on the UAV displacement command, the device transmit power command, and the communication precoding matrix.

[0123] As one possible implementation, the determining module 1320 is used to determine the predicted three-dimensional coordinates of at least one UAV in the next time slot based on the UAV displacement command; determine the path loss between at least one UAV and the ground base station based on the predicted three-dimensional coordinates and the antenna array position of the ground base station; generate the complex baseband channel vector of the downlink from the ground base station to at least one UAV based on the path loss between at least one UAV and the ground base station; and determine the downlink channel state from the ground base station to at least one UAV based on the complex baseband channel vector of the downlink from the ground base station to at least one UAV.

[0124] As one possible implementation, the determination module 1320 is used to determine the distance between at least one UAV and the ground base station based on the predicted three-dimensional coordinates and the antenna array position of the ground base station; based on multiple distances, the carrier frequency for communication between at least one UAV and the ground base station, and setting large-scale fading parameters, the path loss between at least one UAV and the ground base station is calculated; wherein, the large-scale fading parameters include the path loss index and the shadow fading standard deviation.

[0125] As one possible implementation, the first generation module 1310 is used to input the first service perception state information into the policy generation model to obtain the candidate action distribution output by the policy generation model based on the first environmental perception state information; wherein, the candidate action distribution includes the displacement probability distribution of at least one UAV in three-dimensional space and the transmission power probability distribution of multiple ground IoT devices; according to the first feasibility constraint, the UAV displacement command and device transmission power command for the next time slot are sampled from the candidate action distribution; wherein, the UAV displacement command and device transmission power command for the next time slot satisfy the first feasibility constraint.

[0126] As one possible implementation, the first feasibility constraint includes at least one of the following: UAV position constraint; wherein the UAV position constraint is used to indicate that the position of at least one UAV in each time slot is within a preset three-dimensional feasible area; flight displacement constraint; wherein the flight displacement constraint is used to indicate that the displacement of at least one UAV in each direction between adjacent time slots does not exceed the maximum allowable displacement; transmission power constraint; wherein the transmission power constraint is used to indicate that the transmission power of multiple ground IoT devices in each time slot is between a first set transmission power and a second set transmission power; the first set transmission power is less than the second set transmission power; energy consumption constraint; wherein the energy consumption constraint is used to indicate that the remaining energy of at least one UAV in each time slot is not less than a minimum safety threshold and does not exceed a set energy capacity.

[0127] As one possible implementation, the second generation module 1330 is used to construct a downlink channel matrix from the ground base station to at least one UAV based on the downlink channel state; under a second feasibility constraint, it solves an objective function to maximize the total downlink data rate of all UAVs to obtain a communication precoding matrix; wherein the second feasibility constraint includes at least one of the following: a positive definiteness constraint for the precoding matrix; wherein the positive definiteness constraint for the precoding matrix is ​​used to indicate that the communication precoding matrix used by the ground base station in each time slot is a positive semi-definite matrix; an antenna power allocation constraint; wherein the antenna power allocation constraint is used to indicate that each diagonal element of the communication precoding matrix is ​​equal to the single antenna power after the total transmit power of the ground base station is evenly distributed to the multiple transmit antennas of the ground base station.

[0128] As one possible implementation, the policy generation model includes a policy network and a value network. The policy generation model is trained using a training module, which performs multiple rounds of training on the policy generation model in a multi-time-slot simulation environment. The i-th round of training includes: executing control decisions in the simulation environment based on the current policy network, and collecting trajectory data including state sequences, action sequences, and reward sequences; wherein the state sequence includes second service perception state information in each time slot, the action sequence includes UAV displacement commands and device transmit power commands output in each time slot, and the reward sequence includes reward values ​​in each time slot, which are generated based on the data acquisition waiting time of multiple ground IoT devices, the image capture waiting time of multiple monitoring points, and the degree of feasibility constraint violation in the corresponding time slot; estimating the value of each state in the trajectory data using the current value network and calculating the advantage metric of the corresponding action; updating the parameters of the current policy network based on the advantage metric, and updating the parameters of the current value network based on the error between the actual cumulative reward corresponding to each state and the predicted value of the current value network, thus obtaining the policy network and value network obtained in the i-th round of training; wherein, when (i = 1), the current policy network and value network are initialized neural networks; when (i > 1), ... When the current policy network and the current value network are obtained after training in round (i-1), the actual cumulative reward is the accumulated value of the discounted reward obtained from the time slot to which the corresponding state belongs until the end of the simulation.

[0129] As one possible implementation, the first service-aware status information includes at least one of the following: the three-dimensional location coordinates of at least one UAV in the current time slot; the remaining energy or energy consumption status of at least one UAV; the data acquisition waiting time of multiple ground IoT devices; the image capture waiting time of multiple farmland monitoring points; the uplink or downlink channel gain from the ground base station to at least one UAV; the communication link status from multiple ground IoT devices to the corresponding UAV; the environmental status or historical sensing data confidence level of multiple monitoring points; and the identification information of ground IoT devices or farmland monitoring points that have completed service in the current time slot.

[0130] The device control apparatus of this disclosure generates a model through a strategy in each time slot, combining the first environmental perception state information and the first feasibility constraint under the current time slot, and synchronously outputs the UAV displacement command and the ground IoT device transmit power command, thereby realizing the joint dynamic optimization of perception task scheduling and uplink communication resources. Furthermore, based on the predicted three-dimensional position of the UAV in the next time slot, the air-to-ground channel state is estimated in real time, and a precoding matrix that maximizes the total downlink data rate is constructed under the second feasibility constraint. This enables multi-antenna ground base stations to adaptively adjust beam direction and power allocation for high-speed moving and spatially densely distributed air users, effectively overcoming the inapplicability of traditional beamforming methods for static ground users in low-altitude scenarios, and significantly improving the throughput, reliability, and anti-interference capability of control command feedback. Finally, by implementing closed-loop control of the UAV flight trajectory, ground device uplink transmission, and base station downlink beam in time slots, not only is the image capture waiting time of farmland monitoring points and the data acquisition delay of IoT devices significantly shortened, but physical constraints such as flight safety, energy consumption limits, and transmit power are also strictly met, thereby providing an efficient, robust, and scalable air-to-ground integrated control framework for applications in low-altitude dense UAV scenarios.

[0131] To implement the above embodiments, this disclosure also proposes a ground base station, including: a processor and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the device control method as described in this disclosure.

[0132] To implement the above embodiments, this disclosure also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the device control method described in this disclosure.

[0133] To implement the above embodiments, this disclosure also provides a computer program product.

[0134] The computer program product includes a computer program that, when executed by a processor, implements the device control method described in the embodiments of this disclosure.

[0135] Figure 14 This is a structural block diagram of a ground base station according to an exemplary embodiment. Figure 14 The ground base station shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments disclosed herein.

[0136] like Figure 14As shown, the electronic device 1400 includes a processor 1401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1402 or a program loaded from memory 1406 into random access memory (RAM) 1403. The RAM 1403 also stores various programs and data required for the operation of the controller 1400. The processor 1401, ROM 1402, and RAM 1403 are interconnected via a bus 1404. An input / output (I / O) interface 1405 is also connected to the bus 1404.

[0137] The following components are connected to I / O interface 1405: memory 1406 including hard disks, etc.; and communication section 1407 including network interface cards such as local area network (LAN) cards, modems, etc., communication section 1407 performs communication processing via a network such as the Internet; drive 1408 is also connected to I / O interface 1405 as needed.

[0138] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1407. When the computer program is executed by processor 1401, it performs the functions defined in the methods of this disclosure.

[0139] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory including instructions, which can be executed by the processor 1401 of the ground base station 1400 to complete the above method. Optionally, the storage medium can be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0140] In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can transmit, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wireline, optical fiber, RF, etc., or any suitable combination thereof.

[0141] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0142] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A device control method characterized by, Applied to a ground base station, the ground base station communicates with at least one unmanned aerial vehicle (UAV), the at least one UAV being configured to acquire images of multiple monitoring points and collect environmental perception data from multiple ground-based IoT devices. The total service period for performing image acquisition and environmental perception data collection is divided into multiple consecutive time slots, the method comprising: Based on the first service perception state information and the first feasibility constraint in the current time slot among the multiple consecutive time slots, the UAV displacement command and equipment transmission power command for the next time slot are generated through the strategy generation model. Based on the predicted position of the at least one UAV in the next time slot indicated by the UAV displacement command, the downlink channel state from the ground base station to the at least one UAV is determined; Based on the downlink channel state of the at least one UAV, under the second feasibility constraint, a communication precoding matrix is ​​generated to maximize the total data rate from the ground base station to the at least one UAV. Based on the UAV displacement command, the device transmit power command, and the communication precoding matrix, the flight of at least one UAV, the uplink communication of the plurality of ground IoT devices, and the downlink beam of the ground base station are controlled in the next time slot.

2. The method of claim 1, wherein, The step of determining the downlink channel state from the ground base station to the at least one UAV based on the predicted position of the at least one UAV in the next time slot as indicated by the UAV displacement command includes: Based on the UAV displacement command, the predicted three-dimensional coordinates of the at least one UAV in the next time slot are determined; Based on the predicted three-dimensional coordinates and the antenna array position of the ground base station, the path loss between the at least one UAV and the ground base station is determined; Based on the path loss between the at least one UAV and the ground base station, a complex baseband channel vector for the downlink from the ground base station to the at least one UAV is generated; The downlink channel state from the ground base station to the at least one UAV is determined based on the complex baseband channel vector of the downlink from the ground base station to the at least one UAV.

3. The method of claim 2, wherein, Determining the path loss between the at least one UAV and the ground base station based on the predicted three-dimensional coordinates and the antenna array position of the ground base station includes: Based on the predicted three-dimensional coordinates and the antenna array position of the ground base station, the distance between the at least one UAV and the ground base station is determined; Based on the distance corresponding to the at least one UAV, the carrier frequency used for communication between the at least one UAV and the ground base station, and the set large-scale fading parameters, the path loss between the at least one UAV and the ground base station is calculated; wherein, the large-scale fading parameters include the path loss index and the shadow fading standard deviation.

4. The method of claim 1, wherein, The process of generating UAV displacement commands and device transmit power commands for the next time slot based on the first service perception state information and the first feasibility constraint in the current time slot among the multiple consecutive time slots includes: The first service-aware state information is input into the policy generation model to obtain the candidate action distribution output by the policy generation model based on the first service-aware state information; wherein, the candidate action distribution includes the displacement probability distribution of the at least one UAV in three-dimensional space and the transmission power probability distribution of the plurality of ground IoT devices; Based on the first feasibility constraint, the UAV displacement command and device transmission power command for the next time slot are sampled from the candidate action distribution; wherein the UAV displacement command and device transmission power command for the next time slot satisfy the first feasibility constraint.

5. The method according to claim 4, characterized in that, The first feasibility constraint includes at least one of the following: UAV position constraints; wherein, the UAV position constraints are used to indicate that the position of the at least one UAV in each time slot is within a preset three-dimensional feasible area; Flight displacement constraint; wherein the flight displacement constraint is used to indicate that the displacement of the at least one UAV in each direction between adjacent time slots does not exceed the maximum allowable displacement; Transmission power constraint; wherein the transmission power constraint is used to indicate that the transmission power of the plurality of ground IoT devices in each time slot is between a first set transmission power and a second set transmission power; the first set transmission power is less than the second set transmission power; Energy consumption constraint; wherein the energy consumption constraint is used to indicate that the remaining energy of the at least one UAV in each of the time slots is not less than a minimum safety threshold and does not exceed a set energy capacity.

6. The method according to claim 1, characterized in that, Based on the downlink channel state of the at least one UAV, and under a second feasibility constraint, a communication precoding matrix is ​​generated to maximize the total data rate from the ground base station to the at least one UAV, including: Based on the downlink channel state, a downlink channel matrix from the ground base station to the at least one UAV is constructed; Under the second feasibility constraint, an objective function is solved to maximize the total downlink data rate of the at least one UAV, in order to obtain the communication precoding matrix; The second feasibility constraint includes at least one of the following: Positive definiteness constraint of precoding matrix; wherein, the positive definiteness constraint of precoding matrix is ​​used to indicate that the communication precoding matrix used by the ground base station in each time slot is a positive semi-definite matrix; Antenna power allocation constraint; wherein the antenna power allocation constraint is used to indicate that each diagonal element of the communication precoding matrix is ​​equal to the single antenna power after the total transmit power of the ground base station is evenly distributed to the multiple transmit antennas of the ground base station.

7. The method according to any one of claims 1-6, characterized in that, The policy generation model includes a policy network and a value network; the policy generation model is trained using the following steps: The policy generation model is trained in multiple rounds in a multi-time-slot simulation environment, wherein the i-th round of training includes: Based on the current policy network, control decisions are executed in the simulation environment, and trajectory data including state sequences, action sequences, and reward sequences are collected. The state sequences include second service perception state information in each time slot, the action sequences include UAV displacement commands and device transmit power commands output in each time slot, and the reward sequences include reward values ​​in each time slot. The reward values ​​are generated based on the data acquisition waiting time of multiple ground IoT devices, the image capture waiting time of multiple monitoring points, and the degree of violation of feasibility constraints in the corresponding time slot. The current value network is used to estimate the value of each state in the trajectory data, and the advantage metric of the corresponding action is calculated. The parameters of the current policy network are updated based on the advantage metric, and the parameters of the current value network are updated based on the error between the actual cumulative reward corresponding to each state and the predicted value of the current value network, so as to obtain the policy network and value network obtained in the i-th round of training. Wherein, when (i = 1), the current policy network and the value network are initialized neural networks; when (i > 1), the current policy network and the current value network are networks obtained in the (i-1)th round of training, and the actual cumulative reward is the accumulated value of the discount reward obtained from the time slot to which the corresponding state belongs until the end of the simulation.

8. The method according to any one of claims 1-6, characterized in that, The first service-aware state information includes at least one of the following: The three-dimensional position coordinates of the at least one UAV in the current time slot; The remaining energy or energy consumption status of at least one UAV; Data acquisition waiting time of the multiple ground IoT devices; Image capture waiting time for the multiple monitoring points; The uplink or downlink channel gain between the ground base station and the at least one UAV; The communication link status between the multiple ground IoT devices and the corresponding UAVs; Confidence level of environmental status or historical sensing data of the multiple monitoring points; The identification information of ground IoT devices or monitoring points that have completed service within the current time slot.

9. A device control system, characterized in that, An apparatus applied to a ground base station, the ground base station communicating with at least one unmanned aerial vehicle (UAV), the at least one UAV configured to acquire images of multiple monitoring points and collect environmental perception data from multiple ground-based IoT devices, the total service period for performing image acquisition and environmental perception data collection being divided into multiple consecutive time slots, the apparatus comprising: The first generation module is used to generate the UAV displacement command and the device transmit power command for the next time slot based on the first service perception state information and the first feasibility constraint in the current time slot among the multiple consecutive time slots, through a strategy generation model. The determination module is used to determine the downlink channel state from the ground base station to the at least one UAV based on the predicted position of the at least one UAV in the next time slot indicated by the UAV displacement command. The second generation module is used to generate a communication precoding matrix for maximizing the total data rate from the ground base station to the at least one UAV, based on the downlink channel state of the at least one UAV and under a second feasibility constraint. The control module is used to control the flight of at least one UAV, the uplink communication of the plurality of ground IoT devices, and the downlink beam of the ground base station in the next time slot based on the UAV displacement command, the device transmit power command, and the communication precoding matrix.

10. A ground base station, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-8.

Images