Robot positioning method and device, and electronic equipment

Through the coordinated work of multiple sensors, matching the subscene feature information and pose information in the pre-constructed database, the problem of low reliability of initial pose positioning of the robot is solved, and higher positioning accuracy and reliability are achieved.

CN118999569APending Publication Date: 2024-11-22NETEASE LINGDONG (HANGZHOU) TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411104165.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

In the prior art, the robot has low reliability in the initial position positioning in an indoor environment, mainly because it relies on a single sensor for positioning, and is susceptible to environmental factors to cause inaccurate scene data.

Method used

Multiple sensors (at least two types) are used to locate the robot's initial pose. By obtaining the scene data collected by multiple sensors, matching the subscene feature information and pose information in the pre-constructed database, filtering the optimal based on preset logic The data determines the initial pose.

Benefits of technology

It improves the reliability and accuracy of the initial positioning of the robot, reduces the possibility of unreliable positioning results caused by inaccurate sensor data, and enhances the positioning ability of the robot in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118999569A_ABST
    Figure CN118999569A_ABST
Patent Text Reader

Abstract

The invention discloses a robot positioning method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring a plurality of scene data collected by a plurality of sensors; for each piece of scene data, acquiring database data matched with the scene data from a pre-constructed database; based on a preset screening logic, screening optimal database data from multiple pieces of database data corresponding to the multiple pieces of scene data; and according to the pose information included in the optimal database data, the initial pose of the robot in the operation scene is determined. According to the method, the initial pose of the robot is positioned through a plurality of sensors, so that the reliability of positioning the initial pose of the robot based on the sensors of the robot is improved, and the problem that the initial pose of the robot is positioned based on a single sensor in the prior art is solved; and the reliability of an initial pose positioning result is low.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a robot positioning method, device, electronic device, and computer-readable storage medium. Background Art

[0002] With the rapid development of artificial intelligence technology, the application of robots in various industries has become increasingly extensive and in-depth. When robots perform tasks in complex environments, an important challenge they face is how to achieve high-precision positioning, and a prerequisite for achieving high-precision positioning is to obtain a reliable initial position of the robot.

[0003] In the prior art, the initial posture of the robot is often located in an outdoor environment by using GPS (Global Positioning System), while in an indoor environment, it relies on the deployed UWB (Ultra WideBand) or specific markers for positioning. However, since GPS cannot be used in indoor environments and the deployment of UWB or specific markers will cause a large deployment cost, some technologies provide methods for initial posture positioning using the robot's own sensors. At present, this method only involves the use of a single sensor, that is, the initial posture of the robot can only be positioned based on the scene data obtained by a single sensor, which will inevitably lead to inaccurate scene data obtained based on the sensor due to the influence of environmental factors, thereby resulting in unreliable initial posture positioning results.

[0004] Therefore, the prior art has a technical problem that the reliability of the initial posture positioning result is low because the initial posture positioning of the robot is performed based on a single sensor. Summary of the invention

[0005] The present application provides a robot positioning method, device, electronic device, and computer-readable storage medium to solve the technical problem in the prior art that the initial posture positioning result has low reliability due to the positioning of the robot's initial posture based on a single sensor.

[0006] In a first aspect, an embodiment of the present application provides a robot positioning method, wherein the robot is configured with multiple sensors, the multiple sensors include at least two sensor types, and the robot is located in a working scene, and the method includes: obtaining multiple scene data collected by the multiple sensors; for each of the scene data, obtaining database data matching the scene data from a pre-built database, the database corresponding to the sensor that collects the scene data, and including multiple database data, each of the database data corresponding to a sub-scene in the working scene, including feature information corresponding to the sub-scene and posture information of the robot in the sub-scene; based on a preset screening logic, screening the optimal database data from the multiple database data corresponding to the multiple scene data; determining the initial posture of the robot in the working scene according to the posture information included in the optimal database data.

[0007] In a second aspect, an embodiment of the present application provides a positioning device for a robot, wherein the robot is configured with multiple sensors, wherein the multiple sensors include at least two sensor types, and the robot is located in a working scene, and the device includes: a first data acquisition unit, a second data acquisition unit, a data screening unit, and a posture determination unit; the first data acquisition unit is used to acquire multiple scene data collected by the multiple sensors; the second data acquisition unit is used to acquire, for each of the scene data, database data matching the scene data from a pre-built database, wherein the database corresponds to the sensor that collects the scene data, and includes multiple database data, each of which corresponds to a sub-scene in the working scene, including feature information corresponding to the sub-scene and the posture information of the robot in the sub-scene; the data screening unit is used to screen the optimal database data from the multiple database data corresponding to the multiple scene data based on a preset screening logic; the posture determination unit is used to determine the initial posture of the robot in the working scene according to the posture information included in the optimal database data.

[0008] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory and a processor; the memory is used to store one or more computer instructions; the processor is used to execute the one or more computer instructions to implement the above method.

[0009] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having one or more computer instructions stored thereon, which, when executed by a processor, executes the above method.

[0010] Compared with the prior art, the robot positioning method provided in the present application includes: first, obtaining multiple scene data collected by multiple sensors configured on the robot; secondly, for each scene data, obtaining database data matching it from a pre-constructed database (the database corresponds to the sensor type of the sensor that collects the scene data, and includes multiple database data constructed based on the sensor) (the database data corresponds to a sub-scene in the working scene, and includes feature information corresponding to the sub-scene and the robot's posture information in the sub-scene); again, based on a preset screening logic, screening out the optimal database data from the multiple database data corresponding to the multiple scene data; finally, determining the initial posture of the robot in the working scene according to the posture information included in the optimal database data. This method uses multiple sensors to locate the initial posture of the robot. On the one hand, since the possibility of multiple sensors being affected by environmental factors at the same time is small, the possibility of unreliable initial posture positioning results due to inaccurate scene data obtained by the sensors is greatly reduced, and the reliability of the robot's initial posture positioning based on the robot's own sensors is improved. On the other hand, since multiple sensors can collect multiple scene data at the same time, the accuracy and reliability of the robot's initial posture determined based on multiple scene data will be higher than that determined based on a single scene data. The robot positioning method provided in the present application improves the reliability of the robot's initial posture positioning based on the robot's own sensors, and solves the technical problem of low reliability of the initial posture positioning result due to the positioning of the robot's initial posture based on a single sensor in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is an application system diagram of the robot positioning method provided in an embodiment of the present application;

[0012] Figure 2 is a flow chart of a robot positioning method provided in the first embodiment of the present application;

[0013] Figure 3 is a schematic diagram of the positioning process of the initial posture of the robot provided in the first embodiment of the present application;

[0014] Figure 4 is a schematic diagram of the robot posture positioning process provided by the first embodiment of the present application;

[0015] Figure 5 is a schematic structural diagram of a robot positioning device provided in a second embodiment of the present application;

[0016] Figure 6 It is a schematic diagram of the structure of an electronic device provided in the third embodiment of the present application. DETAILED DESCRIPTION

[0017] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0018] With the rapid development of artificial intelligence technology, the application of robots in various industries has become increasingly extensive and in-depth. As intelligent entities that replace human labor for distribution, detection, labor, and service, robots have gradually appeared in complex operating scenarios such as warehousing, logistics, factories, and hotels. When robots perform operations in complex environments, an important challenge they face is how to achieve high-precision positioning, and a prerequisite for achieving high-precision positioning is to obtain a reliable initial position of the robot.

[0019] In the prior art, the initial posture of the robot is often located by GPS in outdoor environments, while in indoor environments, it relies on the deployed UWB or specific markers for positioning. However, since GPS cannot be used in indoor environments and the deployment of UWB or specific markers will cause a large deployment cost, some technologies provide a method for using the robot's own sensor to locate the initial posture. At present, this method only involves the use of a single sensor, that is, the initial posture of the robot can only be located based on the scene data obtained by a single sensor. Because different sensors have their own characteristics, for example, it is easier to extract features based on scene pictures obtained by camera sensors, but the camera sensor itself is easily affected by light. For example, laser sensors are generally not affected by light, but fewer features can be extracted from the scene data obtained by laser sensors. Therefore, the positioning of the initial posture of the robot based on a single sensor will inevitably result in inaccurate scene data obtained by the sensor due to the influence of environmental factors, resulting in unreliable initial posture positioning results.

[0020] Therefore, the prior art has a technical problem that the reliability of the initial posture positioning result is low because the initial posture positioning of the robot is performed based on a single sensor.

[0021] In view of this, the present application provides a robot positioning method, which uses multiple sensors to locate the initial posture of the robot. On the one hand, since the possibility of multiple sensors being affected by environmental factors at the same time is small, the possibility of unreliable initial posture positioning results due to inaccurate scene data obtained by the sensors is greatly reduced, and the reliability of locating the initial posture of the robot based on the robot's own sensors is improved. On the other hand, since multiple sensors can collect multiple scene data at the same time, the accuracy and reliability of the robot's initial posture determined based on multiple scene data will be higher than that determined based on a single scene data. Based on this, the robot positioning method provided by the present application improves the reliability of the robot's initial posture positioning based on the robot's own sensors.

[0022] In addition, the existing methods for initial posture positioning based on the robot's own sensors have some other problems, such as: the database required for positioning the initial posture cannot be updated in real time, so when the robot's operating scene changes or the database is not fully covered, it may cause the robot's initial posture to be unable to be positioned. Another example is: the robot cannot identify positioning anomalies during operation, which causes the robot's running trajectory to deviate when the robot is kidnapped or the robot's sensor is abnormal. Based on this, in some embodiments of the robot positioning method provided in the application, optional solutions are also provided for the above problems.

[0023] The robot positioning method, device, electronic device, and computer-readable storage medium described in the present application are further described in detail below in conjunction with specific embodiments and drawings.

[0024] Figure 1 : is an application system diagram of the robot positioning method provided in the embodiment of the present application. Figure 1 As shown, the system includes a robot end 101 and a server end 102. The robot end 101 includes a robot and multiple sensors deployed thereon. The multiple sensors capture scene information of the working scene based on different methods. The server end 102 can be a processing device electrically connected to the robot end 101, or a server connected to multiple robot ends 101 in communication, on which the robot positioning method provided by the present application is deployed. When the robot end 101 transmits the scene information captured by multiple sensors to the server end 102, the server end 102 locates the initial posture of the robot based on the method.

[0025] The first embodiment of the present application provides a robot positioning method, based on which the initial posture of the robot in the working scene can be positioned. Of course, the method provided in this embodiment can also be used to position the real-time posture of the robot in the process of performing the working task, that is, based on the method provided in this embodiment, the posture positioning result of the robot at each working moment is obtained.

[0026] The posture includes two key information: position and attitude. The position refers to the three-dimensional coordinates of the robot in the reference coordinate system, usually represented by (x, y, z); the attitude refers to the orientation of the robot relative to the reference coordinate system, which can usually be represented by rotation matrix, Euler angle, quaternion, etc., which covers the rotation information of the robot around three axes (usually the x-axis, y-axis, and z-axis of the coordinate system).

[0027] The initial posture can be understood as the initial position and initial posture of the robot when it is started but has not yet performed the task. Before the robot performs the task, obtaining an accurate and reliable initial posture is crucial for real-time positioning during the operation process. It can promote the positioning algorithm to converge to an accurate positioning result more quickly and improve the positioning efficiency of the robot during the operation.

[0028] In a specific implementation, the robot described in this embodiment is equipped with multiple sensors, and the multiple sensors include at least two sensor types.

[0029] The sensor can be understood as a sensing device installed on the robot for sensing the surrounding environment, its own state or interacting with the work object. The sensor is the main way for the robot to obtain information, which is equivalent to the robot's sense organs, enabling the robot to understand and adapt to its operating environment. At present, there are many types of sensors, such as visual sensors, optical sensors, mechanical sensors, tactile sensors, auditory sensors, etc. Different sensors have different working principles and collect different scene data. For example: visual sensors such as cameras, which obtain scene pictures of the work scene by taking pictures of the work scene, and optical sensors such as laser radar sensors, which use laser beams to measure distance, shape, speed and other information to achieve the effect of three-dimensional scanning and modeling of the work scene, and the obtained scene data is the point cloud information corresponding to the work scene. Because different sensors have their own characteristics, each type of sensor is affected by environmental factors and other factors. Therefore, in the method provided in this embodiment, the multiple sensors configured on the robot must include at least two different types of sensors to reduce the probability of inaccurate scene data collected due to the influence of the sensor, thereby affecting the initial positioning result of the robot.

[0030] It should be noted that the method provided in this embodiment does not limit the specific number of each type of sensor, and the type of sensor configured for the robot and the specific number of each type of sensor can be determined according to the complexity of the operation scene, the refinement of the operation task, etc. Of course, as the operation scene and the operation task change, the type and number of sensors configured for the robot can also be dynamically changed (e.g., added or deleted), which will not affect the implementation of the robot positioning method provided in this embodiment.

[0031] Figure 2 : is a flow chart of the robot positioning method provided by this embodiment. Figure 2 The robot positioning method provided in this embodiment is described in detail. The embodiments described below are used to explain the technical solution of this application and are not intended to be used as limitations for actual use.

[0032] like Figure 2 As shown, the robot positioning method provided in this embodiment includes the following steps S210 to S240:

[0033] Step S210, acquiring multiple scene data collected by multiple sensors.

[0034] Step S220, for each scene data, obtain database data matching the scene data from a pre-built database, the database corresponds to the sensor that collects the scene data, and includes multiple database data, each database data corresponds to a sub-scene in the working scene, including feature information corresponding to the sub-scene and the posture information of the robot in the sub-scene.

[0035] Step S230, based on a preset screening logic, screening the best database data from the multiple database data corresponding to the multiple scene data.

[0036] Step S240, determining the initial posture of the robot in the working scene according to the posture information included in the optimal database data.

[0037] In a specific implementation, the above method is performed by Figure 1The server 102 shown in the figure is executed. After the robot is started in the working scene, the robot collects the scene data of the sub-scene in which the robot is currently located based on the sensors configured by the robot. Since the robot is configured with multiple sensors, the scene data collected by the robot based on the multiple sensors also include multiple. After the robot collects multiple scene data, the multiple scene data are transmitted to the server 102. After the server 102 obtains the multiple scene data, it will obtain the database data matching the scene data from the database for each scene data. After the server 102 obtains the database data corresponding to each scene data, it will filter out the optimal database data from the multiple database data based on the filtering logic, and use the posture information included in the optimal database data as the posture information corresponding to the initial posture of the robot in the working scene.

[0038] The following is a detailed description of the above steps:

[0039] For step S210:

[0040] In this step, the server will obtain multiple scene data collected by multiple sensors.

[0041] The acquisition of multiple scene data can also be understood as the robot side completing the collection of scene data through sensors and then sending the scene data to the server side.

[0042] The scene data can be understood as the original collected and unprocessed state information of the working scene, which directly reflects the complete state of a certain moment or a certain sub-scene in the working scene. Different sensors collect different types of scene data, for example: the scene data collected by the lidar sensor is point cloud data, the scene data collected by the camera sensor is pictures, the scene data collected by the tactile sensor is the contact pressure between the robot and the environment, and the scene data collected by the auditory sensor is sound signals.

[0043] For example, the robot is equipped with two camera sensors (camera sensor A and camera sensor B) and a laser radar sensor. Then, the multiple scene data collected by the robot may include: scene picture 1 collected by camera sensor A, scene picture 2 collected by camera sensor B, and scene point cloud data collected by the laser radar sensor. After the robot completes the collection of scene data based on the sensor, it will send scene picture 1, scene picture 2, and scene point cloud data to the server.

[0044] For step S220:

[0045] In this step, the server will obtain database data matching the scene data from the pre-built database for each scene data.

[0046] The database is a pre-constructed data set, and in this embodiment, the data included in the database is defined as database data. Specifically, the database includes multiple database data, and each database data includes feature information corresponding to a sub-scene in the operating scene and the posture information of the robot in the sub-scene. It should be noted that the database in this embodiment has a one-to-one correspondence with the sensor, that is, each sensor configured by the robot has its corresponding database. In this step, obtaining database data that matches the scene data from the database specifically refers to obtaining database data that matches the scene data from the database corresponding to the sensor that collects the scene data.

[0047] Exemplarily, the robot is equipped with two camera sensors (camera sensor A and camera sensor B) and a laser radar sensor, and the multiple scene data acquired by the server include scene picture 1 acquired by camera sensor A, scene picture 2 acquired by camera sensor B, and scene point cloud data acquired by the laser radar sensor. Then, the server will acquire database data matching scene picture 1 from database a corresponding to camera sensor A for scene picture 1, acquire database data matching scene picture 2 from database b corresponding to camera sensor B for scene picture 2, and acquire database data matching scene point cloud data from database c corresponding to the laser radar sensor for scene point cloud data.

[0048] In an optional implementation, for each scene data, obtaining database data matching the scene data from a pre-built database may specifically include the following steps S221 to S223:

[0049] Step S221, extracting features from the scene data to obtain feature information corresponding to the scene data.

[0050] Step S222, searching the database for database data corresponding to feature information having the highest similarity to feature information corresponding to the scene data.

[0051] Step S223 , taking the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data as the database data matching the scene data.

[0052] Since scene data is the original, unprocessed complete information, it is often very large and complex, and is high-dimensional data, which contains a lot of detailed content, such as: image pixel values, video frame sequences, sensor readings, sound waveforms, geographic location coordinates, etc., and may also include a lot of noise and irrelevant information. If scene data is directly used as a processing object, it will lead to a more complicated calculation process and even have an adverse effect on the final calculation result.

[0053] Based on this, in the method provided in this embodiment, feature information is extracted from the scene data based on some feature extraction algorithms and methods, such as: trained neural networks, PCA (Principal Component Analysis) dimensionality reduction methods, etc. The process of extracting feature information can be regarded as a process of compressing and optimizing scene data. The extracted feature information is an abstraction and simplification of the scene data. It is the most essential and key information that can best characterize the state of the work scene in the scene data. Compared with scene data, the extracted feature information has a lower dimension and is easier to process and understand. The method of extracting feature information has been used in some technologies, and the detailed process of feature extraction is not described in this embodiment.

[0054] After obtaining the feature information corresponding to each scene data, the feature information with the highest similarity to the feature information corresponding to the scene data is retrieved from the corresponding database. In this embodiment, the database data corresponding to the feature information with the highest similarity to the feature information corresponding to the scene data is used as the database data matching the scene data.

[0055] Exemplarily, it is assumed that feature information 1 is extracted from scene image 1 based on a pre-trained feature extraction model and a PCA dimensionality reduction method, where scene image 1 is scene data collected by camera sensor A. It is assumed that database a corresponding to camera sensor A includes 10 database data. Feature information with the highest similarity to feature information 1 is retrieved from database a according to feature information 1. It is assumed that the feature information with the highest similarity to feature information 1 in the retrieved database a is feature information n, and feature information n is feature information included in database data N. Therefore, database data N is database data matching scene image 1.

[0056] In a specific implementation, the feature information is represented in the form of a vector. Retrieving the database data corresponding to the feature information with the highest similarity to the feature information corresponding to the scene data may specifically include the following steps S222-1 to S222-3:

[0057] Step S222-1, performing vector product of feature information corresponding to the scene data and feature information included in each database data in the database.

[0058] Step S222-2, using the value of the vector product as the retrieval score of the feature information included in the database data.

[0059] Step S222 - 3 , taking the database data corresponding to the feature information with the highest retrieval score as the database data corresponding to the feature information with the highest similarity to the feature information corresponding to the scene data.

[0060] In this implementation, the feature information extracted from the scene data is represented in the form of a vector. The vector is a quantity with size and direction, which can be regarded as an ordered set of numbers, usually represented as an array of one column or one row. In this embodiment, the vector representing the feature information is a set of feature data. The similarity between the feature information is determined by performing a vector product (i.e., dot product) on the vector of the feature information extracted from the scene data and the vector of the feature information included in the database data in the database. Specifically, the larger the value of the vector product, the smaller the difference between the vectors corresponding to the two feature information, and the higher the similarity between the two feature information. The smaller the value of the vector product, the greater the difference between the vectors corresponding to the two feature information, and the lower the similarity between the two feature information.

[0061] In this embodiment, the value of the vector product is defined as a search score. Since the feature information corresponding to the scene data is vector-multiplied with the feature information included in each database data in the database, the obtained vector product is used as the search score of the feature information included in the database data. According to the size of the search score, the feature information in the database that is most similar to the feature information corresponding to the scene data can be determined, thereby obtaining the database data matching the scene data.

[0062] Exemplarily, it is assumed that feature information 1 is extracted from scene image 1 based on a pre-trained feature extraction model and a PCA dimensionality reduction method, where scene image 1 is scene data collected by camera sensor A. It is assumed that database a corresponding to camera sensor A includes 10 database data. The vector corresponding to feature information 1 is vector-multiplied with the vectors corresponding to the feature information included in the 10 database data, and the value of the vector product is used as the retrieval score corresponding to the feature information included in the database data. It is assumed that the retrieval score of feature information n is the highest, then feature information n is the feature information with the highest similarity to feature information 1, and feature information n is the feature information included in database data N. Therefore, database data N is the database data corresponding to the feature information with the highest similarity to feature information 1, that is, the database data matching scene image 1.

[0063] For step S230:

[0064] In this step, the server will filter the best database data from multiple database data corresponding to multiple scene data based on the preset filtering logic.

[0065] Through the above steps, the database data corresponding to the scene data collected by each sensor is obtained. It can be understood that the number of database data will be obtained at this time according to the number of sensors configured on the robot. In this step, these database data will be further processed to screen out the optimal database data. The optimal database data can be understood as the posture information included in the database data that best matches the current posture of the robot in the working scene. When the robot has not yet performed the working task, the current posture is the initial posture of the robot. That is, through the optimization of this step, the credibility of the initial posture is further increased.

[0066] Exemplarily, after the server obtains database data matching scene image 1 (defined as database data 1), database data matching scene image 2 (defined as database data 2), and database data matching scene point cloud data (defined as database data 3), it will filter out the optimal database data from database data 1, database data 2, and database data 3 based on the filtering logic.

[0067] The screening logic refers to a preset method for screening out the best database data from multiple database data. In an optional implementation, the screening logic is an optimal screening based on the posture information included in each database data obtained, which may specifically include the following steps S231 to S232:

[0068] Step S231, inputting the posture information included in the multiple database data into a preset positioning algorithm.

[0069] Step S232, determining the optimal database data according to the convergence status of the positioning algorithm, where the convergence status is used to characterize the robustness of the positioning algorithm when iteratively calculating the current posture of the robot based on the posture information.

[0070] The positioning algorithm described in this embodiment can be any algorithm currently used for robot posture positioning, and is not limited here. The posture information included in each database data is used as output data and input into the positioning algorithm. The positioning algorithm will perform iterative calculations based on the posture information. In this embodiment, the convergence status of the iterative calculation process of the positioning algorithm is used as the screening basis for screening the optimal database data.

[0071] The convergence can be understood as the process in which the data generated by the algorithm gradually approaches a stable value as the number of iterations increases. The convergence condition refers to the degree and characteristics of the data output by the algorithm gradually stabilizing as the number of iterations increases, and is usually determined based on the results output by the algorithm in several adjacent iteration rounds. In this embodiment, the convergence condition of the positioning algorithm can characterize the robustness of the positioning algorithm when iteratively calculating the current posture of the robot based on the posture information, that is, the convergence condition is used as an evaluation algorithm to stably find a reasonable posture based on different posture information. A good convergence condition indicates that the positioning algorithm can stably converge to a reliable posture positioning result based on the posture information, and a poor convergence condition indicates that it is difficult or impossible for the positioning algorithm to obtain a reliable posture positioning result based on the posture information.

[0072] It can be understood that inputting the posture information included in the database data into the positioning algorithm to calculate the current posture of the robot is actually a process of using the posture information as the posture information of the initial posture of the robot to obtain the posture information of the current posture of the robot. If the convergence of the positioning algorithm is good after the posture information included in a certain database data is input into the positioning algorithm, it means that the positioning algorithm can converge to a reliable posture result by using the posture information in the database data as the posture information of the initial posture. If the convergence of the positioning algorithm is not good after the posture information included in a certain database data is input into the positioning algorithm, it means that the positioning algorithm cannot converge to a reliable posture result by using the posture information in the database data as the posture information of the initial posture. Based on this, the convergence state that can be achieved by comparing the posture information in each database data through the positioning algorithm can be realized to filter out the optimal database data from multiple database data.

[0073] In a specific implementation, the input data input into the positioning algorithm includes, in addition to the posture information included in the database data, feature information corresponding to the scene data of the current working scene collected by the sensor. That is, the positioning algorithm calculates the posture information corresponding to the current posture of the robot based on the posture information included in the database data (equivalent to the posture information of the initial posture) and the feature information corresponding to the current working scene.

[0074] Exemplarily, it is assumed that the preset positioning algorithm is a laser positioning algorithm, which is a method for determining the position and posture of the robot by calculating the distance and angle information between the laser transmitter and the receiver (or the reflected laser signal). Specifically, the feature information in the scene data obtained by the laser radar sensor and the posture information of the initial posture of the robot are used as input data, and the posture information of the current posture of the robot is output after iterative calculation. After the server obtains the database data matching the scene image 1 (defined as database data 1), the database data matching the scene image 2 (defined as database data 2), and the database data matching the scene point cloud data (defined as database data 3), the posture information 1 included in the database data 1 and the point cloud information corresponding to the scene point cloud data, the posture information 2 included in the database data 2 and the point cloud information corresponding to the scene point cloud data, and the posture information 3 included in the database data 3 and the point cloud information corresponding to the scene point cloud data are respectively input into the laser positioning algorithm, and according to the convergence status of the laser positioning algorithm in the iterative operation process, one is selected from the database data 1, the database data 2, and the database data 3 as the optimal database data.

[0075] In a specific implementation, the convergence status includes inputting the posture information into the positioning algorithm, and whether the positioning algorithm converges during the iterative calculation process, that is, the convergence status is defined by whether the positioning algorithm converges, which is used as a screening indicator for the optimal database data. Specifically, the following steps S232-11 to S232-14 may be included:

[0076] Step S232-11, inputting the posture information included in the first database data into the positioning algorithm, the first database data being the one with the highest matching degree with the scene data among the multiple database data.

[0077] Step S232-12: in response to the convergence of the positioning algorithm, determining the first database data as the optimal database data.

[0078] Step S232-13, in response to the positioning algorithm not converging, inputting the posture information included in the second database data into the positioning algorithm according to the degree of matching between the second database data and the scene data, the second database data being at least one of the multiple database data whose degree of matching with the scene data is lower than that of the first database data.

[0079] Step S232-14: in response to the convergence of the positioning algorithm, determining the second database data when the positioning algorithm converges as the optimal database data.

[0080] In this implementation, the posture information included in the database data is input into the positioning algorithm in the order of the degree of matching between the database data and the scene data from large to small, and the optimal database data is determined according to whether the algorithm converges. Whether the algorithm converges can be determined based on the data output by the positioning algorithm in several adjacent iteration rounds. If the data output in several iteration rounds is stable, it means that the algorithm converges. If the data output in several iteration rounds is unstable, it means that the algorithm does not converge. Usually, a threshold of the number of iterations and an indicator value for judging whether the data is stable are set, such as a difference threshold corresponding to the data difference. Specifically, when the number of iterations of the algorithm reaches or does not reach the iteration threshold, the difference between the data output by the algorithm is less than or equal to the difference threshold, indicating that the algorithm converges. When the number of iterations of the algorithm reaches the iteration threshold, the difference between the data output by the algorithm is still greater than the difference threshold, indicating that the algorithm does not converge.

[0081] In a specific implementation, the degree of matching between the database data and the scene data can be represented by the retrieval score obtained when determining the database data matching the scene data. The higher the retrieval score, the higher the similarity between the feature information of the matching database data and the feature information of the scene data, and the higher the degree of matching between the database data and the scene data. Therefore, the posture information included in the database data with the highest degree of matching (i.e., the highest retrieval score) is first input into the positioning algorithm. If the positioning algorithm converges, then the database data is determined as the optimal database data, and the other database data will no longer be processed. In this embodiment, the database data with the highest degree of matching is defined as the first database data. If the positioning algorithm does not converge, it means that the positioning algorithm cannot obtain a stable posture positioning result based on the posture information in the first database data. Then, the posture information included in the other database data with a matching degree lower than the first database data is sequentially input into the positioning algorithm according to the size of the matching degree (i.e., the size of the retrieval score) until the positioning algorithm converges. The database data when the positioning algorithm converges is determined as the optimal database data. In this embodiment, the other database data with a matching degree lower than the first database data is defined as the second database data. Of course, there may be a situation where all the database data cannot make the positioning algorithm converge. In this case, it is judged that the initial posture positioning has failed and re-positioning is required.

[0082] Exemplarily, assuming that the preset positioning algorithm is a laser positioning algorithm, the retrieval scores corresponding to the database data (defined as database data 1) matching the scene image 1, the database data (defined as database data 2) matching the scene image 2, and the database data (defined as database data 3) matching the scene point cloud data obtained by the server are 10, 30, and 20, respectively. First, since the retrieval score corresponding to database data 2 is the highest, that is, the degree of matching between database data 2 and scene image 2 is the highest, therefore, the pose information 2 included in database data 2 and the point cloud information corresponding to the scene point cloud data are input into the laser positioning algorithm. Assuming that the laser positioning algorithm has not converged, it is shown that the pose information 2 cannot enable the laser positioning algorithm to converge to a stable pose positioning. Secondly, since the retrieval score corresponding to database data 3 is higher than the retrieval score corresponding to database data 1, therefore, the pose information 3 included in database data 3 and the point cloud information corresponding to the scene point cloud data are input into the laser positioning algorithm. Assuming that the laser positioning algorithm converges, it is shown that the pose information 3 can enable the laser positioning algorithm to converge to a stable pose positioning. Therefore, database data 3 is determined as the optimal database data, and database data 1 is no longer processed.

[0083] In another specific implementation, in addition to outputting the position information, the positioning algorithm can also output the convergence score during the iterative operation. Based on this, the convergence status can also include inputting the position information into the positioning algorithm and outputting the convergence score during the iterative calculation of the positioning algorithm, that is, using the convergence score as the screening index for the optimal database data. Specifically, the following steps S232-21 to S232-22 can be included:

[0084] Step S232-21, inputting the posture information included in each database data into the positioning algorithm to obtain a convergence score corresponding to each database data;

[0085] Step S232-22: taking the database data with the largest convergence score as the optimal database data.

[0086] In this implementation, all the posture information included in the database data will be input into the positioning algorithm, and the optimal database data will be determined based on the convergence score output by the positioning algorithm. The convergence score can be understood as an indicator that quantifies the degree to which the algorithm iteration process tends to the optimal solution. The higher the score, the faster the algorithm reaches the expected convergence state (i.e., the fewer iterations), the more stable the iteration process, and the better the robustness of the algorithm. The lower the score, the more iterations the algorithm needs to reach the expected convergence state, the greater the fluctuations in the iteration process, and the poorer the robustness of the algorithm. Based on this, by comparing the convergence scores corresponding to multiple database data, the optimal database data can be screened out.

[0087] Exemplarily, assuming that the preset positioning algorithm is a laser positioning algorithm, the retrieval scores of the database data (defined as database data 1) matching the scene image 1, the database data (defined as database data 2) matching the scene image 2, and the database data (defined as database data 3) matching the scene point cloud data obtained by the server are 10, 30, and 20, respectively. The point cloud information corresponding to the pose information 1 and the scene point cloud data included in the database data 1, the point cloud information corresponding to the pose information 2 and the scene point cloud data included in the database data 2, and the point cloud information corresponding to the pose information 3 and the scene point cloud data included in the database data 3 are respectively input into the laser positioning algorithm. Assuming that the convergence scores output by the laser positioning algorithm are 0.7, 0.9, and 0.3, respectively, then, since the convergence score corresponding to the database data 2 is the highest, the database data 2 is determined as the optimal database data.

[0088] Compared with the above two implementation methods of determining the optimal database data, the first implementation method can complete the screening of the optimal database data more quickly, and the second implementation method can complete the screening of the optimal database data more accurately. In the actual application process, you can choose to deploy according to your needs.

[0089] For step S240:

[0090] In this step, the server will determine the initial posture of the robot in the working scene based on the posture information included in the optimal database data.

[0091] The optimal database data is obtained through the above steps. The feature information included in the database data is the one with the highest similarity with the feature information corresponding to the scene data among the multiple feature information included in the database. The posture information included in the database data is the one that can make the positioning algorithm converge or obtain the highest convergence score. The initial posture of the robot is determined according to the posture information included in the optimal database data, specifically, the posture information included in the optimal database data is used as the posture information corresponding to the initial posture of the robot.

[0092] Exemplarily, database data 3 is the optimal database data, then the posture information 3 included in the database data 3 is the posture information corresponding to the initial posture of the robot in the working scene, and the posture information 3 can be further converted into the initial posture by converting the data format. The specific method of data format conversion can be defined and adjusted according to the coordinate system and the expected format requirements, and is not limited or specifically described here.

[0093] The above steps provide a variety of optional implementations of the robot positioning method. Through the above steps, a reliable and accurate initial position of the robot can be obtained based on multiple sensors of the robot itself.

[0094] The following is an explanation using a specific example, in which it is assumed that the robot is equipped with two camera sensors (camera sensor A and camera sensor B) and a lidar sensor.

[0095] Figure 3 It is a schematic diagram of the positioning process of the initial posture of the robot provided in this embodiment.

[0096] like Figure 3 The robot's initial posture positioning process may include the following steps S301 to S305:

[0097] Step S301, obtaining scene data of the working scene collected by the robot based on the sensor.

[0098] Specifically, scene picture 1 collected by camera sensor A, scene picture 2 collected by camera sensor B, and scene point cloud data collected by a lidar sensor are obtained.

[0099] Step S302: extract features from the scene data to obtain feature information corresponding to the scene data.

[0100] Specifically, feature extraction is performed on scene image 1, scene image 2, and scene point cloud data to obtain feature information 1 corresponding to scene image 1, feature information 2 corresponding to scene image 2, and feature information 3 corresponding to the scene point cloud data.

[0101] Step S303, retrieving database data corresponding to feature information having the highest similarity to feature information corresponding to the scene data from the database, and taking the database data corresponding to the feature information having the highest similarity as database data matching the scene data.

[0102] Specifically, based on feature information 1, feature information 2, and feature information 3, feature information with the highest similarity to feature information 1, feature information 2, and feature information 3 is retrieved from database 1 corresponding to camera sensor A, database 2 corresponding to camera sensor B, and database 3 corresponding to lidar sensor. Assuming that the feature information with the highest similarity to feature information 1, feature information 2, and feature information 3 are feature information a, feature information b, and feature information c, respectively, then the database data matching scene image 1 is database data a corresponding to feature information a, the database data matching scene image 2 is database data b corresponding to feature information b, and the database data matching scene point cloud data is database data c corresponding to feature information c.

[0103] Step S304: based on a preset positioning algorithm, the best database data is selected from a plurality of database data matching the scene data.

[0104] Specifically, assuming that the similarity between feature information 1 and feature information a is 90%, the similarity between feature information 2 and feature information b is 80%, and the similarity between feature information 3 and feature information c is 70%. The pose information a corresponding to feature information a, the pose information b corresponding to feature information b, and the pose information c corresponding to feature information c are respectively input into the preset positioning algorithm, and the obtained convergence scores are 0.6, 0.9, and 0.2 respectively. Since the convergence score corresponding to pose information b is the highest, the database data b corresponding to pose information b is used as the optimal database data;

[0105] Step S305: Using the posture information included in the optimal database data as the posture information corresponding to the initial posture of the robot.

[0106] Specifically, the database data b is used as the optimal database data, and therefore the posture information b is used as the posture information corresponding to the initial posture of the robot.

[0107] In an optional implementation manner provided by this embodiment, the method provided by this embodiment may further include the following step S200: for each sensor, based on a preset positioning algorithm, constructing a database corresponding to the sensor.

[0108] The construction of the database requires the use of an available positioning algorithm. Specifically, it is necessary to calculate the robot's posture information based on the scene data collected by the sensor based on the positioning algorithm, so as to form database data that can constitute the database.

[0109] At present, there are many available positioning algorithms, such as laser positioning algorithm, wireless signal positioning algorithm, landmark recognition positioning algorithm, etc. Which positioning algorithm is used as the positioning algorithm applied when building the database needs to be determined according to the specific working scene environment conditions and the working task requirements. For example, the method provided in this embodiment is intended to perform posture positioning based on the robot's own sensors, that is, no UWB, markers or other auxiliary objects are deployed in the working scene, but the wireless signal positioning algorithm depends on the deployment of UWB in the working scene, and the landmark recognition positioning algorithm depends on the deployment of markers in the working scene. Therefore, these two positioning algorithms are not applicable to the method provided in this embodiment, while the laser positioning algorithm depends on the point cloud data collected by the laser radar sensor. As long as the laser radar sensor is configured on the robot, it is in line with the method provided in this embodiment. Therefore, the laser positioning algorithm can be used as the positioning algorithm when building the database.

[0110] In a specific implementation, building a database corresponding to the sensor may specifically include the following steps S201 to S205:

[0111] Step S201, in response to controlling the robot to traverse the working scene, scene data collected by a first sensor in each sub-scene is obtained, wherein the first sensor is any one of a plurality of sensors configured for the robot.

[0112] Step S202: extract features for each scene data to obtain feature information corresponding to the scene data.

[0113] Step S203, input the feature information into the positioning algorithm, and obtain the position and posture information corresponding to the feature information output by the positioning algorithm.

[0114] Step S204, combining the feature information and the position and posture information into database data.

[0115] Step S205: The database data corresponding to each sub-scenario is collected into a database corresponding to the first sensor.

[0116] The control of the robot to traverse the working scene may be manual control or automatic control, and its purpose is to enable the robot to collect scene data corresponding to each sub-scene in the working scene through the first sensor as comprehensively as possible.

[0117] The sub-scenes can be understood as partial areas in the working scene. This embodiment does not limit the number and size of sub-scenes, which are specifically determined according to the requirements for the degree of refinement of the robot's working tasks. When the requirements for the degree of refinement of the robot's working tasks are high, the requirements for the positioning of the robot will be high. Then, the working scene will include more sub-scenes, that is, when the robot traverses the working scene, a slight change in the robot's position may cause changes in the sub-scenes. Conversely, when the requirements for the degree of refinement of the robot's working tasks are low, the requirements for the positioning of the robot will be low. Then, the working scene will include relatively fewer sub-scenes, that is, when the robot traverses the working scene, a large change in the robot's position will cause changes in the sub-scenes.

[0118] After obtaining the scene data collected by the first sensor in each sub-scene, feature information can be extracted from the scene data based on some feature extraction methods, such as: trained neural networks, PCA (Principal Component Analysis) dimensionality reduction methods, etc., so as to obtain feature information corresponding to the scene data of each sub-scene. The method of extracting feature information has been used in some technologies and will not be described in detail here. Further, the feature information corresponding to the scene data of each sub-scene is input into a preset positioning algorithm (such as a laser positioning algorithm), and the posture information corresponding to each feature information can be obtained. The corresponding feature information and posture information are combined together to form database data. Since the robot traverses the working scene, multiple database data corresponding to multiple sub-scenes will be obtained based on the above method. These database data are combined together to form a database corresponding to the first sensor.

[0119] In an optional implementation provided in this embodiment, after the initial posture of the robot is obtained, the method provided in this embodiment may further include the following steps S251 to S253:

[0120] Step S251, in response to the robot performing a working task in a working scene, obtaining scene data at the current working moment collected by a second sensor, wherein the second sensor is any one of a plurality of sensors configured for the robot.

[0121] When the robot performs a task in a working scene, the robot side will control the sensor to collect scene data in real time at a certain frequency (e.g., 5 seconds). The server side, based on a preset positioning algorithm, calculates the robot's posture information at the current working moment according to the obtained initial posture of the robot and the scene data at the current working moment collected by the sensor, thereby controlling the robot to move in the working scene in real time.

[0122] Step S252: Acquire database data matching the scene data from a database corresponding to the second sensor.

[0123] The second sensor is any one of the multiple sensors configured by the robot, and the database data matching the scene data needs to be obtained from the database corresponding to the second sensor. In a specific implementation, the following steps S252-1 to S252-3 may be included:

[0124] Step S252-1, extract features from the scene data to obtain feature information corresponding to the scene data.

[0125] Step S252-2, according to the feature information corresponding to the scene data, obtain database data corresponding to the first feature information from the database corresponding to the second sensor, the first feature information being the one with the highest similarity to the feature information corresponding to the scene data among the multiple feature information included in the database corresponding to the second sensor.

[0126] Step S252-3: Use the database data corresponding to the first characteristic information as the database data that matches the scene data.

[0127] The specific implementation details may refer to the detailed description in step S220 and will not be elaborated here. In this embodiment, the feature information in the database with the highest similarity to the feature information corresponding to the scene data is defined as the first feature information.

[0128] Step S253, based on the matching degree between the scene data and the database data, update the database corresponding to the second sensor, or determine whether the posture information at the current operation moment obtained based on the preset positioning algorithm is correct.

[0129] The degree of matching between the scene data and the database data is represented by the degree of similarity between the feature information corresponding to the scene data and the first feature information. The more similar they are, the more likely it is that the feature information corresponding to the scene data is the same as the first feature information. The less similar they are, the less likely it is that the feature information corresponding to the scene data is the same as the first feature information. Based on the matching relationship between the scene data and the database data, the database is updated or the correctness of the pose information is judged. Specifically, the following two situations may be included:

[0130] Case 1: In response to the similarity between the feature information corresponding to the scene data and the first feature information being less than a first preset threshold, the database corresponding to the second sensor is updated based on the feature information corresponding to the scene data.

[0131] Case 2: In response to the similarity between the feature information corresponding to the scene data and the first feature information being greater than or equal to a first preset threshold, it is determined based on the first feature information whether the posture information at the current operation moment obtained based on the positioning algorithm is correct.

[0132] The first preset threshold can be understood as an indicator value for measuring whether the feature information corresponding to the scene data is similar to the first feature information. For example, if the feature information is represented in vector form, then the first preset threshold can be an indicator value set for the retrieval score, such as 90 points.

[0133] Specifically, if the similarity between the two is less than the first preset threshold, it is judged that the two are not similar, indicating that the feature information corresponding to the scene data is not currently stored in the database corresponding to the second sensor. In this case, the feature information corresponding to the scene data can be combined into new database data and stored in the database; if the similarity between the two is greater than or equal to the first preset threshold, it is judged that the two are similar, indicating that the feature information corresponding to the scene data (i.e., the first feature information) is currently stored in the database corresponding to the second sensor. In this case, there is no need to store the scene data again, and the posture information corresponding to the feature information (i.e., the first feature information) stored in the database can be used to make an abnormality judgment on the posture information of the robot at the current working moment calculated by the positioning algorithm.

[0134] In a specific implementation, updating the database corresponding to the second sensor based on the feature information corresponding to the scene data may include the following steps S253-11 to S253-13:

[0135] Step S253-11, obtaining the posture information at the current operation time based on the positioning algorithm.

[0136] Step S253-12, combining the feature information and the posture information into the third database data.

[0137] Step S253-13, collect the data from the third database into the database corresponding to the second sensor.

[0138] Since the feature information corresponding to the scene data does not exist in the database corresponding to the second sensor, the feature information corresponding to the scene data can be combined into new database data and stored in the database. The database data stored in the database includes not only the feature information but also the posture information corresponding to the feature information. Therefore, in this embodiment, the posture information at the current operation moment obtained based on the positioning algorithm is used as the posture information corresponding to the feature information. The feature information and the posture information are combined into new database data. In this embodiment, the new database data is defined as third database data. After the third database data is obtained, the third database data can be collected into the database corresponding to the second sensor to update the database. In a specific implementation method, the third database data can also be stored in a storage medium for use by the positioning algorithm in subsequent update training.

[0139] In a specific implementation, judging whether the posture information at the current operation time obtained based on the positioning algorithm is correct based on the first feature information may include the following steps S253-21 to S253-22:

[0140] Step S253-21, obtaining the posture information at the current operation time based on the positioning algorithm.

[0141] Step S253-22, in response to the difference between the posture information included in the database data corresponding to the first feature information and the posture information obtained based on the positioning algorithm being greater than a second preset threshold, it is determined that the posture information obtained based on the positioning algorithm is wrong.

[0142] If the similarity between the first feature information and the feature information corresponding to the scene data is greater than the first preset threshold, it can be understood that the first feature information and the feature information corresponding to the scene data are basically the same, and the posture information corresponding to the basically same feature information should also be basically the same. When the difference between the two is too large, it means that one of the posture information is incorrect. Since the posture information in the database is pre-constructed based on the available positioning algorithm in the early stage of robot application, the possibility of error is smaller. Therefore, the posture information corresponding to the first feature information in the database can be used to determine whether the posture information obtained in real time based on the positioning algorithm during the robot operation is correct and whether there is an error in the positioning algorithm.

[0143] The second preset threshold can be understood as an indicator value for measuring whether the posture information corresponding to the first feature information is consistent with the posture information obtained based on the positioning algorithm. The second preset threshold can be preset according to the degree of refinement of the robot positioning, such as setting the second preset threshold to a distance of 5 meters and an angle of 30 degrees.

[0144] Specifically, if the difference between the posture information corresponding to the first feature information and the posture information obtained based on the positioning algorithm is less than or equal to the second preset threshold, it is judged that there is no difference between the two, the two are consistent, the posture information obtained based on the positioning algorithm is correct, and there is no abnormality in the positioning algorithm; and if the difference between the posture information corresponding to the first feature information and the posture information obtained based on the positioning algorithm is greater than the second preset threshold, it is judged that there is a difference between the two, the two are inconsistent, the posture information obtained based on the positioning algorithm is incorrect, and there is an abnormality in the positioning algorithm. In a specific implementation method, after judging that the positioning algorithm is abnormal, a warning can be output for the abnormal situation.

[0145] The following is an explanation using a specific example, in which it is assumed that the robot is equipped with two camera sensors (camera sensor A and camera sensor B) and a lidar sensor.

[0146] Figure 4 It is a schematic diagram of the robot posture positioning process provided in this embodiment.

[0147] like Figure 4 The robot's posture positioning process may include the following steps S401 to S408:

[0148] Step S401, obtaining scene data of the working scene collected by the robot based on the sensor.

[0149] Specifically, scene picture 1 collected by camera sensor A, scene picture 2 collected by camera sensor B, and scene point cloud data collected by a lidar sensor are obtained.

[0150] Step S402: extract features from the scene data to obtain feature information corresponding to the scene data.

[0151] Specifically, feature extraction is performed on scene image 1, scene image 2, and scene point cloud data to obtain feature information 1 corresponding to scene image 1, feature information 2 corresponding to scene image 2, and feature information 3 corresponding to the scene point cloud data.

[0152] Step S403, obtaining the position and posture information of the robot at the current working moment based on a preset positioning algorithm.

[0153] Specifically, according to the type of the preset positioning algorithm, feature information 1, and / or feature information 2, and / or feature information 3 are used as input data to obtain the position information A of the robot at the current working moment output by the positioning algorithm. Exemplarily, when the preset positioning algorithm is a laser positioning algorithm, the input data is feature information 3.

[0154] Step S404: Retrieve feature information having the highest similarity to feature information corresponding to the scene data from the database.

[0155] Specifically, based on feature information 1, feature information 2, and feature information 3, feature information with the highest similarity to feature information 1, feature information 2, and feature information 3 is retrieved from database 1 corresponding to camera sensor A, database 2 corresponding to camera sensor B, and database 3 corresponding to the lidar sensor. It is assumed that the feature information with the highest similarity to feature information 1, feature information 2, and feature information 3 are feature information a, feature information b, and feature information c, respectively.

[0156] Step S405, if the similarity between the feature information corresponding to the scene data and the feature information with the highest similarity retrieved from the database is less than a first preset threshold, the feature information corresponding to the scene data and the position information of the robot at the current working moment obtained based on a preset positioning algorithm are combined into database data, and the database is updated.

[0157] Specifically, assuming that the similarity between feature information 1 and feature information a is less than the first preset threshold, feature information 1 and posture information A are combined into database data 1, and database data 1 is collected into database 1 corresponding to camera sensor A to update database 1; assuming that the similarity between feature information 2 and feature information b is less than the first preset threshold, feature information 2 and posture information A are combined into database data 2, and database data 2 is collected into database 2 corresponding to camera sensor B to update database 2; assuming that the similarity between feature information 3 and feature information c is less than the first preset threshold, feature information 3 and posture information A are combined into database data 3, and database data 3 is collected into database 3 corresponding to the lidar sensor to update database 3.

[0158] Step S406, if the similarity between the feature information corresponding to the scene data and the feature information with the highest similarity retrieved from the database is greater than or equal to a first preset threshold, determine the difference between the posture information corresponding to the feature information with the highest similarity and the posture information of the robot at the current working moment obtained based on the preset positioning algorithm. If the difference between the two is greater than a second preset threshold, determine that the posture information of the robot at the current working moment obtained based on the preset positioning algorithm is incorrect.

[0159] Specifically, assuming that the similarity between feature information 1 and feature information a is greater than or equal to the first preset threshold, the difference value between the posture information a corresponding to feature information a and the posture information A is determined. Assuming that the difference value between posture information a and posture information A is greater than the second preset threshold, it is determined that the posture information A is incorrect; assuming that the similarity between feature information 2 and feature information b is greater than or equal to the first preset threshold, the difference value between the posture information b corresponding to feature information b and the posture information A is determined. Assuming that the difference value between posture information b and posture information A is greater than the second preset threshold, it is determined that the posture information A is incorrect; assuming that the similarity between feature information 3 and feature information c is greater than or equal to the first preset threshold, the difference value between the posture information c corresponding to feature information c and the posture information A is determined. Assuming that the difference value between posture information c and posture information A is less than the second preset threshold, it is determined that the posture information A is correct.

[0160] Step S407: If it is determined that the position and posture information of the robot at the current working moment obtained based on the preset positioning algorithm is incorrect, the abnormal position and posture information is recorded and a warning is output.

[0161] Step S408, saving the abnormal posture information and the newly obtained database data to a storage medium for use in subsequent positioning algorithm update training.

[0162] The first embodiment described above provides a robot positioning method, which aims to obtain a robot initial position with higher confidence through multiple sensors. In addition, the method realizes dynamic update of the database, and can continuously update the data in the database during the robot operation process to improve the global positioning capability. It also realizes the warning of abnormal positioning during the operation process, so that when the robot is kidnapped or the robot sensor has an abnormality, it can be discovered in time to avoid problems such as robot running trajectory deviation. It should be noted that the example in the first embodiment is only for explaining the method described in this application, and is not intended to be a limitation for actual use. The robot positioning method provided in this application includes but is not limited to the method described in the first embodiment.

[0163] A second embodiment of the present application provides a positioning device for a robot, wherein the robot is configured with a plurality of sensors, wherein the plurality of sensors include at least two sensor types, and the robot is located in a working scene. Figure 5 It is a schematic diagram of the structure of the positioning device of the robot provided in this embodiment.

[0164] like Figure 5 As shown, the robot positioning device provided in this embodiment includes: a first data acquisition unit 501, a second data acquisition unit 502, a data screening unit 503, and a posture determination unit 504.

[0165] The first data acquisition unit 501 is used to acquire a plurality of scene data collected by the plurality of sensors.

[0166] The second data acquisition unit 502 is used to obtain database data matching the scene data from a pre-built database for each scene data, wherein the database corresponds to a sensor that collects the scene data, and includes a plurality of database data, each of which corresponds to a sub-scene in the operation scene, including feature information corresponding to the sub-scene and posture information of the robot in the sub-scene.

[0167] Optionally, for each of the scene data, obtaining database data matching the scene data from a pre-built database includes:

[0168] Extracting features from the scene data to obtain feature information corresponding to the scene data;

[0169] Retrieving database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data in the database;

[0170] The database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data is used as the database data matching the scene data.

[0171] Optionally, the feature information is represented in the form of a vector;

[0172] The retrieving the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data in the database comprises:

[0173] Performing a vector product of the feature information corresponding to the scene data and the feature information included in each of the database data in the database;

[0174] Using the value of the vector product as a retrieval score of the feature information included in the database data;

[0175] The database data corresponding to the feature information having the highest search score is used as the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data.

[0176] The data screening unit 503 is used to screen the optimal database data from the multiple database data corresponding to the multiple scene data based on a preset screening logic.

[0177] Optionally, the selecting the optimal database data from the plurality of database data corresponding to the plurality of scene data based on a preset selection logic includes:

[0178] Inputting the posture information included in the plurality of database data into a preset positioning algorithm;

[0179] The optimal database data is determined according to a convergence condition of the positioning algorithm, wherein the convergence condition is used to characterize the robustness of the positioning algorithm when iteratively calculating the current posture of the robot based on the posture information.

[0180] Optionally, the convergence status includes inputting the posture information into the positioning algorithm, and whether the positioning algorithm converges during the iterative calculation process; the screening of the optimal database data from the multiple database data corresponding to the multiple scene data based on the preset screening logic includes:

[0181] Inputting the posture information included in the first database data into the positioning algorithm, wherein the first database data is the one of the plurality of database data that has the highest degree of matching with the scene data;

[0182] In response to the positioning algorithm converging, determining the first database data as the optimal database data;

[0183] In response to the positioning algorithm not converging, inputting the pose information included in the second database data into the positioning algorithm according to the matching degree between the second database data and the scene data, the second database data being at least one of the plurality of database data having a matching degree with the scene data lower than that of the first database data;

[0184] In response to the positioning algorithm converging, the second database data at the time when the positioning algorithm converges is determined as the optimal database data.

[0185] Optionally, the convergence status includes inputting the posture information into the positioning algorithm, and a convergence score output by the positioning algorithm during an iterative calculation process; the screening of the optimal database data from the multiple database data corresponding to the multiple scene data based on a preset screening logic includes:

[0186] Inputting the posture information included in each database data into the positioning algorithm to obtain the convergence score corresponding to each database data;

[0187] The database data with the largest convergence score is taken as the optimal database data.

[0188] The posture determination unit 504 is used to determine the initial posture of the robot in the working scene according to the posture information included in the optimal database data.

[0189] Optionally, the device further comprises: a database construction unit;

[0190] The database construction unit is used to:

[0191] For each of the sensors, a database corresponding to the sensor is constructed based on a preset positioning algorithm.

[0192] Optionally, for each of the sensors, based on a preset positioning algorithm, constructing a database corresponding to the sensor includes:

[0193] In response to controlling the robot to traverse the working scene, acquiring scene data collected by a first sensor in each sub-scene, wherein the first sensor is any one of the multiple sensors configured for the robot;

[0194] Performing feature extraction on each of the scene data to obtain feature information corresponding to the scene data;

[0195] Inputting the feature information into the positioning algorithm to obtain the position and posture information output by the positioning algorithm corresponding to the feature information;

[0196] Combining the feature information with the position information into the database data;

[0197] The database data set corresponding to each sub-scenario is formed into a database corresponding to the first sensor.

[0198] Optionally, the device further comprises: a database updating and abnormality judging unit;

[0199] The database updating and abnormality judging unit is used to:

[0200] In response to the robot performing a work task in the work scene, acquiring scene data at the current work moment collected by a second sensor, wherein the second sensor is any one of the multiple sensors configured for the robot;

[0201] Acquire database data matching the scene data from the database corresponding to the second sensor;

[0202] According to the matching degree between the scene data and the database data, the database corresponding to the second sensor is updated, or it is determined whether the posture information at the current operation moment obtained based on a preset positioning algorithm is correct.

[0203] Optionally, the acquiring, from the database corresponding to the second sensor, database data matching the scene data includes:

[0204] Extracting features from the scene data to obtain feature information corresponding to the scene data;

[0205] According to the feature information corresponding to the scene data, database data corresponding to first feature information is acquired from the database corresponding to the second sensor, wherein the first feature information is one of a plurality of feature information included in the database corresponding to the second sensor and having the highest similarity to the feature information corresponding to the scene data;

[0206] The database data corresponding to the first characteristic information is used as database data matching the scene data.

[0207] Optionally, updating the database corresponding to the second sensor according to the degree of matching between the scene data and the database data, or judging whether the posture information at the current operation moment obtained based on a preset positioning algorithm is correct, includes:

[0208] In response to a similarity between the feature information corresponding to the scene data and the first feature information being less than a first preset threshold, updating the database corresponding to the second sensor based on the feature information corresponding to the scene data;

[0209] In response to the similarity between the feature information corresponding to the scene data and the first feature information being greater than or equal to the first preset threshold, it is determined based on the first feature information whether the posture information at the current working moment obtained based on the positioning algorithm is correct.

[0210] Optionally, updating the database corresponding to the second sensor based on the feature information corresponding to the scene data includes:

[0211] Acquire the position information at the current operation time obtained based on the positioning algorithm;

[0212] Combining the feature information with the position information into third database data;

[0213] The third database data set is integrated into the database corresponding to the second sensor.

[0214] Optionally, judging whether the posture information at the current operation moment obtained based on the positioning algorithm is correct based on the first feature information includes:

[0215] Acquire the position information at the current operation time obtained based on the positioning algorithm;

[0216] In response to the difference between the posture information included in the database data corresponding to the first feature information and the posture information obtained based on the positioning algorithm being greater than a second preset threshold, it is determined that the posture information obtained based on the positioning algorithm is wrong.

[0217] A third embodiment of the present application provides an electronic device, Figure 6 It is a schematic diagram of the structure of the electronic device provided in this embodiment.

[0218] like Figure 6 As shown, the electronic device provided in this embodiment includes: a memory 601 and a processor 602;

[0219] The memory 601 is used to store computer instructions for executing the positioning method of the robot;

[0220] The processor 602 is configured to execute computer instructions stored in the memory 601 to perform the following operations:

[0221] Acquire multiple scene data collected by the multiple sensors;

[0222] For each of the scene data, database data matching the scene data is obtained from a pre-built database, the database corresponding to the sensor that collects the scene data, including a plurality of database data, each of which corresponds to a sub-scene in the operation scene, including feature information corresponding to the sub-scene and position information of the robot in the sub-scene;

[0223] Based on a preset screening logic, screening the best database data from the plurality of database data corresponding to the plurality of scene data;

[0224] According to the posture information included in the optimal database data, the initial posture of the robot in the working scene is determined.

[0225] Optionally, for each of the scene data, obtaining database data matching the scene data from a pre-built database includes:

[0226] Extracting features from the scene data to obtain feature information corresponding to the scene data;

[0227] Retrieving database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data in the database;

[0228] The database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data is used as the database data matching the scene data.

[0229] Optionally, the feature information is represented in the form of a vector;

[0230] The retrieving the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data in the database comprises:

[0231] Performing a vector product of the feature information corresponding to the scene data and the feature information included in each of the database data in the database;

[0232] Using the value of the vector product as a retrieval score of the feature information included in the database data;

[0233] The database data corresponding to the feature information having the highest search score is used as the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data.

[0234] Optionally, the selecting the optimal database data from the plurality of database data corresponding to the plurality of scene data based on a preset selection logic includes:

[0235] Inputting the posture information included in the plurality of database data into a preset positioning algorithm;

[0236] The optimal database data is determined according to a convergence condition of the positioning algorithm, wherein the convergence condition is used to characterize the robustness of the positioning algorithm when iteratively calculating the current posture of the robot based on the posture information.

[0237] Optionally, the convergence status includes inputting the posture information into the positioning algorithm, and whether the positioning algorithm converges during the iterative calculation process; the screening of the optimal database data from the multiple database data corresponding to the multiple scene data based on the preset screening logic includes:

[0238] Inputting the posture information included in the first database data into the positioning algorithm, wherein the first database data is the one of the plurality of database data that has the highest degree of matching with the scene data;

[0239] In response to the positioning algorithm converging, determining the first database data as the optimal database data;

[0240] In response to the positioning algorithm not converging, inputting the pose information included in the second database data into the positioning algorithm according to the matching degree between the second database data and the scene data, the second database data being at least one of the plurality of database data having a matching degree with the scene data lower than that of the first database data;

[0241] In response to the positioning algorithm converging, the second database data at the time when the positioning algorithm converges is determined as the optimal database data.

[0242] Optionally, the convergence status includes inputting the posture information into the positioning algorithm, and a convergence score output by the positioning algorithm during an iterative calculation process; the screening of the optimal database data from the multiple database data corresponding to the multiple scene data based on a preset screening logic includes:

[0243] Inputting the posture information included in each database data into the positioning algorithm to obtain the convergence score corresponding to each database data;

[0244] The database data with the largest convergence score is taken as the optimal database data.

[0245] Optionally, before the step of acquiring the plurality of scene data collected by the plurality of sensors, the following steps are further performed:

[0246] For each of the sensors, a database corresponding to the sensor is constructed based on a preset positioning algorithm.

[0247] Optionally, for each of the sensors, based on a preset positioning algorithm, constructing a database corresponding to the sensor includes:

[0248] In response to controlling the robot to traverse the working scene, acquiring scene data collected by a first sensor in each sub-scene, wherein the first sensor is any one of the multiple sensors configured for the robot;

[0249] Performing feature extraction on each of the scene data to obtain feature information corresponding to the scene data;

[0250] Inputting the feature information into the positioning algorithm to obtain the position and posture information output by the positioning algorithm corresponding to the feature information;

[0251] Combining the feature information with the position information into the database data;

[0252] The database data set corresponding to each sub-scenario is formed into a database corresponding to the first sensor.

[0253] Optionally, after the step of determining the initial posture of the robot according to the posture information included in the optimal database data, the following is further performed:

[0254] In response to the robot performing a work task in the work scene, acquiring scene data at the current work moment collected by a second sensor, wherein the second sensor is any one of the multiple sensors configured for the robot;

[0255] Acquire database data matching the scene data from the database corresponding to the second sensor;

[0256] According to the matching degree between the scene data and the database data, the database corresponding to the second sensor is updated, or it is determined whether the posture information at the current operation moment obtained based on a preset positioning algorithm is correct.

[0257] Optionally, the acquiring, from the database corresponding to the second sensor, database data matching the scene data includes:

[0258] Extracting features from the scene data to obtain feature information corresponding to the scene data;

[0259] According to the feature information corresponding to the scene data, database data corresponding to first feature information is acquired from the database corresponding to the second sensor, wherein the first feature information is one of a plurality of feature information included in the database corresponding to the second sensor and having the highest similarity to the feature information corresponding to the scene data;

[0260] The database data corresponding to the first characteristic information is used as database data matching the scene data.

[0261] Optionally, updating the database corresponding to the second sensor according to the degree of matching between the scene data and the database data, or judging whether the posture information at the current operation moment obtained based on a preset positioning algorithm is correct, includes:

[0262] In response to a similarity between the feature information corresponding to the scene data and the first feature information being less than a first preset threshold, updating the database corresponding to the second sensor based on the feature information corresponding to the scene data;

[0263] In response to the similarity between the feature information corresponding to the scene data and the first feature information being greater than or equal to the first preset threshold, it is determined based on the first feature information whether the posture information at the current working moment obtained based on the positioning algorithm is correct.

[0264] Optionally, updating the database corresponding to the second sensor based on the feature information corresponding to the scene data includes:

[0265] Acquire the position information at the current operation time obtained based on the positioning algorithm;

[0266] Combining the feature information with the position information into third database data;

[0267] The third database data set is integrated into the database corresponding to the second sensor.

[0268] Optionally, judging whether the posture information at the current operation moment obtained based on the positioning algorithm is correct based on the first feature information includes:

[0269] Acquire the position information at the current operation time obtained based on the positioning algorithm;

[0270] In response to the difference between the posture information included in the database data corresponding to the first feature information and the posture information obtained based on the positioning algorithm being greater than a second preset threshold, it is determined that the posture information obtained based on the positioning algorithm is wrong.

[0271] A fourth embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium includes computer instructions. When the computer instructions are executed by a processor, they are used to implement the methods described in the embodiments of the present application.

[0272] It should be noted that the relational terms such as "first" and "second" in this article are only used to distinguish one entity or operation from another entity or operation, and do not require or imply any actual relationship or order between these entities or operations. In addition, the words "include", "have", "include" and "includes" and other similar forms are the same in meaning, and the end of any one or more items after any of the above words is open-ended, and any of the above nouns does not mean that the one or more items have been exhaustively listed or are limited to the one or more items listed.

[0273] As used herein, unless expressly stated otherwise, the term "or" includes all possible combinations, except those that are not feasible. For example, if it is expressed that a database may include A or B, then unless otherwise specifically stated or not feasible, it may include database A, or B, or A and B. As a second example, if it is expressed that a database may include A, B, or C, then unless otherwise specifically stated or not feasible, the database may include database A, or B, or C, or A and B, or A and C, or B and C, or A, B, and C.

[0274] It is worth noting that the above embodiments can be implemented by hardware or software (program code), or a combination of hardware and software. If implemented by software, it can be stored in the above-mentioned computer-readable medium. When the software is executed by a processor, it can execute the above-mentioned disclosed method. The computing unit and other functional units described in the present disclosure can be implemented by hardware or software, or a combination of hardware and software. Those of ordinary skill in the art will also understand that the above-mentioned multiple modules / units can be combined into one module / unit, and each of the above-mentioned modules / units can be further divided into multiple sub-modules / sub-units.

[0275] In the above detailed description, the embodiments have been described with reference to many specific details, which may vary from implementation to implementation. Certain adaptations and modifications may be made to the embodiments. For those skilled in the art, other embodiments may be readily apparent from the specific embodiments disclosed in the present application. This specification and examples are for exemplary purposes only, and the true scope and essence of the present application are described by the claims. The sequence of steps shown in the diagrams is also for the purpose of explanation only and is not intended to be limited to any particular steps or sequence. Therefore, those skilled in the art will appreciate that these steps may be performed in different orders when implementing the same method.

[0276] In the drawings and detailed description of the present application, exemplary embodiments are disclosed. However, many variations and modifications may be made to these embodiments. Accordingly, although specific terms are used, these terms are only general and descriptive and not for the purpose of limitation.

Claims

1. A robot positioning method, It is characterized in that The robot is configured with a plurality of sensors, the plurality of sensors include at least two sensor types, the robot is located in a working scene, and the method includes: Acquire multiple scene data collected by the multiple sensors; For each of the scene data, database data matching the scene data is obtained from a pre-built database, the database corresponding to the sensor that collects the scene data, including a plurality of database data, each of which corresponds to a sub-scene in the operation scene, including feature information corresponding to the sub-scene and position information of the robot in the sub-scene; Based on a preset screening logic, screening the best database data from the plurality of database data corresponding to the plurality of scene data; According to the posture information included in the optimal database data, the initial posture of the robot in the working scene is determined.

2. The method according to claim 1, It is characterized in that For each of the scene data, obtaining database data matching the scene data from a pre-built database includes: Extracting features from the scene data to obtain feature information corresponding to the scene data; Retrieving database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data in the database; The database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data is used as the database data matching the scene data.

3. The method according to claim 2, It is characterized in that The feature information is represented in the form of a vector; The retrieving the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data in the database comprises: Performing a vector product of the feature information corresponding to the scene data and the feature information included in each of the database data in the database; Using the value of the vector product as a retrieval score of the feature information included in the database data; The database data corresponding to the feature information having the highest search score is used as the database data corresponding to the feature information having the highest similarity to the feature information corresponding to the scene data.

4. The method according to claim 1, It is characterized in that The selecting the optimal database data from the plurality of database data corresponding to the plurality of scene data based on the preset selecting logic includes: Inputting the posture information included in the plurality of database data into a preset positioning algorithm; The optimal database data is determined according to a convergence condition of the positioning algorithm, wherein the convergence condition is used to characterize the robustness of the positioning algorithm when iteratively calculating the current posture of the robot based on the posture information.

5. The method according to claim 4, It is characterized in that The convergence condition includes inputting the position information into the positioning algorithm, and whether the positioning algorithm converges during the iterative calculation process; The selecting the optimal database data from the plurality of database data corresponding to the plurality of scene data based on the preset selecting logic includes: Inputting the posture information included in the first database data into the positioning algorithm, wherein the first database data is the one of the plurality of database data that has the highest degree of matching with the scene data; In response to the positioning algorithm converging, determining the first database data as the optimal database data; In response to the positioning algorithm not converging, inputting the pose information included in the second database data into the positioning algorithm according to the matching degree between the second database data and the scene data, the second database data being at least one of the plurality of database data having a matching degree with the scene data lower than that of the first database data; In response to the positioning algorithm converging, the second database data at the time when the positioning algorithm converges is determined as the optimal database data.

6. The method according to claim 4, It is characterized in that The convergence condition includes inputting the position information into the positioning algorithm, and the convergence score output by the positioning algorithm during the iterative calculation process; The selecting the optimal database data from the plurality of database data corresponding to the plurality of scene data based on the preset selecting logic includes: Inputting the posture information included in each database data into the positioning algorithm to obtain the convergence score corresponding to each database data; The database data with the largest convergence score is taken as the optimal database data.

7. The method according to claim 1, It is characterized in that Before the step of acquiring the plurality of scene data collected by the plurality of sensors, the method further includes: For each of the sensors, a database corresponding to the sensor is constructed based on a preset positioning algorithm.

8. The method according to claim 7, It is characterized in that For each of the sensors, based on a preset positioning algorithm, a database corresponding to the sensor is constructed, including: In response to controlling the robot to traverse the working scene, acquiring scene data collected by a first sensor in each sub-scene, wherein the first sensor is any one of the multiple sensors configured for the robot; Performing feature extraction on each of the scene data to obtain feature information corresponding to the scene data; Inputting the feature information into the positioning algorithm to obtain the position and posture information output by the positioning algorithm corresponding to the feature information; Combining the feature information with the position information into the database data; The database data set corresponding to each sub-scenario is formed into a database corresponding to the first sensor.

9. The method according to claim 1, It is characterized in that After the step of determining the initial posture of the robot according to the posture information included in the optimal database data, the method further comprises: In response to the robot performing a work task in the work scene, acquiring scene data at the current work moment collected by a second sensor, wherein the second sensor is any one of the multiple sensors configured for the robot; Acquire database data matching the scene data from the database corresponding to the second sensor; According to the matching degree between the scene data and the database data, the database corresponding to the second sensor is updated, or it is determined whether the posture information at the current operation moment obtained based on a preset positioning algorithm is correct.

10. The method according to claim 9, It is characterized in that The acquiring database data matching the scene data from the database corresponding to the second sensor includes: Extracting features from the scene data to obtain feature information corresponding to the scene data; According to the feature information corresponding to the scene data, database data corresponding to first feature information is acquired from the database corresponding to the second sensor, wherein the first feature information is one of a plurality of feature information included in the database corresponding to the second sensor and having the highest similarity to the feature information corresponding to the scene data; The database data corresponding to the first characteristic information is used as database data matching the scene data.

11. The method according to claim 10, It is characterized in that The updating of the database corresponding to the second sensor according to the matching degree between the scene data and the database data, or judging whether the posture information at the current operation moment obtained based on a preset positioning algorithm is correct, includes: In response to a similarity between the feature information corresponding to the scene data and the first feature information being less than a first preset threshold, updating the database corresponding to the second sensor based on the feature information corresponding to the scene data; In response to the similarity between the feature information corresponding to the scene data and the first feature information being greater than or equal to the first preset threshold, it is determined based on the first feature information whether the posture information at the current working moment obtained based on the positioning algorithm is correct.

12. The method according to claim 11, It is characterized in that The updating the database corresponding to the second sensor based on the feature information corresponding to the scene data includes: Acquire the position information at the current operation time obtained based on the positioning algorithm; Combining the feature information with the position information into third database data; The third database data set is integrated into the database corresponding to the second sensor.

13. The method according to claim 11, It is characterized in that The determining, based on the first feature information, whether the posture information at the current operation moment obtained based on the positioning algorithm is correct includes: Acquire the position information at the current operation time obtained based on the positioning algorithm; In response to the difference between the posture information included in the database data corresponding to the first feature information and the posture information obtained based on the positioning algorithm being greater than a second preset threshold, it is determined that the posture information obtained based on the positioning algorithm is wrong.

14. A robot positioning device, It is characterized in that The robot is equipped with a plurality of sensors, the plurality of sensors include at least two sensor types, the robot is located in a working scene, and the device includes: a first data acquisition unit, a second data acquisition unit, a data screening unit, and a posture determination unit; The first data acquisition unit is used to acquire a plurality of scene data collected by the plurality of sensors; The second data acquisition unit is used to acquire, for each of the scene data, database data matching the scene data from a pre-built database, wherein the database corresponds to a sensor that collects the scene data, and includes a plurality of database data, each of which corresponds to a sub-scene in the operation scene, and includes feature information corresponding to the sub-scene and posture information of the robot in the sub-scene; The data screening unit is used to screen the optimal database data from the plurality of database data corresponding to the plurality of scene data based on a preset screening logic; The posture determination unit is used to determine the initial posture of the robot in the working scene according to the posture information included in the optimal database data.

15. An electronic device, It is characterized in that include: Memory, processor; The memory is used to store one or more computer instructions; The processor is used to execute the one or more computer instructions to implement the method according to any one of claims 1-13.

16. A computer-readable storage medium having stored thereon one or more computer instructions, It is characterized in that When the instruction is executed by a processor, the method according to any one of claims 1 to 13 is performed.