Wheel robot repositioning method based on inertial odometer

Abstract

The present application belongs to the technical field of intelligent robot repositioning, and particularly relates to a wheeled robot repositioning method based on inertial odometer. Specifically comprising: fixed distance interval trigger record laser radar pose list and inertial pose list, through calculating the transformation matrix of the far and near poses in the two lists, comparing the Euclidean distance of the translation vector and the rotation angle difference to judge the loss of positioning; when the positioning is lost, sequentially execute the initial repositioning and forced repositioning, the latter uses the inertial pose transformation matrix to calculate the target pose; if the positioning is lost again after forced repositioning, trigger the alarm. The method does not need additional high-precision sensors, and the positioning defects of single sensor are compensated by algorithm optimization, which improves the positioning reliability and environmental adaptability, and reduces the system cost and complexity.

Description

Technical Field

[0001] This invention belongs to the field of intelligent robot relocation technology, and particularly relates to a relocation method for wheeled robots based on inertial odometry. Background Technology

[0002] In the field of intelligent robot applications, wheeled robots are widely used in logistics warehousing, environmental detection, and service guidance due to their simple structure and flexible movement. However, when wheeled robots are in complex environments, especially those with many similar scene features, localization loss becomes a key factor restricting their stable operation. For example, in a large warehouse, neatly arranged similar shelves, or in a hospital corridor, similar doors and walls, make it difficult for robots to accurately identify their own position using sensors such as vision and lidar. Localization loss prevents the robot from accurately knowing its own location, thus hindering its ability to perform tasks according to the predetermined path and severely impacting its normal operation.

[0003] Currently, mainstream robot localization technologies each have their limitations. While lidar-based localization technology can provide relatively accurate pose information in most cases, in complex environments, the reflected signals of lidar can be interfered with, such as by specular reflection and multipath effects, leading to a gradual accumulation of positioning errors and eventual loss of positioning. Vision-based localization methods are susceptible to changes in lighting and occlusion, making it difficult to obtain reliable positioning data stably in low light or when the target is obstructed. Inertial odometry (IoU) calculates robot pose using data from wheel speed sensors and an IMU (Inertial Measurement Unit). Although cumulative errors exist, its accuracy is sufficient for certain positioning requirements within short distances, such as within 5 meters. Therefore, utilizing this characteristic of IoU to assist in robot relocalization in complex environments is of great significance. Summary of the Invention

[0004] To address the technical problems existing in the background art described above, this invention proposes a repositioning method for wheeled robots based on inertial odometry.

[0005] To achieve the above objectives, the technical solution adopted by the present invention includes the following steps:

[0006] S1. Data Recording: During the robot's movement, two key data lists are maintained in real time, including the LiDAR pose list and the inertial pose list;

[0007] LiDAR pose list: Triggered at fixed distance intervals, it records the LiDAR pose calculated by the robot pose localization algorithm. When the total travel distance meets the set conditions, a new recording cycle is started.

[0008] Inertial pose list: Triggered at fixed distance intervals, the inertial pose calculated by the inertial odometry is recorded. When the total travel distance meets the set conditions, a new recording cycle is started.

[0009] S2. Location Loss Detection: Periodically analyze the data from both lists to determine if the robot has lost its location; the specific steps are as follows:

[0010] Calculate the inertial pose transformation matrix: from the inertial pose vector list data Select the inertial pose vector that is closest to the current time. and the inertial pose vector furthest from the current time. Calculate the transformation matrix between two inertial poses. ;

[0011] Calculate the lidar pose transformation matrix: from lidar pose vector list data Select the lidar pose vector closest to the current time. and the lidar pose vector furthest from the current time Calculate the transformation matrix between the poses of the two lidar sensors. ;

[0012] Location loss detection: by comparing the transformation matrix and The system determines whether the robot has lost its localization based on two dimensions: position and rotation. This includes position difference assessment, rotation difference assessment, and a comprehensive assessment to determine whether the robot has lost its localization and requires a redirection operation.

[0013] Location difference judgment: extraction and The translation parts in the matrix are denoted as follows: Translation vector and Translation vector Calculate the Euclidean distance between two translation vectors. ;like If the location exceeds a preset threshold, it is determined as the first location loss;

[0014] Rotation difference determination: Calculate the rotation angle value of the Z-axis used in the transformation matrix. and The difference in the corresponding rotation angle is denoted as rotation difference. ;like If the rotation threshold is exceeded, it is determined that the second positioning is lost;

[0015] Comprehensive judgment: If both the first and second positioning are determined to be lost, it is determined that the robot has lost positioning and a repositioning operation is performed.

[0016] S3. Relocation operation: including initial relocation and forced relocation;

[0017] Initial relocation: After detecting a loss of localization, the robot stops moving, calls the relocation interface to perform relocation, and verifies the relocation result. If the relocation is successful, the localization is successful.

[0018] Forced relocation: If the initial relocation fails, a forced relocation operation is performed, starting from... Extract the lidar pose vector that is furthest from the current time. Using the previously obtained inertial pose transformation matrix The updated latest lidar pose is calculated using matrix multiplication. The calculation method is as follows: , position As the target pose for the robot's forced relocation;

[0019] S4. Trigger Warning: After forced repositioning, if the robot detects a loss of positioning again when it has traveled a preset distance, it will stop moving and trigger an alarm through sound, light, network, or other means, while recording the running data.

[0020] Preferably, the robot pose localization algorithm in step S1 includes algorithms based on vision, laser SLAM, and GPS localization.

[0021] Preferably, in step S1, the new recording cycle for the lidar pose list and the inertial pose list is achieved by using a rolling recording mechanism to maintain a fixed length. When new data is written, the oldest recorded data is automatically removed, and the list content is continuously updated.

[0022] Preferably, the data recording method for the lidar pose list in step S1 is as follows:

[0023] With a fixed distance interval d as the trigger condition, whenever the robot moves a distance of d, At that time, the lidar pose vector calculated by the robot pose localization algorithm at the current moment is used. Recorded in a list, where, This is the pose vector of the lidar. This represents the robot's position coordinates in a two-dimensional plane. This indicates the robot's rotation angle around the Z-axis. A new recording cycle is started when the total travel distance represented by the recorded LiDAR poses reaches a certain threshold or the number of LiDAR poses in the list exceeds a preset threshold.

[0024] Preferably, the data recording method for the inertial pose list in step S1 is as follows:

[0025] With a fixed distance interval d as the trigger condition, whenever the robot moves a distance of d, the robot's inertial pose vector at the current moment is calculated by fusing wheel speed and the rotation angle of the Z-axis of the inertial measurement unit (IMU) using inertial odometry. ,in The position coordinates calculated by the inertial odometer. The rotation angle along the Z-axis is calculated. For inertial pose vectors, a new recording cycle is started when the total distance traveled represented by the recorded inertial poses reaches a certain threshold or the number of inertial poses in the list exceeds a preset threshold.

[0026] Preferably, the transformation matrix between the two inertial poses in step S2 is... The calculation method is as follows:

[0027] .

[0028] Preferably, the transformation matrix between the poses of the two lidar sensors in step S2 The calculation method is as follows:

[0029] .

[0030] Preferably, in step S2, the position difference determination involves calculating the Euclidean distance between the two translation vectors. The implementation method is as follows: .

[0031] Preferably, in step S2, the rotation difference determination involves... The implementation method is as follows: .

[0032] Preferably, in the initial repositioning of step S3, the verification is successful if the position difference judgment threshold and the rotation difference judgment threshold are met simultaneously after the initial repositioning.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] 1. Improve positioning reliability: By combining the high precision of inertial odometry over short distances with LiDAR pose information, the positioning defects of a single sensor in complex environments are effectively compensated. When LiDAR positioning is interfered with, positioning loss can be detected and repositioned in a timely manner, thus improving the reliability of robot positioning.

[0035] 2. Enhanced environmental adaptability: To address the localization challenges caused by similar scenarios in complex environments, this method does not rely on the uniqueness of environmental features. It can achieve localization status judgment and relocalization simply by comparing the pose data of inertial odometry and lidar, thereby enhancing the robot's adaptability in complex environments.

[0036] 3. Reduced cost and complexity: Compared to introducing high-precision but high-cost positioning sensors (such as high-precision GPS and UWB positioning devices), this method is based on the robot's existing wheel speed sensors, IMU, and LiDAR. Repositioning is achieved through algorithm optimization, which improves the robot's positioning performance and reduces system complexity without increasing hardware costs. Attached Figure Description

[0037] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the overall structure and process of a wheeled robot relocation method based on inertial odometry.

[0039] Figure 2 A lost positioning detection map in a wheeled robot relocalization method based on inertial odometry;

[0040] Figure 3 This is a flowchart of a relocation method for a wheeled robot based on inertial odometry. Detailed Implementation

[0041] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0042] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0043] In practice, current mainstream robot localization technologies each have their limitations. While lidar-based localization technology can provide relatively accurate pose information in most cases, in complex environments, the reflected signals of lidar may be interfered with, such as by specular reflection and multipath effects, leading to a gradual accumulation of localization errors and eventual loss of positioning. Vision-based localization methods are easily affected by factors such as changes in lighting and occlusion; in dim lighting or when the target is obscured, it is difficult to obtain reliable localization data stably. Inertial odometry (IOM) calculates robot pose using data measured by wheel speed sensors and an IMU (Inertial Measurement Unit). Although cumulative errors exist, its accuracy is sufficient for certain localization requirements within a short distance, such as within 5 meters. Therefore, utilizing this characteristic of IOM to assist in robot relocalization in complex environments is of great significance. Therefore, this invention proposes a wheeled robot relocalization method based on inertial odometry, the specific implementation process of which is as follows: Figure 1 As shown, leveraging the high accuracy of inertial odometry over short distances, the robot determines whether it has lost localization by comparing the LiDAR pose and the inertial pose, and relocalizes using the inertial pose when localization is lost. Specifically, the LiDAR pose is calculated from environmental data scanned by the LiDAR using a visual algorithm; the inertial pose is calculated from wheel speed data collected by wheel speed sensors combined with the θ angle measured by the IMU, using a fusion algorithm.

[0044] First, data recording is performed. During the robot's movement, the control system maintains two key data lists in real time: the LiDAR pose list and the LiDAR pose list. Triggered at fixed intervals d, whenever the robot moves a distance d, the pose (including position coordinates (x, y) and Z-axis rotation angle θ) calculated by the LiDAR using the robot pose localization algorithm is recorded in this list. The algorithm includes vision-based, LiDAR SLAM, and GPS-based localization algorithms. Let the LiDAR pose vector be... Where x and y represent the robot's position coordinates in the two-dimensional plane, and θ represents the robot's rotation angle around the Z-axis. This is the LiDAR pose vector. A new recording cycle begins when the total distance traveled represented by the recorded pose data reaches a 5-meter threshold, or when the number of poses in the list exceeds a preset threshold. Inertial pose list: Similar to the LiDAR pose list, it is triggered by a distance interval d. It calculates the robot's current inertial pose by fusing wheel speed and IMU angle θ using inertial odometry. Record it in this list. Inertial pose vector ,in The position coordinates calculated by the inertial odometer. The rotation angle along the Z-axis is calculated. For inertial pose vectors, a new recording cycle is started when the total travel distance represented by the recorded inertial pose reaches a threshold of 5 meters or the number of inertial poses in the list exceeds a preset threshold.

[0045] The new recording cycle for the lidar pose list and inertial pose list is achieved by using a rolling recording mechanism to maintain a fixed length. When new data is written, the oldest recorded data is automatically removed, and the list content is continuously updated.

[0046] Next, location loss detection is performed, and the steps are as follows: Figure 2 As shown, during the robot's operation, the control system periodically updates the lidar pose vector list data. and inertial pose vector list data The data is analyzed to determine whether the robot has lost its location. The specific steps are as follows:

[0047] Calculate the inertial pose transformation matrix: from the inertial pose vector list data Select the inertial pose vector that is closest to the current time. and the inertial pose vector furthest from the current time. Calculate the transformation matrix between two inertial poses. The transformation matrix The calculation method is as follows: This matrix describes the robot's... Position to The relationship between the pose and the positional translation and rotation about the Z-axis in a two-dimensional plane is shown in the matrix. The first two columns of the first and second rows are composed of trigonometric function values ​​of the difference between two angles, reflecting the rotational change; the third column contains the positional coordinate difference, reflecting the translational change.

[0048] Calculate the lidar pose transformation matrix: from lidar pose vector list data Select the lidar pose vector closest to the current time. and the lidar pose vector furthest from the current time Calculate the transformation matrix between the poses of the two lidar sensors. The transformation matrix The calculation method is as follows: This matrix also reflects the changes in the position and rotation around the Z-axis of the lidar during the corresponding time period.

[0049] Then, a location loss determination is performed by comparing the transformation matrix. and The system determines whether the robot has lost its localization based on two dimensions: position and rotation. This includes position difference assessment, rotation difference assessment, and a comprehensive assessment to determine whether the robot has lost its localization and requires a redirection operation.

[0050] Among them, positional difference judgment: extraction and The translation parts in the matrix are denoted as follows: Translation vector and Translation vector Calculate the Euclidean distance between two translation vectors. The Euclidean distance The implementation method is as follows: ,like If the position exceeds a preset threshold, it is determined as a first positioning loss. Rotation difference judgment: Calculate using the Z-axis rotation angle value used in the transformation matrix. and The difference in the corresponding rotation angle is denoted as rotation difference. The rotational difference The implementation method is as follows: ;like If the rotation threshold is exceeded, it is determined to be a second positioning loss. Comprehensive judgment: If both the first and second positioning losses are determined simultaneously, it is determined that the robot has experienced positioning loss, and a repositioning operation is performed.

[0051] Relocation operations include initial relocation and forced relocation, and the implementation process is as follows: Figure 3 As shown in the diagram. Initial repositioning involves the control system immediately sending a stop command upon detecting a loss of robot positioning, halting the robot's current movement and preventing collisions or other dangerous situations caused by incorrect positioning. Subsequently, a pre-defined repositioning interface (integrating various repositioning algorithms, such as global map-based matching and particle filter repositioning algorithms) is invoked to perform the repositioning operation. After repositioning, the lost-positioning detection method described above is used again to verify the repositioned pose. Passing the verification means that both the positional difference threshold and the rotational difference threshold must be met after the initial repositioning; otherwise, the initial repositioning is considered successful. If the verification passes, meaning the repositioned pose meets the pre-defined threshold requirements for both positional and rotational differences, the positioning is considered successful, and the robot can continue operating normally according to its predetermined task.

[0052] Forced relocation is performed if the initial relocation fails, starting from... Extract the lidar pose vector that is furthest from the current time. Using the previously obtained inertial pose transformation matrix The updated latest lidar pose is calculated using matrix multiplication. The calculation method is as follows: , position As the target pose for the robot's forced repositioning, the robot's movement is controlled to... Forced repositioning is completed, allowing the robot to return to a relatively reasonable position based on the short-distance accuracy guaranteed by inertial odometry.

[0053] Finally, an early warning operation is triggered. After forced relocation, the robot continues to travel a preset distance. If a loss of positioning is detected again during this process, it indicates that the robot may be in a complex dynamic environment with numerous dynamic obstacles or drastic changes in environmental characteristics, or that the robot's sensors or positioning system have other unknown serious malfunctions. To avoid accidents, the robot immediately stops moving and triggers an alarm mechanism, sending alarm information to the operator via audible and visual alarms and network communication. Simultaneously, it records current operating data for subsequent manual intervention and troubleshooting.

[0054] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A repositioning method for a wheeled robot based on inertial odometry, characterized in that, Includes the following steps: S1. Data Recording: During the robot's movement, two key data lists are maintained in real time, including the LiDAR pose list and the inertial pose list; LiDAR pose list: Triggered at fixed distance intervals, it records the LiDAR pose calculated by the robot pose localization algorithm. When the total travel distance meets the set conditions, a new recording cycle is started. Inertial pose list: Triggered at fixed distance intervals, the inertial pose calculated by the inertial odometry is recorded. When the total travel distance meets the set conditions, a new recording cycle is started. S2. Location Loss Detection: Periodically analyze the data from both lists to determine if the robot has lost its location; the specific steps are as follows: Calculate the inertial pose transformation matrix: from the inertial pose vector list data Select the inertial pose vector that is closest to the current time. and the inertial pose vector furthest from the current time. Calculate the transformation matrix between two inertial poses. ; Calculate the lidar pose transformation matrix: from lidar pose vector list data Select the lidar pose vector closest to the current time. and the lidar pose vector furthest from the current time Calculate the transformation matrix between the poses of the two lidar sensors. ; Location loss detection: by comparing the transformation matrix and The system determines whether the robot has lost its localization based on both position and rotation dimensions. This includes position difference judgment, rotation difference judgment, and comprehensive judgment to determine whether the robot has lost its positioning and needs to be redirected. Location difference judgment: extraction and The translation parts in the matrix are denoted as follows: Translation vector and Translation vector Calculate the Euclidean distance between two translation vectors. ;like If the location exceeds a preset threshold, it is determined as the first location loss; Rotation difference determination: Calculate the rotation angle value of the Z-axis used in the transformation matrix. and The difference in the corresponding rotation angle is denoted as rotation difference. ;like If the rotation threshold is exceeded, it is determined that the second positioning is lost; Comprehensive judgment: If both the first and second positioning are determined to be lost, it is determined that the robot has lost positioning and a repositioning operation is performed. S3. Relocation operation: including initial relocation and forced relocation; Initial relocation: After detecting a loss of localization, the robot stops moving, calls the relocation interface to perform relocation, and verifies the relocation result. If the relocation is successful, the localization is successful. Forced relocation: If the initial relocation fails, a forced relocation operation is performed, starting from... Extract the lidar pose vector that is furthest from the current time. Using the previously obtained inertial pose transformation matrix The updated latest lidar pose is calculated using matrix multiplication. The calculation method is as follows: , position As the target pose for the robot's forced relocation; S4. Trigger Warning: After forced repositioning, if the robot detects a loss of positioning again when it has traveled a preset distance, it will stop moving and trigger an alarm via sound, light, or network, while recording the running data.

2. The repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, The robot pose localization algorithm in step S1 includes algorithms based on vision, laser SLAM, and GPS localization.

3. The repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, In step S1, the new recording cycle for the lidar pose list and the inertial pose list is achieved by using a rolling recording mechanism to maintain a fixed length. When new data is written, the oldest recorded data is automatically removed, and the list content is continuously updated.

4. The repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, The specific method for recording the lidar pose list in step S1 is as follows: With a fixed distance interval d as the trigger condition, whenever the robot moves a distance of d, At that time, the lidar pose vector calculated by the robot pose localization algorithm at the current moment is used. Recorded in a list, where, This is the pose vector of the lidar. This represents the robot's position coordinates in a two-dimensional plane. This indicates the robot's rotation angle around the Z-axis. A new recording cycle is started when the total travel distance represented by the recorded LiDAR poses reaches a certain threshold or the number of LiDAR poses in the list exceeds a preset threshold.

5. The repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, The data recording method for the inertial pose list in step S1 is as follows: With a fixed distance interval d as the trigger condition, whenever the robot moves a distance of d, the robot's inertial pose vector at the current moment is calculated by fusing wheel speed and the rotation angle of the Z-axis of the inertial measurement unit (IMU) using inertial odometry. ,in The position coordinates calculated by the inertial odometer. The rotation angle along the Z-axis is calculated. For inertial pose vectors, a new recording cycle is started when the total distance traveled represented by the recorded inertial poses reaches a certain threshold or the number of inertial poses in the list exceeds a preset threshold.

6. The repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, The transformation matrix between the two inertial poses in step S2 The calculation method is as follows: 。 7. The repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, The transformation matrix between the poses of the two lidars in step S2 The calculation method is as follows: 。 8. A repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, In step S2, the position difference determination involves calculating the Euclidean distance between the two translation vectors. The implementation method is as follows: .

9. A repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, In step S2, the rotation difference judgment involves rotation difference. The implementation method is as follows: .

10. A repositioning method for a wheeled robot based on inertial odometry according to claim 1, characterized in that, In the initial repositioning of step S3, the verification is successful when the position difference judgment threshold and the rotation difference judgment threshold are met simultaneously after the initial repositioning.

Images