Positioning equipment position planning method and device, electronic equipment and medium

Abstract

The invention provides a positioning equipment position planning method and device, electronic equipment and a medium, and the method comprises the steps: obtaining an indoor three-dimensional map and wireless signal data, obtaining environment feature data according to the indoor three-dimensional map, inputting the environment feature data and the wireless signal data into a trained positioning equipment position prediction model for processing, and obtaining a positioning equipment position prediction model; and outputting the initial positioning equipment position, calculating the signal coverage intensity, the positioning equipment uniformity and the positioning equipment deployment cost in the indoor three-dimensional map according to the initial positioning equipment position, further calculating the positioning equipment fitness, adjusting the initial positioning equipment position according to the positioning equipment fitness, and obtaining the target positioning equipment position. According to the method, the accurate three-dimensional map data and the positioning equipment position prediction model are utilized to realize automatic planning of the position of the positioning equipment, the optimal positioning equipment arrangement scheme can be obtained through optimization, and the comprehensiveness and uniformity of positioning signal coverage are ensured.

Description

Technical Field

[0001] The present invention relates to the field of positioning equipment deployment, and in particular to a positioning equipment position planning method, device, electronic equipment and medium. Background Art

[0002] In the deployment of existing indoor positioning systems, the currently commonly used technical solution is to first use drawing software to draw an indoor floor plan, and then the technicians manually design the physical installation location of the positioning equipment based on their experience. Since the actual application usually involves the planning of dozens or even hundreds of positioning devices, the existing planning method is inefficient and requires a lot of manpower. In addition, the existing planning method requires technicians to combine manual observation of the indoor environment to determine whether the location of the set positioning device can fully cover the area, and it is impossible to confirm whether the location of the positioning device is the optimal point, resulting in no signal coverage or poor signal quality in some locations during the actual application process, or too many positioning devices covering a certain area resulting in increased coverage costs.

[0003] Existing technologies rely heavily on the experience of technicians and on-site surveys to plan the installation location of positioning equipment. This method can easily lead to unreasonable point planning, which may result in incomplete signal coverage or multipath effects due to building obstructions, thereby reducing the accuracy of the positioning system. Secondly, the positioning accuracy is affected by the error between the drawings and the actual indoor environment, as well as the problem that it is difficult to install the equipment exactly according to the planned points during installation. The combination of these two errors greatly reduces the final positioning accuracy. In addition, current technologies lack effective automated tools and methods to simplify the deployment and maintenance process, which not only increases the complexity and cost of deployment, but also affects the stability and maintainability of the system. Summary of the invention

[0004] In view of the above problems, the embodiments of the present application propose a positioning device position planning method, apparatus, electronic device and medium.

[0005] In a first aspect of the implementation of the present application, a positioning device position planning method is first provided, characterized in that the method comprises:

[0006] Obtain indoor 3D maps and wireless signal data;

[0007] Acquiring environmental feature data according to the indoor three-dimensional map;

[0008] Inputting the environmental feature data and the wireless signal data into a trained positioning device position prediction model for processing, and outputting an initial positioning device position;

[0009] Calculate the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map according to the initial positioning device position;

[0010] Calculate the adaptability of the positioning device according to the signal coverage strength, the uniformity of the positioning device, and the deployment cost of the positioning device;

[0011] The initial positioning device position is adjusted according to the positioning device fitness to obtain the target positioning device position.

[0012] Optionally, the indoor three-dimensional map is collected by a vehicle, the vehicle includes a laser radar and a camera, and the acquisition of the indoor three-dimensional map and wireless signal data includes:

[0013] Get the preset driving route;

[0014] In the process of driving along the driving route, the three-dimensional point cloud data is collected by the laser radar, and the two-dimensional image data is collected by the camera;

[0015] synthesizing an initial three-dimensional map according to the three-dimensional point cloud data and the two-dimensional image data;

[0016] Performing feature extraction on the three-dimensional point cloud data to obtain three-dimensional feature data;

[0017] Processing the initial three-dimensional map according to the three-dimensional feature data to obtain the indoor three-dimensional map;

[0018] The wireless signal data is acquired according to preset attribute information of the positioning device.

[0019] Optionally, the extracting features from the three-dimensional point cloud data to obtain three-dimensional feature data includes:

[0020] Acquire voxel data and original point data according to the three-dimensional point cloud data;

[0021] Performing feature extraction on the voxel data to obtain candidate region features;

[0022] Performing feature extraction on the original point data to obtain key point features;

[0023] The three-dimensional feature data is determined according to the candidate region features and the key point features.

[0024] Optionally, the positioning device position prediction model is trained in the following manner:

[0025] Acquire a standard indoor three-dimensional map, wherein the standard indoor three-dimensional map includes a plurality of standard positioning devices arranged at positions of standard indoor positioning devices;

[0026] Acquiring standard environmental feature data according to the standard indoor three-dimensional map;

[0027] Acquiring standard wireless signal data according to preset attribute information of the standard positioning device;

[0028] Preprocessing at least one of the standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, wherein the preprocessing is one or more of data cleaning, missing data filling, normalization, and classification coding;

[0029] According to the pre-processed standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, the training environment feature data, the training wireless signal data, the training indoor positioning device position, the verification environment feature data, and the verification wireless signal data are obtained;

[0030] Using the training environment feature data, the training wireless signal data, and the position of the training indoor positioning device as training samples;

[0031] Using the training environment feature data and the training wireless signal data as inputs of the positioning device position prediction model, using the training indoor positioning device position as output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model to obtain a positioning device position prediction model to be verified;

[0032] Using the verification environment feature data and the verification wireless signal data as verification samples;

[0033] Inputting the verification environment feature data and the verification wireless signal data into the positioning device position prediction model to be verified, and obtaining the verification indoor positioning device position output by the positioning device position prediction model to be verified;

[0034] Calculate the positioning device position prediction model loss function according to the training indoor positioning device position and the verification indoor positioning device position;

[0035] The positioning device position prediction model loss function is used to adjust the model parameters of the positioning device position prediction model to be verified, and the steps of using the training environment feature data and the training wireless signal data as the input of the positioning device position prediction model, using the training indoor positioning device position as the output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model are re-executed until a preset stop condition is reached to obtain a trained positioning device position prediction model.

[0036] Optionally, the positioning device position prediction model is composed of a sequential model, the sequential model includes an input layer, a plurality of hidden layers, and an output layer, and the environmental feature data and the wireless signal data are input into a trained positioning device position prediction model for processing, and an initial positioning device position is output, including:

[0037] Inputting the environmental feature data and the wireless signal data into the input layer for processing to obtain a first positioning device feature output by the input layer;

[0038] Inputting the first positioning device feature into the hidden layer for processing to obtain a second positioning device feature output by the hidden layer;

[0039] The second positioning device feature is input into the output layer for processing to obtain the initial positioning device position output by the output layer.

[0040] Optionally, the calculating, according to the initial positioning device position, the signal coverage strength, the uniformity of the positioning device, and the positioning device deployment cost in the indoor three-dimensional map includes:

[0041] Acquire a positioning device signal radius according to preset attribute information of the positioning device;

[0042] Calculate the signal coverage strength in the indoor three-dimensional map according to the initial positioning device position and the positioning device signal radius;

[0043] The positioning device uniformity and the positioning device deployment cost are calculated according to the initial positioning device position.

[0044] Optionally, adjusting the initial positioning device position according to the positioning device fitness to obtain the target positioning device position includes:

[0045] Performing a random perturbation process on the initial positioning device position for a preset number of times to obtain a plurality of groups of perturbed positioning device positions;

[0046] Calculating the disturbance signal coverage strength, the disturbance positioning device uniformity, and the disturbance positioning device deployment cost in the indoor three-dimensional map according to each group of disturbance positioning device positions;

[0047] Calculating the fitness of the disturbance positioning device corresponding to each group of the disturbance positioning device positions according to the disturbance signal coverage strength, the uniformity of the disturbance positioning device, and the deployment cost of the disturbance positioning device;

[0048] The fitness of the disturbance positioning device corresponding to several groups of the disturbance positioning device positions is compared, and a group of the disturbance positioning device positions corresponding to the maximum fitness of the disturbance positioning device is used as the target positioning device position.

[0049] In a second aspect of the implementation of the present application, a positioning device position planning device is also provided, characterized in that the device comprises:

[0050] A data acquisition module, used to acquire indoor three-dimensional maps and wireless signal data;

[0051] An environmental feature data extraction module, used to obtain environmental feature data according to the indoor three-dimensional map;

[0052] An initial positioning device position acquisition module is used to input the environmental feature data and the wireless signal data into a trained positioning device position prediction model for processing, and output an initial positioning device position;

[0053] An optimization parameter calculation module, used to calculate the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map according to the initial positioning device position;

[0054] A fitness calculation module, used to calculate the fitness of the positioning device according to the signal coverage strength, the uniformity of the positioning device, and the deployment cost of the positioning device;

[0055] The target positioning device position acquisition module is used to adjust the initial positioning device position according to the positioning device fitness to obtain the target positioning device position.

[0056] In the third aspect of the implementation of the present application, there is also provided an electronic device, characterized in that it includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, and when the computer program is executed by the processor, it implements the positioning device position planning method as described above.

[0057] In a fourth aspect of the implementation of the present application, a computer-readable storage medium is further provided, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the positioning device position planning method as described above is implemented.

[0058] The embodiments of the present application have the following advantages:

[0059] In the embodiment of the present application, firstly, by obtaining the indoor three-dimensional map and wireless signal data, the environmental feature data is obtained according to the indoor three-dimensional map, which provides more accurate and reliable map data support for the planning of the positioning device point. This method not only reduces the dependence on manual survey, but also improves the rationality of point planning, and effectively avoids problems such as incomplete signal coverage and multipath effect. Secondly, the environmental feature data and wireless signal data are input into the trained positioning device position prediction model for processing, and the initial positioning device position is output. The automatic planning of the positioning device point can be realized in the current indoor three-dimensional map. Through accurate three-dimensional map data and advanced algorithms, automatic planning can more accurately determine the optimal installation position of the positioning device, effectively avoid errors and deviations that may occur in manual planning, and improve the efficiency of positioning device deployment. Finally, the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map are calculated according to the initial positioning device position, and the positioning device fitness is calculated according to the signal coverage strength, positioning device uniformity, and positioning device deployment cost. The initial positioning device position is adjusted according to the positioning device fitness to obtain the target positioning device position. Deep learning network technology can be used to verify and optimize the positioning network signal coverage in indoor environments, which not only ensures the comprehensiveness and uniformity of signal coverage, but also takes into account the economic efficiency of deployment costs, thereby achieving optimal resource allocation and ensuring the reliability and stability of the positioning network. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art.

[0061] Figure 1 This is a flowchart of a method for positioning a positioning device according to an embodiment of the present application;

[0062] Figure 2 It is a structural diagram of a positioning device position planning device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in the embodiments of the present invention, many technical details are proposed in order to enable readers to better understand the present invention. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present invention can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined and referenced with each other without contradiction.

[0064] Reference Figure 1 , showing a step flow chart of a positioning device position planning method provided by an embodiment of the present application.

[0065] The method may specifically include the following steps:

[0066] Step 101, obtaining indoor three-dimensional map and wireless signal data.

[0067] It should be noted that an indoor three-dimensional map can be a digital map that depicts the structure and layout of the indoor space in detail. It accurately reproduces the physical environment and spatial characteristics of the room through three-dimensional modeling technology. The creation of indoor three-dimensional maps usually involves the use of technologies such as laser scanning, photogrammetry or computer-aided design (CAD) to obtain high-precision spatial data. The accuracy and detail of the indoor three-dimensional map are crucial to improving the performance of the indoor positioning system. It provides the necessary data support for automatic planning and optimization of the position of the positioning device, ensuring the accuracy and reliability of the positioning service. In an embodiment of the present application, the indoor three-dimensional map can be obtained by SLAM (Simultaneous Localization and Mapping) technology.

[0068] Among them, SLAM (Simultaneous Localization and Mapping) technology allows robots to simultaneously build a map of the environment and determine their own position when moving in an unknown environment. SLAM technology has a wide range of applications in robot navigation, driverless cars, augmented reality (AR) and virtual reality (VR).

[0069] The basic principles of SLAM include the following key steps:

[0070] Sensor data collection: Collect environmental data using sensors such as laser scanners, cameras, ultrasonic sensors, and inertial measurement units (IMUs).

[0071] Data association: Matching newly collected sensor data with known map data to determine relationships between data points.

[0072] Position estimation: Using sensor data and data association results, the current position of the robot is estimated through algorithms (such as Kalman filtering, particle filtering, etc.).

[0073] Map Update: Update and improve the environment map based on the robot's position estimate and sensor data.

[0074] In an embodiment of the present application, the indoor three-dimensional map obtained by using SLAM technology may be a three-dimensional map of the indoor environment constructed when the vehicle moves in an unknown environment.

[0075] It should be noted that wireless signal data can be the basic information required to achieve accurate location estimation. Wireless signal data is crucial in indoor positioning systems. The following is a detailed description of these characteristics and how they are collected:

[0076] Frequency: The frequency of a wireless signal refers to the number of times the signal oscillates per second, measured in Hertz (Hz). Commonly used wireless communication technologies in indoor positioning, such as Zigbee, Bluetooth, and Wi-Fi, usually operate in the 2.4 GHz frequency band.

[0077] Wavelength: Wavelength is the distance a signal travels in one complete cycle and is inversely proportional to frequency. Wavelength can be calculated using the formula λ = c / f, where λ is the wavelength, c is the speed of light, and f is the frequency.

[0078] Transmit power: This refers to the energy intensity of the signal transmitted by the wireless device, usually measured in milliwatts (mW) or decibel milliwatts (dBm). The transmit power is determined by the hardware specifications of the wireless device and can be set and adjusted by consulting the device's technical manual or through the software interface.

[0079] The methods for collecting these wireless signal characteristic data usually include:

[0080] Hardware query: Directly read parameters such as frequency, wavelength, and transmission power through the device's hardware interface.

[0081] Software configuration: Use the software tools or APIs provided by the device to obtain and set these parameters.

[0082] Signal Analysis: Use specialized signal analysis tools or software to measure and record the characteristics of wireless signals.

[0083] By precisely controlling and optimizing the transmission power and frequency of wireless signals, multipath effects and signal interference can be reduced, thereby improving the performance of the positioning system.

[0084] In an embodiment of the present application, the wireless signal data may be used as an input to a positioning device location prediction model to obtain the location of the positioning device.

[0085] In an embodiment of the present application, indoor three-dimensional maps and wireless signal data can be obtained to facilitate subsequent positioning device location planning.

[0086] Step 102: Acquire environmental feature data according to the indoor three-dimensional map.

[0087] It should be noted that environmental feature data is a key parameter in indoor 3D maps, including but not limited to the size, shape characteristics and obstacle distribution of indoor space. The specific meaning can be explained as follows:

[0088] Indoor space dimensions can include the length, width, and height of the indoor space. These basic dimensions determine the size of the indoor space and directly affect the propagation distance and coverage of the wireless signal. Larger spaces may require more positioning devices to ensure full coverage of the signal. Indoor space dimensions can be obtained through architectural drawings, laser scanning, or direct measurement using measuring tools (such as tape measures).

[0089] Shape characteristics can be the specific shape of the indoor space. If the shape of the indoor space is complex or irregular, special feature description methods may be needed to quantify the complexity of its shape. For example, using the ratio of the perimeter to the area of ​​the space can be used as a simple shape characteristic indicator. The shape characteristics of the indoor space can be accurately captured by 3D modeling software or laser scanning technology.

[0090] Obstacle distribution can include obstacle type, quantity, and size. Obstacles such as walls, furniture, and equipment can affect the propagation path of wireless signals, causing signal obstruction, reflection, or diffraction. Identifying and quantifying the characteristics of these obstacles is critical to optimizing the layout of positioning equipment. Obstacle distribution can be identified and recorded through indoor 3D maps, sensor data, or manual surveys.

[0091] In the design and deployment of indoor positioning systems, comprehensive consideration of these environmental factors can help optimize the layout of positioning equipment and improve the uniformity of signal coverage and positioning accuracy. For example, in irregularly shaped indoor spaces, a denser layout of positioning equipment may be required to compensate for the unevenness of signal propagation; in environments with many obstacles, the location of the equipment may need to be adjusted or the number of equipment may need to be increased to overcome the impact of signal obstruction. With accurate environmental feature data, a more intelligent and efficient indoor positioning solution can be achieved.

[0092] In an embodiment of the present application, environmental feature data may be used as input to a positioning device position prediction model to obtain the positioning device position.

[0093] In an embodiment of the present application, environmental feature data can be acquired based on an indoor three-dimensional map.

[0094] This application obtains environmental feature data based on indoor three-dimensional maps, providing more accurate and reliable map data support for the planning of positioning equipment points. This method not only reduces the reliance on manual surveys, but also improves the rationality of point planning, effectively avoiding problems such as incomplete signal coverage and multipath effects.

[0095] Step 103: Input the environmental feature data and the wireless signal data into a trained positioning device position prediction model for processing, and output an initial positioning device position.

[0096] It should be noted that the positioning device location prediction model can be an algorithm model for determining the optimal installation location of the indoor positioning device. The positioning device location prediction model can be based on a machine learning algorithm and use indoor environment feature data and wireless signal propagation characteristics to predict the layout of the positioning device.

[0097] In the embodiment of the present application, the environmental feature data and the wireless signal data can be input into the trained positioning device position prediction model for processing, and the initial positioning device position can be output. The initial positioning device position can be a number of three-dimensional position coordinates in the indoor three-dimensional map, and the initial positioning device position represents the initial arrangement position of a number of positioning devices.

[0098] This application processes environmental feature data and wireless signal data into a trained positioning device location prediction model and outputs the initial positioning device location. This can realize automatic planning of positioning device points in the current indoor three-dimensional map. Through precise three-dimensional map data and advanced algorithms, automatic planning can more accurately determine the optimal installation position of the positioning device, effectively avoiding errors and deviations that may occur in manual planning, and improving the deployment efficiency of positioning equipment.

[0099] Step 104: Calculate the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map according to the initial positioning device position.

[0100] It should be noted that in indoor 3D maps, signal coverage strength, positioning device uniformity and positioning device deployment cost are key indicators for evaluating and optimizing indoor positioning system performance. The following is a detailed description of these indicators:

[0101] Signal coverage strength can refer to the strength and uniformity of wireless signals in indoor spaces. Ideal signal coverage should provide stable and strong enough signals throughout the indoor space to ensure the accuracy and reliability of positioning services. The signal strength of each location can be determined through simulation or actual measurement, and the signal coverage uniformity of the entire space can be calculated.

[0102] The uniformity of positioning devices refers to the uniformity of the distribution of positioning devices in the indoor space. Uniform distribution of devices can ensure comprehensive signal coverage and avoid signal blind spots. The uniformity of distribution can be evaluated by calculating the distance between positioning devices or the device density.

[0103] The deployment cost of positioning equipment can include equipment purchase cost, installation cost and maintenance cost. When optimizing the layout of positioning equipment, cost-effectiveness needs to be considered to ensure that the overall cost is minimized while meeting performance requirements. The deployment cost can be evaluated by calculating the number of devices, installation difficulty and expected maintenance cost.

[0104] In an embodiment of the present application, the signal coverage strength in the indoor three-dimensional map, the uniformity of the positioning devices in the indoor three-dimensional map, and the positioning device deployment cost in the indoor three-dimensional map can be calculated according to the initial positioning device position.

[0105] Step 105: Calculate the positioning device fitness according to the signal coverage strength, the positioning device uniformity, and the positioning device deployment cost.

[0106] It should be noted that in the indoor positioning system, the positioning device fitness can be a comprehensive evaluation indicator that can reflect the comprehensive performance of the positioning device layout in terms of signal coverage strength, positioning device uniformity and deployment cost. The higher the positioning device fitness score, the better the overall performance of the layout scheme in terms of signal coverage, uniform device distribution and cost-effectiveness.

[0107] In an embodiment of the present application, the adaptability of the positioning device can be calculated based on the signal coverage strength, the uniformity of the positioning device, and the deployment cost of the positioning device in the indoor three-dimensional map.

[0108] Step 106: adjust the initial positioning device position according to the positioning device fitness to obtain the target positioning device position.

[0109] It should be noted that the target positioning device position can be an optimized positioning device layout in an indoor three-dimensional map, which can achieve optimal signal coverage, uniform distribution of equipment and cost-effectiveness. In an embodiment of the present application, the target positioning device position can be the optimal positioning device layout in an indoor three-dimensional map.

[0110] In the embodiment of the present application, the initial positioning device position can be adjusted according to the fitness of the positioning device to obtain the target positioning device position. The method of adjusting the initial positioning device position according to the fitness of the positioning device can be iterative optimization. The target positioning device position can be a set of two-dimensional position coordinates in an indoor three-dimensional map, and the target positioning device position represents the optimal arrangement position of several positioning devices. The installation height of the positioning device can be selected according to the actual installation environment.

[0111] In the embodiment of the present application, firstly, by obtaining the indoor three-dimensional map and wireless signal data, the environmental feature data is obtained according to the indoor three-dimensional map, which provides more accurate and reliable map data support for the planning of the positioning device point. This method not only reduces the dependence on manual survey, but also improves the rationality of point planning, and effectively avoids problems such as incomplete signal coverage and multipath effect. Secondly, the environmental feature data and wireless signal data are input into the trained positioning device position prediction model for processing, and the initial positioning device position is output. The automatic planning of the positioning device point can be realized in the current indoor three-dimensional map. Through accurate three-dimensional map data and advanced algorithms, automatic planning can more accurately determine the optimal installation position of the positioning device, effectively avoid errors and deviations that may occur in manual planning, and improve the efficiency of positioning device deployment. Finally, the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map are calculated according to the initial positioning device position, and the positioning device fitness is calculated according to the signal coverage strength, positioning device uniformity, and positioning device deployment cost. The initial positioning device position is adjusted according to the positioning device fitness to obtain the target positioning device position. Deep learning network technology can be used to verify and optimize the positioning network signal coverage in indoor environments, which not only ensures the comprehensiveness and uniformity of signal coverage, but also takes into account the economic efficiency of deployment costs, thereby achieving optimal resource allocation and ensuring the reliability and stability of the positioning network.

[0112] In an optional embodiment of the present application, the indoor three-dimensional map is collected by a vehicle, and the vehicle includes a laser radar and a camera. Step 101 also includes the following sub-steps:

[0113] S111: Obtaining a preset driving route;

[0114] S112: In a process of driving along the driving route, collecting three-dimensional point cloud data by the laser radar and collecting two-dimensional image data by the camera;

[0115] S113: synthesizing an initial three-dimensional map according to the three-dimensional point cloud data and the two-dimensional image data;

[0116] S114: extracting features from the three-dimensional point cloud data to obtain three-dimensional feature data;

[0117] S115: Processing the initial three-dimensional map according to the three-dimensional feature data to obtain the indoor three-dimensional map;

[0118] S116: Acquire the wireless signal data according to preset attribute information of the positioning device.

[0119] It should be noted that the vehicle can be a vehicle including a computing unit, wherein the computing unit can process data collected by sensors installed on the vehicle, and the data processing process of the computing unit can be performed on the vehicle or on a cloud server. In a specific implementation, the computing unit can be a computer or other device.

[0120] It should be noted that the driving route may be a preset driving route of the vehicle in the current indoor environment. The current indoor environment includes but is not limited to an underground parking lot. In the embodiment of the present application, the vehicle may travel according to the driving route, and the vehicle may travel according to the driving route in the case of automatic driving.

[0121] It should be noted that LiDAR and cameras are two commonly used sensor technologies, which are widely used in autonomous driving, robot navigation, augmented reality (AR), virtual reality (VR) and indoor positioning. The following is a detailed introduction to these two technologies:

[0122] LiDAR can determine the distance of a target by emitting laser pulses and measuring their return time. It is able to generate a high-precision three-dimensional map of the environment, including the shape, size and position of objects. LiDAR can provide centimeter-level distance measurement accuracy, can acquire a large number of data points in a short period of time, achieve rapid environmental perception, and can work stably in various lighting conditions.

[0123] Cameras can generate images by capturing light and converting it into electrical signals. Cameras can be equipped with image processing chips that enable real-time image analysis and recognition.

[0124] In an embodiment of the present application, the lidar can provide accurate distance and three-dimensional structure information, while the camera can provide rich visual details and color information. The combination of the two can achieve more comprehensive and reliable environmental perception.

[0125] It should be noted that three-dimensional point cloud data and two-dimensional image data are two different types of data, which are widely used in computer vision, robotics, geographic information system (GIS) and other fields.

[0126] 3D point cloud data can be a data set consisting of a large number of 3D points, each of which usually contains its coordinates (X, Y, Z) in 3D space and possible other attributes (such as color, intensity, etc.). It can be obtained through LiDAR, structured light scanning, stereo vision system or other 3D scanning technology.

[0127] The two-dimensional image data may be a two-dimensional array of pixels, each pixel containing its position (X, Y) in the image and color information (such as RGB value). It may be acquired by a camera, scanner or other image acquisition device. In an embodiment of the present application, the three-dimensional point cloud data and the two-dimensional image data may be used to generate an initial three-dimensional map.

[0128] It should be noted that the initial three-dimensional map can be a high-precision indoor three-dimensional map generated by fusing the three-dimensional point cloud data and the two-dimensional image data collected by different sensors through the perception fusion technology. Among them, fusing the data collected by different sensors through the perception fusion technology can refer to fusing multimodal data at the feature level of the lidar branch, but fusing at the data set and feature level of the image branch. In the embodiment of the present application, the initial three-dimensional map is the initial indoor three-dimensional map obtained by data fusion.

[0129] It should be noted that the three-dimensional feature data may be three-dimensional feature data in an indoor three-dimensional map. Taking the indoor three-dimensional map as an example, which is a map of an underground parking lot, the three-dimensional feature data may be wall corners, columns, garage floor signs, etc. This application does not restrict which three-dimensional features the three-dimensional feature data specifically includes.

[0130] In an embodiment of the present application, the indoor three-dimensional map may be a more accurate indoor three-dimensional map obtained by post-processing the initial three-dimensional map, and the post-processing includes but is not limited to denoising, feature extraction, etc.

[0131] It should be noted that the preset attribute information of the positioning device may be an inherent attribute of the positioning device to be deployed. The following are some common preset attribute information of positioning devices:

[0132] The device type may be a type of positioning device, such as a Wi-Fi access point, a Bluetooth beacon, an ultra-wideband (UWB) tag, an RFID reader, and the like.

[0133] The communication protocol may be a communication protocol used by the positioning device, such as IEEE 802.11 (Wi-Fi), Bluetooth, Zigbee, LoRa, etc.

[0134] Transmit power can be the power of the signal transmitted by the positioning device, usually measured in milliwatts (mW) or decibel milliwatts (dBm).

[0135] The frequency range may be a frequency range in which the positioning device operates, such as 2.4 GHz, 5 GHz, etc.

[0136] Signal coverage can be the range of physical space that a device can effectively cover, which is usually related to transmission power and environmental factors.

[0137] Before deploying positioning equipment, it is very important to understand and record these preset attribute information because they directly affect the installation, configuration and performance of the equipment. For example, the transmission power and frequency range will affect the propagation and coverage of the signal. By comprehensively considering these attributes, the layout of positioning equipment can be planned and optimized more effectively to ensure that the system can provide accurate and reliable positioning services.

[0138] In the embodiment of the present application, the vehicle can obtain wireless signal data according to the preset attribute information of the positioning device, wherein the wireless signal data can be basic parameters such as the frequency, wavelength, and transmission power of the wireless signal.

[0139] In the embodiment of the present application, the vehicle can obtain a preset driving route, and in the process of driving along the driving route, collect three-dimensional point cloud data through the laser radar, collect two-dimensional image data through the camera, and then synthesize the initial three-dimensional map based on the three-dimensional point cloud data and the two-dimensional image data. Perform feature extraction on the three-dimensional point cloud data to obtain three-dimensional feature data, and then process the initial three-dimensional map based on the three-dimensional feature data to obtain an indoor three-dimensional map. And wireless signal data can be obtained according to the preset attribute information of the positioning device.

[0140] In the specific implementation, the autonomous driving vehicle can be started from a preset starting point in the indoor environment used for model training, and is ready to start the scanning process. The sensors carried by the vehicle, such as lidar and cameras, start to automatically collect indoor environment data. Then, the perception fusion technology of autonomous driving is used to fuse the data collected by different sensors to generate a high-precision indoor three-dimensional map to ensure the accuracy and completeness of the map. Finally, the generated map is post-processed, including denoising, feature extraction (such as corners, columns, garage floor signs, etc.), map optimization, etc., to improve the usability and accuracy of the map.

[0141] In an optional embodiment of the present application, step S114 further includes the following sub-steps:

[0142] S1141: Acquire voxel data and original point data according to the three-dimensional point cloud data;

[0143] S1142: extracting features from the voxel data to obtain candidate region features;

[0144] S1143: Extracting features from the original point data to obtain key point features;

[0145] S1144: Determine the three-dimensional feature data according to the candidate region features and the key point features.

[0146] It should be noted that voxel data can be a form of data representation in three-dimensional space, similar to pixels in two-dimensional images, but extended to three-dimensional space. Voxel is the abbreviation of "Volumetric Pixel", which represents a basic unit in three-dimensional space. Each voxel usually contains its position information in three-dimensional space (X, Y, Z coordinates) and possible other attributes (such as color, density, transparency, etc.). In an embodiment of the present application, voxel data can be used to obtain candidate region features.

[0147] Raw point cloud data may refer to unprocessed three-dimensional point cloud data directly obtained through three-dimensional scanning technology (such as laser radar, structured light scanning, stereo vision, etc.). These data contain surface information of the scanned object or environment, and each point usually contains its coordinates (X, Y, Z) in three-dimensional space and possible other attributes (such as color, reflection intensity, timestamp, etc.). In the embodiment of the present application, voxel data can be used to obtain key point features.

[0148] Region Proposal Features may refer to a series of features used in computer vision and image processing to identify and extract regions in an image that may contain objects of interest (such as people, animals, vehicles, etc.). These features are typically used to generate Region Proposals, i.e., potential regions that may contain target objects in an image. In an embodiment of the present application, the Region Proposal Features may be potential regions that contain target objects such as corners, columns, and garage floor signs.

[0149] Keypoint features may refer to a set of attributes used to describe significant and stable local features in an image in image processing and computer vision. These key points are usually points in an image that are not easily affected by changes in illumination, scale, rotation, and perspective, such as corner points, edge points, or texture-rich areas. In an embodiment of the present application, keypoint features may be points in an indoor three-dimensional map that are not easily affected by changes in illumination, scale, rotation, and perspective.

[0150] In an embodiment of the present application, voxel data and original point data can be obtained based on the three-dimensional point cloud data, and then feature extraction can be performed on the voxel data to obtain candidate area features, and feature extraction can be performed on the original point data to obtain key point features, and finally three-dimensional feature data can be determined based on the candidate area features and key point features.

[0151] In a specific implementation, a three-dimensional target detection algorithm model based on fused-perceived 3D point cloud data can be used to extract features from the three-dimensional point cloud data.

[0152] The point cloud data is input into the global feature extraction model and the local feature extraction model for processing. In the global feature extraction model, the three-dimensional point cloud data is processed into point cloud voxels. The processed voxel data is first subjected to down-sampling feature extraction, and then subjected to up-sampling feature extraction to obtain the candidate region features.

[0153] In the local feature extraction model, the original point data is obtained according to the three-dimensional point cloud data. The key point sampling results are obtained by sampling the key points of the original point data, and the key point sampling results are input into the point cloud learning network (multi-layer perceptron) to obtain the key point features.

[0154] The candidate region features and key point features are input into the detection model for feature fusion to generate a candidate region containing the target object. The target object is then classified and the candidate box is used for regression processing to obtain the position of the target object in the three-dimensional map. The detected target object and its position are output.

[0155] Among them, the global feature extraction model can refer to a type of model used to extract global features from the entire image in image processing and computer vision. These models usually focus on the overall properties of the image, such as color distribution, texture structure, shape information, etc., rather than local details. In a specific implementation, the entire 3D point cloud data can be processed.

[0156] Local feature extraction models can refer to a class of models used to extract features from local areas of an image in image processing and computer vision. These models focus on small areas or local structures in the image, such as corners, edges, textures, etc. In specific implementations, key points in 3D point cloud data can be processed.

[0157] Detection models can be used in computer vision and image processing to identify and locate specific target objects in images or videos. These models usually need to complete two tasks at the same time: target classification (determine whether a certain category of objects exists in the image) and target localization (determine the position and size of the object in the image). In specific implementations, they can be used to detect target objects such as corners, columns, and garage floor signs in indoor 3D maps.

[0158] In an optional embodiment of the present application, the positioning device position prediction model is trained in the following manner:

[0159] S201: Acquire a standard indoor three-dimensional map, where the standard indoor three-dimensional map includes a number of standard positioning devices arranged at positions of standard indoor positioning devices;

[0160] S202: Acquire standard environment feature data according to the standard indoor three-dimensional map;

[0161] S203: Acquire standard wireless signal data according to preset attribute information of the standard positioning device;

[0162] S204: preprocessing at least one of the standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, wherein the preprocessing is one or more of data cleaning, missing data filling, normalization, and classification coding;

[0163] S205: Acquire training environment feature data, training wireless signal data, training indoor positioning device location, verification environment feature data, and verification wireless signal data according to the pre-processed standard environment feature data, the standard wireless signal data, and the standard indoor positioning device location;

[0164] S206: Using the training environment feature data, the training wireless signal data, and the position of the training indoor positioning device as training samples;

[0165] S207: using the training environment feature data and the training wireless signal data as inputs of the positioning device position prediction model, using the training indoor positioning device position as output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model to obtain a positioning device position prediction model to be verified;

[0166] S208: Using the verification environment feature data and the verification wireless signal data as verification samples;

[0167] S209: Inputting the verification environment feature data and the verification wireless signal data into the positioning device position prediction model to be verified, and obtaining the verification indoor positioning device position output by the positioning device position prediction model to be verified;

[0168] S210: Calculating a positioning device position prediction model loss function according to the training indoor positioning device position and the verification indoor positioning device position;

[0169] S211: Use the positioning device position prediction model loss function to adjust the model parameters of the positioning device position prediction model to be verified, and re-execute the steps of using the training environment feature data and the training wireless signal data as the input of the positioning device position prediction model, using the training indoor positioning device position as the output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model until a preset stop condition is reached to obtain a trained positioning device position prediction model.

[0170] It should be noted that the standard indoor three-dimensional map can be a number of optimal indoor positioning network environments formed in advance in multiple indoor environments for model training through multiple rounds of configuration and debugging by wireless positioning experts. The standard indoor three-dimensional map can be used to provide the training data and test data required for subsequent machine learning.

[0171] In an embodiment of the present application, a vehicle can obtain a standard indoor three-dimensional map, and the standard indoor three-dimensional map includes several standard positioning devices arranged at the positions of standard indoor positioning devices. The method of obtaining the standard indoor three-dimensional map can be that the vehicle travels along a preset driving route in an optimal indoor positioning network environment, and in the process of traveling along the driving route, three-dimensional point cloud data is collected by a laser radar, two-dimensional image data is collected by a camera, and then an initial three-dimensional map is synthesized based on the three-dimensional point cloud data and the two-dimensional image data. Feature extraction is performed on the three-dimensional point cloud data to obtain three-dimensional feature data, and then the initial three-dimensional map is processed based on the three-dimensional feature data to obtain an indoor three-dimensional map.

[0172] In the embodiment of the present application, standard environment feature data may be obtained according to a standard indoor three-dimensional map. The standard environment feature data includes but is not limited to the size, shape characteristics and obstacle distribution of the indoor space in the optimal indoor positioning network environment.

[0173] In the embodiment of the present application, standard wireless signal data can be obtained according to preset attribute information of the standard positioning device, wherein the standard wireless signal data can be basic parameters such as frequency, wavelength, and transmission power of wireless signals in the optimal indoor positioning network environment.

[0174] In an embodiment of the present application, information such as the location of a standard indoor positioning device can be obtained based on a standard indoor three-dimensional map. In a specific implementation, it can be data on the location of wireless positioning points, including the number of positioning devices, the location of standard indoor positioning devices, a unique identifier, signal strength and other information. The specific collection method is as follows: Through the indoor high-precision map collected by the autonomous driving vehicle, after coordinate conversion, the deployed positioning devices can be identified, thereby obtaining the coordinate information of the positioning device points, and then bound and corresponding with the positioning device coding information sorted in advance. At the same time, a wireless signal acquisition terminal is used to follow the autonomous driving vehicle to collect wireless signals at various locations indoors. The data that can be collected by the wireless acquisition terminal include the unique identifier of the positioning device and the signal strength from the device to the location of the acquisition terminal. Since the position of the acquisition terminal is relative and fixed to the position of the autonomous driving vehicle, the indoor relative position of the acquisition terminal at different acquisition times can be obtained later through coordinate conversion.

[0175] Among them, the acquisition terminal refers to a device or system used to collect and transmit data. They are usually located at the source of data generation, responsible for collecting raw data and sending it to a central processing system or cloud server through the network. The acquisition terminal can be various types of devices, including sensors, data loggers, mobile devices, industrial control systems, etc., depending on the application scenario and data type. In an embodiment of the present application, the signal strength from the positioning device to the acquisition terminal location can be collected through the acquisition terminal.

[0176] It should be noted that data cleaning, missing data filling, normalization, and classification coding are several key steps in the data preprocessing process, which are crucial to improving data quality and ensuring the performance of machine learning models. The following is a detailed introduction to each step:

[0177] Data cleaning can refer to the process of identifying and correcting errors and inconsistencies in a data set. This includes removing duplicate records, correcting erroneous values, dealing with outliers, etc. The purpose of data cleaning is to ensure the quality of the data and make it more suitable for analysis and modeling.

[0178] Missing Data Imputation can refer to the process of dealing with missing values ​​in a dataset. Missing values ​​may be caused by data collection errors, equipment failure, or other reasons. Methods for dealing with missing values ​​include: Deleting records with missing values: If there are few missing values, these records can be deleted directly. Mean / Median Filling: Fill missing values ​​with the mean or median of the variable. Mode Filling: Fill missing values ​​with the mode (most frequently occurring value) of the variable. Interpolation: Estimate missing values ​​using linear interpolation or other interpolation methods. Machine Learning Methods: Use regression models or other machine learning algorithms to predict missing values.

[0179] Normalization can refer to scaling data to a common range so that the values ​​of different features are comparable. This is very important for many machine learning algorithms (especially distance-based algorithms such as K-nearest neighbors). Common normalization methods include: Min-Max Normalization: Scaling data to the range of [0,1] or [-1,1]. Z-score Normalization: Converting data to a distribution with a mean of 0 and a standard deviation of 1.

[0180] Categorical encoding can refer to converting categorical variables into numerical form so that machine learning algorithms can process them. Categorical variables are usually in the form of text or labels, such as color, gender, country, etc. Common categorical encoding methods include:

[0181] One-Hot Encoding: Create a binary feature for each category, where only one feature is 1 and the rest are 0.

[0182] Label Encoding: Assigns a unique integer value to each category.

[0183] Ordinal Encoding: Assigning integer values ​​to ordinal categorical variables, maintaining the order of categories.

[0184] Binary Encoding: Convert categories into binary representation to reduce the number of features.

[0185] These preprocessing steps help improve the quality and consistency of the data, thereby improving the performance and reliability of the machine learning model. In practical applications, the appropriate preprocessing method can be determined based on the characteristics of the data and the analysis objectives.

[0186] In an embodiment of the present application, at least one of standard environmental feature data, standard wireless signal data, and standard indoor positioning device position can be preprocessed, and the preprocessing is one or more of data cleaning, missing data filling, normalization, and classification coding.

[0187] In the specific implementation, data cleaning can be understood as follows: if the signal strength at a certain beacon location is significantly higher or lower than the average signal strength of the beacon at the current location when the same location is collected multiple times, and the absolute value of its Z-Score is greater than a specific threshold (such as 3), then the signal strength data at this beacon location may be considered an outlier. These may be caused by measurement errors or the directivity of the antenna. The Z-Score method can be used to identify outliers and delete them from the data set. Z-score is an outlier detection method based on standard scores. Z-score represents the distance between the data point and the mean. The Z-score is obtained by calculating the difference between each data point and the mean and dividing it by the standard deviation. If the Z-score is greater than a specific threshold (such as 3), the data point is considered an outlier. The advantage of this method is that it does not depend on the distribution shape of the data, but only considers the relative distance of the data point from the mean.

[0188] The data cleaning process can use Python's Pandas library to process a DataFrame containing RSSI (received signal strength indication) data. First, create a sample dataset that contains some NaN values, which represent missing values ​​in the data. Next, calculate the mean of the RSSI column and use this mean to fill all NaN values ​​in the DataFrame. The purpose of this is to ensure that the dataset will not be affected by missing values ​​in subsequent analysis or machine learning model training. Finally, the filled DataFrame will no longer contain any NaN values, and all missing data will be replaced with the mean of the column.

[0189] In the specific implementation, the process of data filling can be reflected as follows: for data missing from the training data, you can choose to fill it. For numerical features, this application can use mean filling. Data filling can use Python's NumPy library to process a NumPy array containing RSSI (received signal strength indication) signal strength. First, define an array called data, which contains the actual collected signal quality data. Next, calculate the mean (mean_value) and standard deviation (std_dev) of this array. Then, calculate the Z-score of each data point, which is obtained by subtracting the mean from each data point and dividing it by the standard deviation. Z-score is a standardized method that can help identify outliers in the data. Finally, all data points with a Z-score greater than 3 are found. These data points are considered outliers and are stored in an array called outliers. In this way, extreme values ​​that may be caused by measurement errors or other reasons can be identified and processed.

[0190] In specific implementations, normalization can be due to the fact that sensor data may have different dimensions and numerical ranges, and it is beneficial to scale all features to a uniform range (usually 0 to 1, or -1 to 1). Normalization helps the model converge faster and avoids some features from having a disproportionate impact on model training due to their large numerical range. Normalization can use Python's Pandas library and scikit-learn library to process the DataFrame containing RSSI (received signal strength indication) data. First, create a sample dataset containing a column of data called 'RSSI'. Subsequently, import the MinMaxScaler class in the scikit-learn library for data normalization. Create a MinMaxScaler object to normalize the 'RSSI' column in the DataFrame and scale the data to a range of 0 to 1. The normalized data is stored in a new column 'Normalized_RSSI'. Finally, print the normalized DataFrame to view the processing results. This operation ensures that data of different ranges have the same weight in the machine learning model and improves model performance.

[0191] In a specific implementation, the classification encoding can be encoding categorical data. If the data set contains categorical data (such as the type of indoor space or the material of obstacles), it needs to be converted into numerical data. This application can use one-hot encoding to convert each category into a new binary column, where only the column corresponding to the category is 1 and the rest are 0.

[0192] In a specific implementation, features useful for optimizing the indoor positioning network can also be extracted and constructed based on a standard indoor three-dimensional map or several standard positioning devices arranged at the positions of standard indoor positioning devices. Specifically, features may include but are not limited to:

[0193] Standard environmental characteristics data, such as:

[0194] Indoor space dimensions: Extract the length, width, and height of the indoor environment. These dimensions can directly affect the propagation and coverage of the signal.

[0195] Shape features: If the interior space is irregularly shaped, features can be constructed to describe the complexity of its shape, such as using the ratio of the perimeter to the area of ​​the space.

[0196] Obstacle distribution: Identify and quantify the type, number, and size of obstacles, which have a significant impact on the occlusion and reflection of signal propagation.

[0197] Standard wireless signal data, such as:

[0198] Signal frequency: The frequency of a signal affects its wavelength and penetrating power and can be used as one of its characteristics.

[0199] Signal wavelength: Wavelength is related to frequency and can be used to evaluate the propagation characteristics of signals in different materials.

[0200] Signal power: The power of signal transmission directly affects the coverage and strength of the signal.

[0201] Standard indoor 3D maps, such as:

[0202] Map feature extraction: Extract features from the high-precision indoor three-dimensional map generated by SLAM technology, including wall corners, columns, floor signs, lamps, office desks and chairs, etc. These features can be used as reference points for signal propagation.

[0203] Standard indoor positioning device locations, specific location coordinates of indoor positioning device layout, and beacon spacing: construct the distance characteristics between beacons, the spacing will affect signal overlap and coverage uniformity.

[0204] Signal propagation model simulation results, such as:

[0205] Signal coverage: Based on the signal propagation model, extract the coverage range of each positioning device.

[0206] Signal strength distribution: By analyzing the strength of the signal at different locations, the characteristics of the signal strength distribution can be constructed.

[0207] Signal attenuation: Extract the attenuation rate of the signal during propagation.

[0208] Through the above feature construction steps, a rich and representative feature set can be generated, one or more of which can be used as input to the machine learning model to predict or optimize the layout of the indoor positioning network. These features include not only direct measurements of the original data, but also derived features obtained by analyzing and processing the original data, thereby providing the model with a more comprehensive perspective to understand and solve the indoor positioning network optimization problem.

[0209] In the specific implementation, Python's TensorFlow, TensorFlow Data Validation (TFDV), Pandas, and scikit-learn libraries can be used to process the Pandas DataFrame containing the preliminary layout point data of the indoor environment. First, create a sample DataFrame containing data such as the length and width of the indoor space, signal frequency and power, and obstacle material. Then, convert this Pandas DataFrame into a TensorFlow dataset for further processing and analysis.

[0210] To better understand the dataset, you can use Python’s TFDV library to infer the schema of the dataset and visualize statistics. This is an optional step that helps to better understand the structure and distribution of the data.

[0211] In the data preprocessing step, you can use Python to define a preprocessing function preprocess_data, which is responsible for cleaning missing values, feature normalization, and encoding categorical data. For missing values, use the mean to fill; for numerical features, use StandardScaler for normalization; for categorical features, use one-hot encoding. Finally, convert the processed data back to the TensorFlow dataset and convert it into a format suitable for model training, setting the batch size to 32. Through the above steps, you can get the preprocessed dataset.

[0212] By following these steps, we ensure that the dataset is properly cleaned and transformed before being fed into the neural network model, thereby improving the performance and reliability of the model.

[0213] In an embodiment of the present application, training environment characteristic data, training wireless signal data, training indoor positioning device location and verification environment characteristic data, verification wireless signal data can be obtained based on the pre-processed standard environment characteristic data, standard wireless signal data, and standard indoor positioning device location. Among them, the training environment characteristic data, training wireless signal data, and training indoor positioning device location are part of the standard environment characteristic data, standard wireless signal data, and standard indoor positioning device location, and the verification environment characteristic data and verification wireless signal data are another part of the standard environment characteristic data and standard wireless signal data. It can be understood that the standard data set is divided into a training set and a verification set according to a certain ratio.

[0214] In an embodiment of the present application, the training environment feature data, the training wireless signal data, and the training indoor positioning device position can be used as training samples, the training environment feature data and the training wireless signal data can be used as inputs of the positioning device position prediction model, and the training indoor positioning device position can be used as outputs of the positioning device position prediction model. The positioning device position prediction model is trained using the training samples to obtain a positioning device position prediction model to be verified.

[0215] In an embodiment of the present application, the verification environment feature data and the verification wireless signal data can be used as verification samples, and the verification environment feature data and the verification wireless signal data are input into the positioning device position prediction model to be verified to obtain the verification indoor positioning device position output by the positioning device position prediction model to be verified;

[0216] In an embodiment of the present application, a positioning device position prediction model loss function is calculated based on the training indoor positioning device position and the verification indoor positioning device position, wherein the Adam optimizer and mean square error can be selected as the loss function, and the mean absolute error is added as an evaluation indicator.

[0217] It should be noted that the Adam optimizer (Adaptive Moment Estimation Optimizer) can be an optimization algorithm widely used in the field of deep learning. It combines the advantages of the Momentum and RMSprop algorithms to provide a fast and stable model training convergence process. Adam adjusts the size of the gradient update by calculating the adaptive learning rate of each parameter, which makes it more flexible in dealing with the gradients of different parameters. Specifically, Adam maintains the first-order moment (i.e., the exponentially weighted average of the gradient) and the second-order moment (i.e., the exponentially weighted average of the square of the gradient) of each parameter, and uses these moments to adjust the learning rate of each parameter. This adaptability makes Adam perform well in dealing with sparse gradients and non-stationary targets, so it has been widely used in various deep learning tasks. In an embodiment of the present application, the Adam optimizer can be used to obtain the position prediction model loss function of the positioning device.

[0218] Mean Squared Error (MSE) can be a commonly used loss function, which is widely used in regression problems to measure the difference between the model prediction value and the true value. The mean squared error quantifies the difference by calculating the square of the difference between the predicted value and the true value, and then taking the average of these square values. In an embodiment of the present application, the mean squared error can be used to obtain the loss function of the positioning device position prediction model.

[0219] Mean Absolute Error (MAE) is a commonly used evaluation metric, particularly for regression problems, to measure the average absolute difference between the model's predicted value and the true value. The mean absolute error quantifies the difference by calculating the absolute value of the difference between the predicted value and the true value, and then taking the average of these absolute values. In an embodiment of the present application, the mean absolute error can be used to evaluate the positioning device position prediction model loss function.

[0220] In the embodiment of the present application, the positioning device position prediction model loss function can be used to adjust the model parameters of the positioning device position prediction model to be verified, and the steps of using the training environment feature data and the training wireless signal data as the input of the positioning device position prediction model, using the training indoor positioning device position as the output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model are re-executed until the preset stop condition is reached to obtain the trained positioning device position prediction model. The preset stop condition can be the number of training rounds, which is not limited by the present application.

[0221] This application can effectively obtain the layout of indoor positioning equipment by obtaining standard indoor three-dimensional maps and related data, performing preprocessing and model training. According to standard environmental feature data and wireless signal data, the quality and consistency of the data are improved through preprocessing steps such as data cleaning, missing data filling, normalization and classification coding. Subsequently, through the training and verification process, the model parameters are continuously adjusted until the preset stop condition is reached, thereby obtaining a trained positioning device location prediction model. It can automatically process and analyze large amounts of data, improve the deployment efficiency and accuracy of positioning equipment, and ensure the reliability and accuracy of the prediction results through model training and verification. In addition, it can also adapt to different indoor environments and equipment configurations, and has strong versatility and flexibility.

[0222] In an optional embodiment of the present application, the positioning device position prediction model is composed of a sequential model, and the sequential model includes an input layer, a plurality of hidden layers, and an output layer. Step 103 also includes the following sub-steps:

[0223] S301: Inputting the environmental feature data and the wireless signal data into the input layer for processing to obtain a first positioning device feature output by the input layer;

[0224] S302: Inputting the first positioning device feature into the hidden layer for processing to obtain a second positioning device feature output by the hidden layer;

[0225] S303: Input the second positioning device feature into the output layer for processing to obtain the initial positioning device position output by the output layer.

[0226] It should be noted that the sequential model can be a common model structure in deep learning, especially when using frameworks such as Keras or TensorFlow. This model consists of a series of network layers stacked in sequence, each layer takes the output of the previous layer as input, and gradually processes the data until the final output result. The sequential model is widely used in various neural network architectures because of its simple structure, easy to understand and implement. In an embodiment of the present application, the sequential model can be used to construct a positioning device position prediction model.

[0227] The input layer can be the first layer of the neural network, which is responsible for receiving raw data or features. In a sequential model, the input layer usually does not perform any calculations, but defines the shape and structure of the network input. For example, when processing image data, the input layer may define the width, height, and number of color channels of the image. The main function of the input layer is to provide a clear input interface for the network to ensure that the data can be correctly passed to the subsequent hidden layers. In an embodiment of the present application, the positioning device position prediction model may include an input layer.

[0228] Hidden layers can be network layers located between the input layer and the output layer. They are the core parts of neural networks for complex data processing and feature extraction. Each hidden layer consists of multiple neurons (or nodes), which process input data through activation functions and pass the results to the next layer. The number of hidden layers and the number of neurons in each layer are key parameters when designing a neural network. They determine the depth and width of the network, which in turn affects the learning ability and performance of the model. In an embodiment of the present application, the positioning device position prediction model may include several hidden layers. The specific number of hidden layers can be set according to actual needs.

[0229] The output layer can be the last layer of the neural network, which is responsible for generating the final output of the model. The structure and activation function of the output layer are usually designed according to the type of task. For example, in a classification task, the output layer may contain neurons equal to the number of categories, and use a softmax activation function to generate a probability distribution; in a regression task, the output layer may have only one neuron, and use a linear activation function to directly output the predicted value. The result of the output layer is the final conclusion of the model after processing the input data, and is the basis for evaluating model performance and making decisions. In an embodiment of the present application, the positioning device position prediction model may include an output layer.

[0230] In an embodiment of the present application, environmental feature data and wireless signal data can be input into the input layer for processing to obtain a first positioning device feature output by the input layer, and then the first positioning device feature can be input into the hidden layer for processing to obtain a second positioning device feature output by the hidden layer, and finally the second positioning device feature can be input into the output layer for processing to obtain an initial positioning device position output by the output layer.

[0231] The first positioning device feature and the second positioning device feature may be a multi-dimensional array or matrix that reflects environmental feature data and wireless signal data.

[0232] In the specific implementation, you can first define a Sequential model in Python and add several Dense layers as hidden layers, each of which is followed by a Dropout layer to reduce overfitting. The parameters of the input layer are set to the number of features, and the number of neurons in the output layer is set to the number of beacon positions we predict. The ReLU activation function can be used as the activation function of the hidden layer, and the linear activation function is selected for the output layer to solve the regression problem of this application. Next, compile the model, select the Adam optimizer and mean square error as the loss function, and add the mean absolute error as the evaluation indicator. Finally, use the model.fit method to train the model.

[0233] In an embodiment of the present application, the model can be further optimized by adjusting parameters such as the network structure, number of hidden layers, number of neurons, activation function, optimizer, loss function, etc. in the positioning device position prediction model and combining the test results of the test data.

[0234] In an optional embodiment of the present application, step 104 further includes the following sub-steps:

[0235] S401: Acquire a positioning device signal radius according to preset attribute information of the positioning device;

[0236] S402: Calculating the signal coverage strength in the indoor three-dimensional map according to the initial positioning device position and the positioning device signal radius;

[0237] S403: Calculate the uniformity of the positioning device and the deployment cost of the positioning device according to the initial positioning device position.

[0238] The positioning device signal radius may refer to the maximum distance that the positioning device can effectively cover and provide positioning services. This radius depends on a variety of factors, including the signal transmission power of the device, the receiving sensitivity, the interference and obstacles in the environment, and the type of positioning technology used. In an embodiment of the present application, the positioning device signal radius can be obtained based on the attribute information of the positioning device.

[0239] Signal coverage strength may refer to the signal quality level that a wireless signal can provide in a specific area. It is often used to evaluate the performance of wireless communication systems (such as Wi-Fi, Bluetooth, mobile networks, etc.), especially in positioning and communication applications. In an embodiment of the present application, the signal coverage strength may be determined by calculating the average proportion of predicted coverage that exceeds a threshold.

[0240] The uniformity of the positioning device may refer to the uniformity of the distribution of the positioning device in the target area. A positioning device layout with high uniformity means that the distribution of the devices in space is more balanced, which can ensure more comprehensive signal coverage, reduce signal blind spots, and improve the quality and reliability of positioning services. In an embodiment of the present application, the uniformity of the positioning device can be obtained by calculating the variance of the positioning device position or the density distribution in different areas. Among them, the smaller the variance, the higher the uniformity, and the more uniform the density, the higher the uniformity.

[0241] The positioning device deployment cost may refer to the economic cost and resource consumption required to deploy the positioning device in the target area. The deployment cost includes the cost of the device itself, the installation cost, the maintenance cost, etc. Optimizing the deployment cost means reducing the cost as much as possible while ensuring the quality of the positioning service. In an embodiment of the present application, the positioning device deployment cost can be obtained by calculating the Euclidean distance between the position of each positioning device and the average position, where the smaller the distance, the lower the cost.

[0242] In an embodiment of the present application, the signal radius of the positioning device can be obtained according to the preset attribute information of the positioning device, the signal coverage strength in the indoor three-dimensional map can be calculated according to the initial positioning device position and the positioning device signal radius, and the positioning device uniformity and positioning device deployment cost can be calculated according to the initial positioning device position.

[0243] In an optional embodiment of the present application, step 106 further includes the following sub-steps:

[0244] S501: performing a preset number of random disturbance processing on the initial positioning device position to obtain a plurality of groups of disturbed positioning device positions;

[0245] S502: Calculating the disturbance signal coverage strength, the disturbance positioning device uniformity, and the disturbance positioning device deployment cost in the indoor three-dimensional map according to each group of disturbance positioning device positions;

[0246] S503: Calculating the fitness of the disturbance positioning device corresponding to each group of the disturbance positioning device positions according to the disturbance signal coverage strength, the uniformity of the disturbance positioning device, and the deployment cost of the disturbance positioning device;

[0247] S504: comparing the disturbance positioning device fitness corresponding to several groups of the disturbance positioning device positions, and taking a group of the disturbance positioning device positions corresponding to the maximum disturbance positioning device fitness as the target positioning device positions.

[0248] It should be noted that the random perturbation process can be a random perturbation, which is a commonly used technique in optimization algorithms, especially in local search and heuristic algorithms. Its basic idea is to explore other possible solutions in the solution space by making slight random changes to the current solution in the hope of finding a better solution. In an embodiment of the present application, the perturbation process can be a random slight change to the coordinate value in the two-dimensional position of the initial positioning device.

[0249] In the embodiment of the present application, the initial positioning device position can be subjected to a preset number of random disturbances to obtain several groups of disturbed positioning device positions, and then the disturbance signal coverage strength, the disturbance positioning device uniformity, and the disturbance positioning device deployment cost in the indoor three-dimensional map are calculated according to each group of the disturbed positioning device positions. The specific calculation method can refer to the above-mentioned method for calculating signal coverage strength, positioning device uniformity, and positioning device deployment cost. Among them, the preset number can be set according to the actual situation.

[0250] In the embodiment of the present application, the fitness of the disturbance positioning device corresponding to each group of disturbance positioning device positions can be calculated according to the coverage strength of the disturbance signal, the uniformity of the disturbance positioning device, and the deployment cost of the disturbance positioning device, and then the fitness of the disturbance positioning device corresponding to several groups of disturbance positioning device positions is compared, and a group of disturbance positioning device positions corresponding to the maximum disturbance positioning device fitness is used as the target positioning device position. Among them, the maximum disturbance positioning device fitness can be the maximum value of the disturbance positioning device fitness.

[0251] In the embodiment of the present application, firstly, by obtaining the indoor three-dimensional map and wireless signal data, the environmental feature data is obtained according to the indoor three-dimensional map, which provides more accurate and reliable map data support for the planning of the positioning device point. This method not only reduces the dependence on manual survey, but also improves the rationality of point planning, and effectively avoids problems such as incomplete signal coverage and multipath effect. Secondly, the environmental feature data and wireless signal data are input into the trained positioning device position prediction model for processing, and the initial positioning device position is output. The automatic planning of the positioning device point can be realized in the current indoor three-dimensional map. Through accurate three-dimensional map data and advanced algorithms, automatic planning can more accurately determine the optimal installation position of the positioning device, effectively avoid errors and deviations that may occur in manual planning, and improve the efficiency of positioning device deployment. Finally, the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map are calculated according to the initial positioning device position, and the positioning device fitness is calculated according to the signal coverage strength, positioning device uniformity, and positioning device deployment cost. The initial positioning device position is adjusted according to the positioning device fitness to obtain the target positioning device position. Deep learning network technology can be used to verify and optimize the positioning network signal coverage in indoor environments, which not only ensures the comprehensiveness and uniformity of signal coverage, but also takes into account the economic efficiency of deployment costs, thereby achieving optimal resource allocation and ensuring the reliability and stability of the positioning network.

[0252] Reference Figure 2 , shows a schematic diagram of the structure of a positioning device position planning device provided in an embodiment of the present application, the device comprising:

[0253] Data acquisition module 201, used to acquire indoor three-dimensional map and wireless signal data;

[0254] An environmental feature data extraction module 202 is used to obtain environmental feature data according to the indoor three-dimensional map;

[0255] An initial positioning device position acquisition module 203 is used to input the environmental feature data and the wireless signal data into a trained positioning device position prediction model for processing, and output an initial positioning device position;

[0256] An optimization parameter calculation module 204 is used to calculate the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map according to the initial positioning device position;

[0257] A fitness calculation module 205 is used to calculate the fitness of the positioning device according to the signal coverage strength, the uniformity of the positioning device, and the deployment cost of the positioning device;

[0258] The target positioning device position acquisition module 206 is used to adjust the initial positioning device position according to the positioning device fitness to obtain the target positioning device position.

[0259] In an optional embodiment of the present application, the indoor three-dimensional map is collected by a vehicle, the vehicle includes a laser radar and a camera, and the data acquisition module 201 includes:

[0260] A driving route acquisition module, used to acquire a preset driving route;

[0261] A two-dimensional image data acquisition module, used to collect three-dimensional point cloud data through the laser radar and two-dimensional image data through the camera during driving along the driving route;

[0262] An initial three-dimensional map synthesis module, used for synthesizing an initial three-dimensional map according to the three-dimensional point cloud data and the two-dimensional image data;

[0263] A three-dimensional feature data extraction module is used to extract features from the three-dimensional point cloud data to obtain three-dimensional feature data;

[0264] An indoor three-dimensional map acquisition module, used for processing the initial three-dimensional map according to the three-dimensional feature data to acquire the indoor three-dimensional map;

[0265] The wireless signal data acquisition module is used to acquire the wireless signal data according to the preset attribute information of the positioning device.

[0266] In an optional embodiment of the present application, the environmental feature data extraction module 202 includes:

[0267] A three-dimensional point cloud data processing module, used for acquiring voxel data and original point data according to the three-dimensional point cloud data;

[0268] A candidate region feature acquisition module, used to extract features from the voxel data to obtain candidate region features;

[0269] A key point feature acquisition module is used to extract features from the original point data to obtain key point features;

[0270] The three-dimensional feature data acquisition module is used to determine the three-dimensional feature data according to the candidate area features and the key point features.

[0271] In an optional embodiment of the present application, the positioning device position prediction model is trained in the following manner:

[0272] Acquire a standard indoor three-dimensional map, wherein the standard indoor three-dimensional map includes a plurality of standard positioning devices arranged at positions of standard indoor positioning devices;

[0273] Acquiring standard environmental feature data according to the standard indoor three-dimensional map;

[0274] Acquiring standard wireless signal data according to preset attribute information of the standard positioning device;

[0275] Preprocessing at least one of the standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, wherein the preprocessing is one or more of data cleaning, missing data filling, normalization, and classification coding;

[0276] According to the pre-processed standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, the training environment feature data, the training wireless signal data, the training indoor positioning device position, the verification environment feature data, and the verification wireless signal data are obtained;

[0277] Using the training environment feature data, the training wireless signal data, and the position of the training indoor positioning device as training samples;

[0278] Using the training environment feature data and the training wireless signal data as inputs of the positioning device position prediction model, using the training indoor positioning device position as output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model to obtain a positioning device position prediction model to be verified;

[0279] Using the verification environment feature data and the verification wireless signal data as verification samples;

[0280] Inputting the verification environment feature data and the verification wireless signal data into the positioning device position prediction model to be verified, and obtaining the verification indoor positioning device position output by the positioning device position prediction model to be verified;

[0281] Calculate the positioning device position prediction model loss function according to the training indoor positioning device position and the verification indoor positioning device position;

[0282] The positioning device position prediction model loss function is used to adjust the model parameters of the positioning device position prediction model to be verified, and the steps of using the training environment feature data and the training wireless signal data as the input of the positioning device position prediction model, using the training indoor positioning device position as the output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model are re-executed until a preset stop condition is reached to obtain a trained positioning device position prediction model.

[0283] In an optional embodiment of the present application, the positioning device position prediction model is composed of a sequential model, and the sequential model includes an input layer, a plurality of hidden layers, and an output layer. The initial positioning device position acquisition module 203 includes:

[0284] A first positioning device feature acquisition module, used for inputting the environmental feature data and the wireless signal data into the input layer for processing, and obtaining a first positioning device feature output by the input layer;

[0285] A second positioning device feature acquisition module, used for inputting the first positioning device feature into the hidden layer for processing, and obtaining a second positioning device feature output by the hidden layer;

[0286] The initial positioning device position acquisition submodule is used to input the second positioning device feature into the output layer for processing to obtain the initial positioning device position output by the output layer.

[0287] In an optional embodiment of the present application, the optimization parameter calculation module 204 includes:

[0288] A positioning device signal radius acquisition module, used to acquire the positioning device signal radius according to preset attribute information of the positioning device;

[0289] A signal coverage strength acquisition module, used to calculate the signal coverage strength in the indoor three-dimensional map according to the initial positioning device position and the positioning device signal radius;

[0290] The uniformity and deployment cost acquisition module is used to calculate the uniformity of the positioning device and the deployment cost of the positioning device according to the initial positioning device position.

[0291] In an optional embodiment of the present application, the target positioning device position acquisition module 206 includes:

[0292] A disturbance positioning device position acquisition module is used to perform a preset number of random disturbances on the initial positioning device position to obtain a plurality of groups of disturbance positioning device positions;

[0293] A disturbance positioning device position processing module, used to calculate the disturbance signal coverage strength, the disturbance positioning device uniformity, and the disturbance positioning device deployment cost in the indoor three-dimensional map according to each group of disturbance positioning device positions;

[0294] A disturbance positioning device fitness calculation module, used to calculate the disturbance positioning device fitness corresponding to each group of the disturbance positioning device positions according to the disturbance signal coverage strength, the uniformity of the disturbance positioning device, and the deployment cost of the disturbance positioning device;

[0295] The target positioning device position acquisition submodule is used to compare the disturbance positioning device fitness corresponding to several groups of the disturbance positioning device positions, and take a group of the disturbance positioning device positions corresponding to the maximum disturbance positioning device fitness as the target positioning device position.

[0296] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0297] An embodiment of the present application further provides an electronic device, which may include a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, the positioning device position planning method as described above is implemented.

[0298] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the positioning device position planning method as described above is implemented.

[0299] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0300] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0301] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0302] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0303] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0304] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0305] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0306] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the above elements.

[0307] The above provides a detailed introduction to a positioning device position planning method, device, electronic device and medium. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A positioning device location planning method, characterized in that: The method comprises: Obtain indoor 3D maps and wireless signal data; Acquiring environmental feature data according to the indoor three-dimensional map; Inputting the environmental feature data and the wireless signal data into a trained positioning device position prediction model for processing, and outputting an initial positioning device position; Calculate the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map according to the initial positioning device position; Calculate the adaptability of the positioning device according to the signal coverage strength, the uniformity of the positioning device, and the deployment cost of the positioning device; The initial positioning device position is adjusted according to the positioning device fitness to obtain the target positioning device position.

2. The method according to claim 1, characterized in that The indoor three-dimensional map is collected by a vehicle, the vehicle includes a laser radar and a camera, and the acquisition of the indoor three-dimensional map and wireless signal data includes: Get the preset driving route; In the process of driving along the driving route, the three-dimensional point cloud data is collected by the laser radar, and the two-dimensional image data is collected by the camera; synthesizing an initial three-dimensional map according to the three-dimensional point cloud data and the two-dimensional image data; Performing feature extraction on the three-dimensional point cloud data to obtain three-dimensional feature data; Processing the initial three-dimensional map according to the three-dimensional feature data to obtain the indoor three-dimensional map; The wireless signal data is acquired according to preset attribute information of the positioning device.

3. The method according to claim 2, characterized in that The step of extracting features from the three-dimensional point cloud data to obtain three-dimensional feature data includes: Acquire voxel data and original point data according to the three-dimensional point cloud data; Performing feature extraction on the voxel data to obtain candidate region features; Performing feature extraction on the original point data to obtain key point features; The three-dimensional feature data is determined according to the candidate region features and the key point features.

4. The method according to claim 1, characterized in that The positioning device position prediction model is trained in the following way: Acquire a standard indoor three-dimensional map, wherein the standard indoor three-dimensional map includes a plurality of standard positioning devices arranged at positions of standard indoor positioning devices; Acquiring standard environmental feature data according to the standard indoor three-dimensional map; Acquiring standard wireless signal data according to preset attribute information of the standard positioning device; Preprocessing at least one of the standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, wherein the preprocessing is one or more of data cleaning, missing data filling, normalization, and classification coding; According to the pre-processed standard environment feature data, the standard wireless signal data, and the standard indoor positioning device position, the training environment feature data, the training wireless signal data, the training indoor positioning device position, the verification environment feature data, and the verification wireless signal data are obtained; Using the training environment feature data, the training wireless signal data, and the position of the training indoor positioning device as training samples; Using the training environment feature data and the training wireless signal data as inputs of the positioning device position prediction model, using the training indoor positioning device position as output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model to obtain a positioning device position prediction model to be verified; Using the verification environment feature data and the verification wireless signal data as verification samples; Inputting the verification environment feature data and the verification wireless signal data into the positioning device position prediction model to be verified, and obtaining the verification indoor positioning device position output by the positioning device position prediction model to be verified; Calculate the positioning device position prediction model loss function according to the training indoor positioning device position and the verification indoor positioning device position; The positioning device position prediction model loss function is used to adjust the model parameters of the positioning device position prediction model to be verified, and the steps of using the training environment feature data and the training wireless signal data as the input of the positioning device position prediction model, using the training indoor positioning device position as the output of the positioning device position prediction model, and using the training samples to train the positioning device position prediction model are re-executed until a preset stop condition is reached to obtain a trained positioning device position prediction model.

5. The method according to claim 4, characterized in that The positioning device location prediction model is composed of a sequential model, which includes an input layer, a plurality of hidden layers, and an output layer. The environmental feature data and the wireless signal data are input into the trained positioning device location prediction model for processing, and the initial positioning device location is output, including: Inputting the environmental feature data and the wireless signal data into the input layer for processing to obtain a first positioning device feature output by the input layer; Inputting the first positioning device feature into the hidden layer for processing to obtain a second positioning device feature output by the hidden layer; The second positioning device feature is input into the output layer for processing to obtain the initial positioning device position output by the output layer.

6. The method according to claim 1, characterized in that The calculating, according to the initial positioning device position, the signal coverage strength, the uniformity of the positioning device, and the positioning device deployment cost in the indoor three-dimensional map includes: Acquire a positioning device signal radius according to preset attribute information of the positioning device; Calculate the signal coverage strength in the indoor three-dimensional map according to the initial positioning device position and the positioning device signal radius; The positioning device uniformity and the positioning device deployment cost are calculated according to the initial positioning device position.

7. The method according to claim 6, characterized in that The step of adjusting the initial positioning device position according to the positioning device fitness to obtain the target positioning device position includes: Performing a random perturbation process on the initial positioning device position for a preset number of times to obtain a plurality of groups of perturbed positioning device positions; Calculating the disturbance signal coverage strength, the disturbance positioning device uniformity, and the disturbance positioning device deployment cost in the indoor three-dimensional map according to each group of disturbance positioning device positions; Calculating the fitness of the disturbance positioning device corresponding to each group of the disturbance positioning device positions according to the disturbance signal coverage strength, the uniformity of the disturbance positioning device, and the deployment cost of the disturbance positioning device; The fitness of the disturbance positioning device corresponding to several groups of the disturbance positioning device positions is compared, and a group of the disturbance positioning device positions corresponding to the maximum fitness of the disturbance positioning device is used as the target positioning device position.

8. A positioning device position planning device, characterized in that: The device comprises: A data acquisition module, used to acquire indoor three-dimensional maps and wireless signal data; An environmental feature data extraction module, used to obtain environmental feature data according to the indoor three-dimensional map; An initial positioning device position acquisition module is used to input the environmental feature data and the wireless signal data into a trained positioning device position prediction model for processing, and output an initial positioning device position; An optimization parameter calculation module, used to calculate the signal coverage strength, positioning device uniformity, and positioning device deployment cost in the indoor three-dimensional map according to the initial positioning device position; A fitness calculation module, used to calculate the fitness of the positioning device according to the signal coverage strength, the uniformity of the positioning device, and the deployment cost of the positioning device; The target positioning device position acquisition module is used to adjust the initial positioning device position according to the positioning device fitness to obtain the target positioning device position.

9. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method for positioning device position planning according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the positioning device position planning method according to any one of claims 1 to 7 is implemented.

Images