Unmanned aerial vehicle path planning method and device based on radio map reconstruction
By combining deep learning and denoising techniques, radio maps can be reconstructed using sparse sampled data, solving the problem of high-precision map reconstruction for UAVs in complex environments and achieving efficient UAV path planning and communication strategy optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-07
- Publication Date
- 2026-03-31
AI Technical Summary
Traditional radio map reconstruction technology struggles to meet the high-precision map requirements of drones in complex environments, especially in urban areas where signal obstruction and interference significantly impact map reconstruction accuracy and efficiency.
A method based on deep learning and denoising technology is adopted to reconstruct radio maps from sparse sampled data. The method combines the depth image prior (DIP) and the denoising regularization (RED) algorithm, and uses the alternating direction multiplier method (ADMM) for iterative optimization to generate high-precision radio maps. Finally, the deep adversarial dual Q network (D3QN) algorithm is used to plan the flight path of the UAV.
It significantly improves the reconstruction accuracy of radio maps under sparse data conditions, reduces data collection costs and difficulties, and enhances the flight safety and communication efficiency of UAVs in complex environments.
Smart Images
Figure CN121761908A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of wireless communication and machine learning technology, specifically to a method and apparatus for unmanned aerial vehicle (UAV) path planning based on radio map reconstruction. Background Technology
[0002] With the continuous development of 6th Generation (6G) communication technology, integrated air-space-ground networks are gradually becoming a key research direction in the field of communications. This network aims to achieve seamless connectivity between ground, air, and space platforms to support a wide range of application scenarios and highly dynamic communication needs. In this network, unmanned aerial vehicles (UAVs) play a crucial role, and their applications in agricultural monitoring, urban planning, emergency rescue, and rapid deployment of communication networks are becoming increasingly important.
[0003] During missions, unmanned aerial vehicles (UAVs) rely heavily on accurate radio maps. These maps provide essential signal strength and distribution data to optimize flight paths and communication strategies, ensuring efficient and safe operation in complex environments. Traditional radio map reconstruction techniques typically depend on large amounts of dense ground measurement data, which is not only costly but also involves complex and time-consuming processing, making it difficult to adapt to rapidly changing application requirements.
[0004] Specifically, traditional technologies face numerous challenges when dealing with complex environments such as cities. For example, factors such as building obstruction, terrain undulations, and interference from electronic devices in cities can all affect the propagation of wireless signals. These factors combined limit the accuracy and efficiency of radio map reconstruction. Therefore, existing map reconstruction methods are insufficient to meet the high-precision map requirements of UAVs in complex environments. Summary of the Invention
[0005] At least one embodiment of this application provides a method and apparatus for UAV path planning based on radio map reconstruction, which solves the problem that existing map reconstruction methods are difficult to meet the high-precision map requirements of UAVs in complex environments.
[0006] To solve the above-mentioned technical problems, this application is implemented as follows:
[0007] In a first aspect, embodiments of this application provide a method for UAV path planning based on radio map reconstruction, including:
[0008] The radio signal data of multiple sampling points obtained by the UAV in the target area are acquired, wherein the radio signal data of each sampling point includes the three-dimensional coordinates of the sampling point and the received signal strength of at least one network device received by the UAV at the sampling point;
[0009] Based on the radio signal data from the multiple sampling points, sparse sampling data for the multiple sampling points is generated, wherein the sparse sampling data for each sampling point includes the signal-to-interference ratio (SIR) received by the UAV from each network device and the probability of connection interruption between the UAV and each network device;
[0010] Based on the sparse sampling data of the multiple sampling points, a radio map model is established by combining the depth image prior (DIP) and the denoising regularization (RED) algorithm; and the alternating direction multiplier method (ADMM) is applied to iteratively optimize the radio map model to obtain the reconstructed radio map.
[0011] Based on the radio map, plan the flight path of the target drone.
[0012] Optionally, based on the radio signal data of the multiple sampling points, sparse sampling data of the multiple sampling points is generated, including:
[0013] Iterate through each sampling point, and for the current sampling point in the current iteration, perform the following calculation:
[0014] The process involves iterating through each of the at least one network device, wherein, for the currently traversed network device, the signals of other network devices besides the current network device are used as interference signals to calculate the signal-to-interference ratio (SIR) received by the UAV from the current network device; and, based on whether the SIR of the signal from the current network device exceeds a preset threshold, the probability of connection interruption between the UAV and the current network device is determined.
[0015] Optionally, based on the sparse sampling data of the multiple sampling points, a radio map model is established by combining depth image prior (DIP) and denoising regularization (RED) algorithms, including:
[0016] Initialize a deep convolutional network and set random parameters;
[0017] The sparse sampling data from the multiple sampling points is input into the deep convolutional network and iteratively optimized multiple times to obtain the radio map model; wherein: each iteration calculates the reconstruction error based on the current output of the deep convolutional network and the sparse sampling data from the multiple sampling points, wherein the reconstruction error is... It is the sum of the data fidelity term and the regularization term. It is the observation matrix. It is the output of a deep convolutional network. It is sparse sampling data from the multiple sampling points. It is a regularization term. It is a regularization parameter; the parameters of the current deep convolutional network are updated with the goal of minimizing the reconstruction error.
[0018] Optionally, the Alternating Direction Multiplier Method (ADMM) is applied for iterative optimization of the radio map model to obtain a reconstructed radio map, including:
[0019] Based on the sparse sampling data of the multiple sampling points, the observation matrix, the additive white Gaussian noise level, the regularization term and ADMM parameters, the number of iterations, network parameters, image reconstruction variables and Lagrange multipliers are initialized;
[0020] Iterative optimization is performed using an alternating update strategy, which involves updating network parameters using the current image reconstruction variables and Lagrange multipliers; updating image reconstruction variables based on the current network parameters and Lagrange multipliers, combined with denoising regularization; and adjusting the Lagrange multipliers to meet the constraints.
[0021] When the preset iteration optimization termination condition is met, the iteration optimization ends and the reconstructed radio map is obtained.
[0022] Optionally, after obtaining the reconstructed radio map, the method further includes:
[0023] The accuracy and communication coverage efficiency of the radio map are evaluated at preset intervals.
[0024] Based on the accuracy and communication coverage efficiency of the obtained radio map, the ADMM parameters and / or the UAV's flight strategy are adjusted to iteratively update the radio map.
[0025] Optionally, the flight path of the target UAV is planned based on the radio map, including: learning the flight trajectory of the target UAV offline using the Deep Adversarial Double Q Network (D3QN) algorithm based on the radio map.
[0026] Optional, also includes:
[0027] During the flight of the target UAV based on the flight trajectory, a radio map is drawn, and the flight path and communication strategy of the target UAV are adjusted according to the currently drawn radio map.
[0028] Secondly, embodiments of this application provide a UAV path planning device based on radio map reconstruction, comprising:
[0029] The acquisition module is used to acquire radio signal data from multiple sampling points obtained by the UAV in the target area. The radio signal data of each sampling point includes the three-dimensional coordinates of the sampling point and the received signal strength of at least one network device received by the UAV at the sampling point.
[0030] The generation module is used to generate sparse sampling data for the multiple sampling points based on the radio signal data of the multiple sampling points. The sparse sampling data for each sampling point includes the signal-to-interference ratio (SIR) of the UAV to each network device and the probability of connection interruption between the UAV and each network device.
[0031] The reconstruction module is used to establish a radio map model based on the sparse sampling data of the multiple sampling points by combining the depth image prior (DIP) and the denoising regularization (RED) algorithm; and to perform iterative optimization of the radio map model by applying the alternating direction multiplier method (ADMM) to obtain the reconstructed radio map.
[0032] The planning module is used to plan the flight path of the target UAV based on the radio map.
[0033] Thirdly, embodiments of this application provide a UAV path planning device based on radio map reconstruction, comprising: a transceiver, a processor, a memory, and a program or instructions stored in the memory and executable on the processor; when the processor executes the program or instructions, it implements the steps of the method described in any of the first aspects.
[0034] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that, when executed by a processor, implements the steps of the method described in any of the first aspects.
[0035] Fifthly, embodiments of this application provide a computer program product including computer instructions that, when executed by a processor, implement the steps of the method described in any of the first aspects.
[0036] Compared with existing technologies, the UAV path planning method and apparatus based on radio map reconstruction provided in this application, by combining deep learning and denoising techniques, can significantly improve the reconstruction accuracy of radio maps under sparse data conditions, effectively addressing signal interference and data loss issues in complex environments. This application reduces reliance on large-scale data collection; through deep denoising regularization, high-quality map reconstruction can be achieved even with incomplete data, reducing the cost and difficulty of data collection. Furthermore, this application utilizes the Alternating Direction Multiplier Method (ADMM) to optimize the computation process, effectively managing the algorithm's computational resources, accelerating iteration speed, and improving overall computational efficiency. Additionally, this application dynamically adjusts the UAV's flight path and communication strategy based on the reconstructed radio map, enhancing the safety and efficiency of UAV mission execution. Attached Figure Description
[0037] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application;
[0038] Figure 2 This is a flowchart of a UAV path planning method based on radio map reconstruction according to an embodiment of this application;
[0039] Figure 3 This is a schematic diagram of a UAV path planning device based on radio map reconstruction according to an embodiment of this application.
[0040] Figure 4 This is another structural schematic diagram of the UAV path planning device based on radio map reconstruction according to an embodiment of this application. Detailed Implementation
[0041] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, without limiting the number of objects; for example, the first object can be one or more. Furthermore, the term "and / or" in this application describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.
[0042] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc.; an indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on the judgment result.
[0043] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used in the systems and radio technologies mentioned above, as well as in other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th Generation (6G) communication systems.
[0044] Figure 1This diagram illustrates a block diagram of a wireless communication system applicable to embodiments of this application. The wireless communication system includes a drone 11 and a network device 12. The drone 11 can be a terminal with flight and wireless communication capabilities. The network device 12 can include access network equipment, which may also be referred to as a Radio Access Network (RAN) device, a Radio Access Network function, or a Radio Access Network unit. The access network equipment can include a base station, a Wireless Local Area Network (WLAN) access point (AP), or a Wireless Fidelity (WiFi) node, etc. In this context, a base station may be referred to as a Node B (NB), an evolved Node B (eNB), a next-generation Node B (gNB), a New Radio Node B (NR Node B), an access point, a relay base station (RBS), a serving base station (SBS), a base transceiver station (BTS), a radio base station, a radio transceiver, a basic service set (BSS), an extended service set (ESS), a home Node B (HNB), a home evolved Node B, a transmission reception point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved. The base station is not limited to any specific technical terminology. It should be noted that in this application embodiment, only a base station in an NR system is used as an example for introduction, and the specific type of base station is not limited.
[0045] This application provides a method for UAV path planning based on radio map reconstruction. This method reconstructs UAV communication radio maps using Deep Denoising Regularization (RMC-DDR), offering an innovative solution to the shortcomings of existing technologies. By integrating deep learning and denoising techniques, this application can effectively reconstruct high-quality radio maps even with sparse data, significantly improving the accuracy and robustness of map reconstruction. Compared to traditional techniques, this application can achieve accurate mapping and simulation of complex wireless signal environments with far less data than conventional methods, providing strong technical support for UAV path planning and communication strategy optimization. The above-described method of this application enables UAVs to reduce their reliance on actual physical environment testing during mission execution, thereby expanding their application range in complex environments while ensuring flight safety and communication efficiency.
[0046] This application provides a UAV path planning method based on radio map reconstruction, which can be executed by a UAV or other devices with computing capabilities. Please refer to... Figure 2 The method includes:
[0047] Step 21: Obtain radio signal data from multiple sampling points sampled by the UAV in the target area. The radio signal data of each sampling point includes the three-dimensional coordinates of the sampling point and the received signal strength of at least one network device received by the UAV at the sampling point.
[0048] This application embodiment uses a drone to perform sparse sampling in a target area to collect radio signal data. Each sampling point includes the three-dimensional coordinates of the drone at the sampling point and the received signal strength of at least one network device received by the drone at that sampling point. Specifically, the network device can be an access network device, such as a base station. The following embodiments will mainly use a base station (cellular base station) as an example for illustration.
[0049] In some implementations, drones equipped with highly sensitive signal receivers and high-precision GPS positioning systems are used to ensure that the collected radio signal data has high spatiotemporal resolution. Assuming there are C cellular base stations in the target area, using... Let t represent the equivalent channel gain from base station c to the UAV at time t.
[0050] Therefore, the signal power received by the UAV from base station c at time t Represented as:
[0051] (1)
[0052] in This represents the transmit power of base station c, which is usually a fixed value; and These represent the large-scale channel gain and antenna gain of base station c, respectively. Represents the position of the drone at time t; a random variable. This represents small-scale fading. (Used...) This represents the cellular base station that is connected to the drone at time t.
[0053] Step 22: Generate sparse sampling data for the multiple sampling points based on the radio signal data from the multiple sampling points. The sparse sampling data for each sampling point includes the signal-to-interference ratio (SIR) received by the UAV from each network device and the probability of connection interruption between the UAV and each network device.
[0054] This application embodiment generates sparse sampling data for the plurality of sampling points by traversing each sampling point, wherein the following calculation is performed for the current sampling point being traversed:
[0055] The process involves iterating through each of the at least one network device, wherein, for the currently iterated network device, the signals of other network devices besides the current network device are used as interference signals to calculate the signal-to-interference ratio (SIR) of the UAV to the current network device; and, based on whether the SIR of the signal of the current network device exceeds a preset threshold, the probability of connection interruption between the UAV and the current network device is determined.
[0056] Due to the randomness of small-scale fading, at time t, for any UAV location and the base station associated with the UAV, the received signal-to-noise ratio is a random number. Therefore, the relationship between the UAV and network equipment (cellular base station) is also random. The probability of connection interruption is and The function is represented as
[0057] (2)
[0058] in, For the signal-to-interference ratio of the drone, when Less than the preset threshold ,Right now If the signal-to-interference ratio (SIR) of the drone at time t is interrupted, then the connection between the drone and the cellular base station is determined to be interrupted.
[0059] (3)
[0060] Step 23: Based on the sparse sampling data of the multiple sampling points, a radio map model is established by combining the depth image prior (DIP) and the denoising regularization (RED) algorithm; and the alternating direction multiplier method (ADMM) is applied to iteratively optimize the radio map model to obtain the reconstructed radio map.
[0061] In this embodiment, the RMC-DDR algorithm uses random initialization of the neural network as the image prior, utilizes depth denoising regularization to process sparse sampled data, and automatically adjusts network parameters to optimize map reconstruction through iterative learning, independent of external labeled datasets. Specifically, based on the sparse sampled data of the multiple sampling points, a radio map model is established by combining Deep Image Prior (DIP) and Regularization by Denoising (RED) algorithms, including:
[0062] (1) Initialize a deep convolutional network and set random parameters;
[0063] (2) Input the sparse sampling data of the multiple sampling points into the deep convolutional network and perform multiple iterations of optimization to obtain the radio map model; wherein: each iteration calculates the reconstruction error based on the current output of the deep convolutional network and the sparse sampling data of the multiple sampling points, wherein the reconstruction error is:
[0064] (4)
[0065] The reconstruction error is the sum of the data fidelity term and the regularization term. It is the observation matrix. It is the output of a deep convolutional network. It is sparse sampling data from the multiple sampling points. It is a regularization term. It is a regularization parameter; the parameters of the current deep convolutional network are updated with the goal of minimizing the reconstruction error.
[0066] here, Used to quantize and reconstruct the signal With observation data The degree of matching between them; regularization terms By introducing signals Prior knowledge (e.g., the smoothness, sparsity, or specific statistical properties of the signal) is used to stabilize the solution to the inverse problem and suppress the effects of noise; regularization parameters Adjust the trade-off between data fidelity terms and regularization terms.
[0067] In this embodiment, the Alternating Direction Method of Multipliers (ADMM) optimization includes sequentially updating network parameters, image reconstruction variables, and Lagrange multipliers to ensure that equality constraints during the reconstruction process are met. Specifically, ADMM is applied to iteratively optimize the radio map model to obtain the reconstructed radio map, including:
[0068] (1) Based on the sparse sampling data of the multiple sampling points, the observation matrix, the additive white Gaussian noise level, the regularization term and ADMM parameters, initialize the number of iterations, network parameters, image reconstruction variables and Lagrange multipliers.
[0069] Here, the input data includes sparse sampling data (damaged measurement data) from the multiple sampling points. Observation matrix (linear degenerate matrix) Additive white Gaussian noise level Regularization parameters ADMM parameters Initialize the number of iterations. Initialize network parameters Image reconstruction variables Lagrange multipliers .
[0070] (2) Iterative optimization is performed using an alternating update strategy, wherein the network parameters are updated using the current image reconstruction variables and Lagrange multipliers; the image reconstruction variables are updated based on the current network parameters and Lagrange multipliers, combined with denoising regularization; and the Lagrange multipliers are adjusted to meet the constraints.
[0071] Here, in the iterative optimization, network parameter updates include: updating the network parameters by solving the optimization problem using the current image reconstruction variables and Lagrange multipliers. This may include gradient descent or other optimization techniques to reduce reconstruction errors; image reconstruction variable updates include: updating the image reconstruction variables based on the current network parameters and Lagrange multipliers, combined with denoising regularization techniques. The key to this step is improving image quality and reducing noise; Lagrange multiplier updates include: adjusting the Lagrange multipliers to further satisfy equality constraints. This includes adjusting the multipliers to ensure that constraints are properly considered and satisfied during the optimization process.
[0072] (3) When the preset iteration optimization termination condition is met, the iteration optimization ends and the reconstructed radio map is obtained.
[0073] Here, after each iteration, it can be determined whether the termination condition is met. For example, a convergence check can be performed to check whether the algorithm has reached convergence, including minimizing the error and reducing constraint violations. If the preset termination condition is met (such as reaching the maximum number of iterations or the error and constraint violations reaching an acceptable minimum level), the iterative optimization is terminated.
[0074] Step 24: Plan the flight path of the target UAV based on the radio map.
[0075] In this embodiment, the flight path of the target UAV is planned based on the radio map obtained in step 23. For example, based on the radio map, the flight trajectory of the target UAV is learned offline using a Deep Adversarial Double Q Network (D3QN) algorithm. As one implementation, the above process includes:
[0076] (1) Analyze the radio map to identify weak signal coverage points; wherein, the locations with low SIR values in the radio map (such as SIR values below a preset threshold) are weak signal coverage points.
[0077] (2) Establishing the trajectory optimization problem for the UAV. A specific example is provided below. It should be noted that the following example is only an example of a trajectory optimization problem that can be used in this application and is not intended to limit this application.
[0078] (5)
[0079] st ,
[0080] ,
[0081] ,
[0082] ,
[0083] in, Indicates the flight direction of the target drone; This represents the time step for each training round of the target drone; It is a non-negative number representing the number of extra steps the drone is willing to take to avoid communication interruption; This indicates the maximum displacement of the drone in each step. and These represent the initial position and the final destination of the drone, respectively. and These represent the upper and lower boundaries of the drone's flight area, respectively.
[0084] (3) Initialize two deep neural networks. Based on the radio map, iterate through multiple iterations to approximate the action value function and update the network parameters to reduce the error. Specifically, the following loss function can be used:
[0085] (6)
[0086] in, The coefficient is The Q-network is used for action selection; while the loss function contains... The coefficient is The target network is used to evaluate the bootstrap actions within it; , representing the contribution of the UAV to the objective function of the optimization problem at a given time step.
[0087] Through the above steps, this embodiment of the application, by combining deep learning and denoising techniques, can significantly improve the reconstruction accuracy of radio maps under sparse data conditions, effectively addressing signal interference and data loss issues in complex environments. This embodiment reduces reliance on large-scale data collection; through deep denoising regularization, high-quality map reconstruction can be achieved even with incomplete data, reducing the cost and difficulty of data collection. Furthermore, this embodiment utilizes the Alternating Direction Multiplier Method (ADMM) to optimize the computation process, effectively managing the algorithm's computational resources, accelerating iteration speed, and improving overall computational efficiency.
[0088] Considering that radio maps may change over time, in some embodiments, after step 23 above, this application may further evaluate the accuracy and communication coverage efficiency of the radio map at a preset period; and, based on the evaluated accuracy and communication coverage efficiency of the radio map, adjust the ADMM parameters and / or the UAV's flight strategy to iteratively update the radio map. The specific method for iteratively updating the radio map can refer to steps 21-23 above, that is, by repeatedly executing steps 21-23 above, the radio map is iteratively updated.
[0089] Considering that radio maps may change over time and affect the path planning performance of the target UAV, this embodiment can also maintain real-time updates to the radio map by drawing it online. Specifically, during the flight of the target UAV based on the flight trajectory, a radio map is drawn, and the flight path and communication strategy of the target UAV are adjusted according to the currently drawn radio map. This embodiment can have the target UAV draw the radio map online and adjust its flight path and communication strategy, or an external computing device can draw the radio map online and send the updated radio map to the target UAV. The target UAV can then adjust its flight path and communication strategy according to the updated radio map. Alternatively, an external computing device can draw the radio map online, adjust the target UAV's flight path and communication strategy according to the updated radio map, and send the adjusted flight path to the target UAV so that the target UAV can fly according to the adjusted flight path.
[0090] Through the above steps, this application embodiment dynamically adjusts flight and communication strategies, dynamically adjusting the UAV's flight path and communication strategy based on the reconstructed radio map, thereby improving the safety and efficiency of the UAV in performing tasks, especially in urban and other structurally complex environments.
[0091] The various methods of the embodiments of this application have been described above. Apparatus for implementing the above methods will now be provided.
[0092] Please refer to Figure 3 This application also provides a drone path planning device based on radio map reconstruction, comprising:
[0093] The acquisition module 301 is used to acquire radio signal data of multiple sampling points obtained by the UAV in the target area, wherein the radio signal data of each sampling point includes the three-dimensional coordinates of the sampling point and the received signal strength of at least one network device received by the UAV at the sampling point;
[0094] The generation module 302 is used to generate sparse sampling data for the multiple sampling points based on the radio signal data of the multiple sampling points, wherein the sparse sampling data for each sampling point includes the signal-to-interference ratio of the UAV to each network device and the probability of connection interruption between the UAV and each network device;
[0095] The reconstruction module 303 is used to establish a radio map model based on the sparse sampling data of the multiple sampling points by combining the depth image prior (DIP) and the denoising regularization (RED) algorithm; and to perform iterative optimization of the radio map model by applying the alternating direction multiplier method (ADMM) to obtain the reconstructed radio map.
[0096] The planning module 304 is used to plan the flight path of the target UAV based on the radio map.
[0097] Optionally, the acquisition module 301 is further configured to traverse each sampling point, wherein, for the currently traversed sampling point, the following calculations are performed: traversing each of the at least one network device, wherein, for the currently traversed network device, the signals of other network devices besides the current network device are used as interference signals to calculate the signal-to-interference ratio (SIR) received by the UAV from the current network device; and, based on whether the SIR of the signal of the current network device exceeds a preset threshold, determining the probability of connection interruption between the UAV and the current network device.
[0098] Optionally, the reconstruction module 303 is further configured to initialize a deep convolutional network and set random parameters; input the sparse sampling data of the multiple sampling points into the deep convolutional network, perform multiple iterative optimizations, and obtain the radio map model; wherein: each iteration calculates the reconstruction error based on the current output of the deep convolutional network and the sparse sampling data of the multiple sampling points, wherein the reconstruction error is... It is the sum of the data fidelity term and the regularization term. It is the observation matrix. It is the output of a deep convolutional network. It is sparse sampling data from the multiple sampling points. It is a regularization term. It is a regularization parameter; the parameters of the current deep convolutional network are updated with the goal of minimizing the reconstruction error.
[0099] Optionally, the reconstruction module 303 is further configured to initialize the number of iterations, network parameters, image reconstruction variables, and Lagrange multipliers based on the sparse sampling data of the multiple sampling points, the observation matrix, the additive white Gaussian noise level, the regularization term, and the ADMM parameters; perform iterative optimization using an alternating update strategy, wherein the network parameters are updated using the current image reconstruction variables and Lagrange multipliers; the image reconstruction variables are updated based on the current network parameters and Lagrange multipliers, combined with denoising regularization; and the Lagrange multipliers are adjusted to meet the constraints; and the iterative optimization ends when the preset iterative optimization termination condition is met, thereby obtaining the reconstructed radio map.
[0100] Optionally, the above-mentioned device further includes:
[0101] The update module is used to evaluate the accuracy and communication coverage efficiency of the reconstructed radio map at a preset period after obtaining the reconstructed radio map; and to adjust the ADMM parameters and / or the flight strategy of the UAV based on the evaluated accuracy and communication coverage efficiency of the radio map, and to iteratively update the radio map.
[0102] Optionally, the planning module 304 is further configured to learn the flight trajectory of the target UAV offline using the Deep Adversarial Double Q Network (D3QN) algorithm based on the radio map.
[0103] Optionally, the above-mentioned device further includes:
[0104] The adjustment module is used to draw a radio map during the flight of the target UAV based on the flight trajectory, and to adjust the flight path and communication strategy of the target UAV according to the currently drawn radio map.
[0105] It should be noted that the device in this embodiment corresponds to the method applied to the UAV described above. The implementation methods in each of the above embodiments are also applicable to the embodiments of this device and can achieve the same technical effect. The device provided in this application embodiment can implement all the method steps implemented in the above method embodiments and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiments and the beneficial effects will not be described in detail.
[0106] Another embodiment of this application provides a drone path planning device based on radio map reconstruction, such as... Figure 4 As shown, the device includes a transceiver 410, a processor 400, a memory 420, and a program or instructions stored in the memory 420 and executable on the processor 400. When the processor 400 executes the program or instructions, it implements the various processes of the above method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here. This UAV path planning device can be installed in the UAV or located in a computing device outside the UAV.
[0107] The transceiver 410 is used to receive and send data under the control of the processor 400.
[0108] Among them, Figure 4In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 400 and memory represented by memory 420 together. The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. Transceiver 410 can be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over a transmission medium. For different user equipment, user interface 430 can also be an interface capable of connecting external or internal devices, including but not limited to keypads, displays, speakers, microphones, joysticks, etc.
[0109] The processor 400 is responsible for managing the bus architecture and general processing, while the memory 420 can store the data used by the processor 400 when performing operations.
[0110] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the above-described embodiments of the UAV path planning method based on radio map reconstruction, and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.
[0111] This application also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the various processes of the above-described UAV path planning method embodiment based on radio map reconstruction and achieve the same technical effect. To avoid repetition, they will not be described again here.
[0112] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0113] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0114] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A method for unmanned aerial vehicle path planning based on radio map reconstruction, characterized in that, The method comprises: obtaining radio signal data of a plurality of sampling points sampled by a UAV in a target area, wherein the radio signal data of each sampling point comprises three-dimensional coordinates of the sampling point and received signal strength of at least one network device received by the UAV at the sampling point; generating sparse sampling data of the plurality of sampling points according to the radio signal data of the plurality of sampling points, wherein the sparse sampling data of each sampling point comprises received signal-to-interference ratio of each network device by the UAV and connection interruption probability of each network device by the UAV; based on the sparse sampling data of the plurality of sampling points, establishing a radio map model by combining deep image prior (DIP) and a denoising regularization (RED) algorithm; and applying an alternating direction method of multipliers (ADMM) to iteratively optimize the radio map model to obtain a reconstructed radio map; planning a flight path of a target UAV according to the radio map.
2. The method of claim 1, wherein, The method comprises: traversing each sampling point, wherein the following calculations are performed for a currently traversed sampling point: traversing each network device in the at least one network device, wherein the following calculations are performed for a currently traversed network device:
3. The method of claim 1, wherein, treating signals of other network devices except the current network device as interference signals, calculating received signal-to-interference ratio of the current network device by the UAV, and determining connection interruption probability of the current network device by the UAV according to whether the signal-to-interference ratio of the current network device exceeds a preset threshold. based on the sparse sampling data of the plurality of sampling points, establishing a radio map model by combining deep image prior (DIP) and a denoising regularization (RED) algorithm, comprises: inputting the sparse sampling data of the plurality of sampling points into the deep convolutional network, performing multiple iterations of optimization to obtain the radio map model; wherein: each iteration is based on an output of a current deep convolutional network and the sparse sampling data of the plurality of sampling points to calculate a reconstruction error, wherein the reconstruction error is , is an observation matrix, is an output of a deep convolutional network, is sparse sampling data of the plurality of sampling points, is a regularization term, is a regularization parameter; and parameters of the current deep convolutional network are updated with a goal of minimizing the reconstruction error.
4. The method of claim 3, wherein, initializing a deep convolutional network and setting random parameters; applying an alternating direction method of multipliers (ADMM) to iteratively optimize the radio map model to obtain a reconstructed radio map, comprises: initializing iteration times, network parameters, image reconstruction variables, and Lagrange multipliers based on the sparse sampling data of the plurality of sampling points, an observation matrix, an additive white Gaussian noise level, a regularization term, and ADMM parameters; using an alternating update strategy for iterative optimization, wherein network parameters are updated using current image reconstruction variables and Lagrange multipliers, image reconstruction variables are updated based on current network parameters and Lagrange multipliers in combination with denoising regularization, and Lagrange multipliers are adjusted to satisfy a constraint condition; 5. The method of claim 4, wherein, when a preset iteration optimization end condition is met, ending the iteration optimization to obtain the reconstructed radio map. after obtaining the reconstructed radio map, the method further comprises: evaluating accuracy and communication coverage efficiency of the radio map at a preset period; 6. The method of claim 1, wherein, based on the evaluated accuracy and communication coverage efficiency of the radio map, adjusting the ADMM parameters and / or flight strategy of the UAV to iteratively update the radio map. planning a flight path of a target UAV according to the radio map, comprises: using a deep Q-network (D3QN) algorithm to learn a flight trajectory of the target UAV offline according to the radio map.
7. The method of claim 6, wherein, Also included are: During the process of the target UAV flying based on the flight trajectory, a radio map is drawn, and the flight path and communication strategy of the target UAV are adjusted according to the currently drawn radio map.
8. An unmanned aerial vehicle path planning apparatus based on radio map reconstruction, characterized by, Comprise: An acquisition module is configured to acquire radio signal data of a plurality of sampling points sampled by a UAV in a target area, wherein the radio signal data of each sampling point comprises three-dimensional coordinates of the sampling point and received signal strength of at least one network device received by the UAV at the sampling point; A generation module is configured to generate sparse sampling data of the plurality of sampling points according to the radio signal data of the plurality of sampling points, wherein the sparse sampling data of each sampling point comprises received signal-to-interference ratio of each network device by the UAV and connection interruption probability of the UAV and each network device; A reconstruction module is configured to establish a radio map model by combining a deep image prior (DIP) and a denoising regularization (RED) algorithm based on the sparse sampling data of the plurality of sampling points, and to obtain a reconstructed radio map by applying an alternating direction method of multipliers (ADMM) to iterative optimization of the radio map model; A planning module is configured to plan a flight path of a target UAV according to the radio map.
9. An unmanned aerial vehicle path planning device based on radio map reconstruction, characterized by, Comprise: A transceiver, a processor, a memory, and a program or instructions stored on the memory and executable on the processor; The processor executes the program or instructions to implement the steps of the method of any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium, and when executed by the processor, the steps of the method of any one of claims 1 to 7 are implemented.
11. A computer program product, characterised in that, The computer program comprises computer instructions, and when executed by the processor, the steps of the method of any one of claims 1 to 7 are implemented.