Robot large-range positioning test system and method based on multiple laser radars

Abstract

The invention belongs to the technical field of robot positioning test, and discloses a robot large-range positioning test system and method based on multiple laser radars, and the test system, and the system comprises a sensing module which forms an array through the laser radars, and is used for covering a preset test area; the data transmission module is connected with the sensing module and is used for collecting and transmitting the point cloud data acquired by the laser radar; the processing module is connected with the data transmission module and used for processing the point cloud data; the time service module is connected with the sensing module and / or the processing module and is used for performing time synchronization on the point cloud data acquired by the laser radar; the target module is installed on the robot to be tested, and the target module comprises a polyhedral reflective target with a feature structure; the invention provides an efficient, accurate and flexible solution with a wide application range for testing the positioning performance of the robot, and fills the blank in the field of large-range and high-precision positioning testing of the robot in the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot positioning testing, and particularly relates to a large-range robot positioning testing system and method based on multiple lidars. Background Art

[0002] Robot positioning is the basis for robot autonomous navigation and task completion, and it is crucial to accurately and reliably test its performance. Existing robot positioning testing equipment and methods have many limitations.

[0003] Traditional measuring devices such as meter rulers and laser rangefinders can only perform single-point or limited-point measurements and cannot continuously track and measure the movement trajectory of robots. Although laser trackers and motion capture systems can measure trajectories, their application scope is mainly limited to indoor environments, and the measurement range is usually limited. For example, the measurement range of the cameras used in motion capture systems mostly does not exceed 20m. When it is necessary to test the positioning performance of robots in a large-range space, a large number of cameras need to be deployed for the motion capture system, resulting in a complex system, cumbersome installation, and high cost. In addition, the motion capture system is easily affected by the reflective ground and generates measurement noise. The laser tracker relies on continuous tracking of the target. When the measured robot moves quickly or turns around, the target is easily blocked, resulting in tracking loss and affecting the continuity of the test.

[0004] GNSS (Global Navigation Satellite System) positioning equipment is suitable for outdoor large-range testing, but its accuracy is easily affected by satellite signals, multipath effects, etc., and it cannot be used in indoor environments. At the same time, meter rulers, laser rangefinders, and laser trackers usually can only measure single targets and are not applicable to scenarios where multiple robots need to be tested simultaneously.

[0005] Therefore, the existing technology has deficiencies in realizing large-range, high-precision, indoor-outdoor universal, continuous trajectory measurement, and multi-target testing of robots, and it is difficult to meet the comprehensive and efficient testing requirements for robot positioning performance in complex environments. Summary of the Invention

[0006] The present invention aims to solve at least to some extent the above technical problems. For this purpose, the object of the present invention is to provide a large-range robot positioning testing system and method based on multiple lidars.

[0007] The technical solution adopted by the present invention is as follows:

[0008] One aspect of the present invention provides a large-range robot positioning testing system based on multiple lidars, including:

[0009] A perception module, which consists of an array of at least three lidars with three-dimensional scanning capabilities, and the lidars are configured to cover a preset test area;

[0010] The data transmission module is connected to the perception module and is used to collect and transmit the point cloud data acquired by the lidar;

[0011] A processing module, connected to the data transmission module, for processing point cloud data;

[0012] A timing module, connected to the perception module and / or the processing module, is used to synchronize the time of the point cloud data acquired by the laser radar;

[0013] A target module is installed on the robot to be tested, and the target module includes a polyhedral reflective target with a characteristic structure;

[0014] The processing module is configured to: receive the point cloud data of the laser radar synchronized by the timing module from the data transmission module; perform spatial registration on the point cloud data of the laser radar to establish a unified coordinate system; identify the point cloud cluster corresponding to the target module from the point cloud data in the unified coordinate system; calculate the real-time position and posture of the target module based on the identified point cloud cluster and the pre-stored point cloud template of the target module; and fuse the position and posture solution results from multiple laser radars to output the positioning test results of the robot.

[0015] Optionally, the perception module is composed of at least three 1550nm multi-line automotive-grade laser radars with three-dimensional scanning capabilities, and the measurement distance of the laser radar is greater than 100m.

[0016] Optionally, the lidars in the perception module are deployed according to an optimized layout predetermined based on the test area and lidar parameters.

[0017] Optionally, the data transmission module is configured with an industrial-grade Gigabit Ethernet switch that supports the IEEE 802.3at PoE power supply standard.

[0018] Optionally, the processing module is equipped with a GPU accelerated computing server and a dedicated acceleration card for point cloud processing.

[0019] Optionally, the timing module supports GPS, PTP and gPTP time synchronization.

[0020] Optionally, the spatial registration uses an ICP (Iterative Closest Point) algorithm to achieve point cloud stitching to establish a unified coordinate system.

[0021] Optionally, identifying the point cloud cluster corresponding to the target module is achieved using a clustering algorithm.

[0022] Optionally, the real-time position and posture of the target module are solved by using a modified PointPillars algorithm and a pre-stored target point cloud template for model matching.

[0023] Optionally, the position and attitude calculation results from multiple lidars are fused using a Kalman filter to achieve data fusion and spatio-temporal alignment.

[0024] Another aspect of the present invention provides a large-scale positioning test method for a robot based on multiple lidars, which is used to perform positioning tests on a robot equipped with a target module. The method includes the following steps:

[0025] A. Determine the optimized deployment layout of at least three 3D lidars within the test area. The layout constructs a coverage model based on the 3D model of the test area and lidar parameters, and determines the installation positions and attitudes of the lidars that meet the coverage requirements through an optimization algorithm or iterative adjustment.

[0026] B. Install and calibrate at least three 3D lidars according to the optimized deployment layout. The calibration includes spatially registering the point cloud data from different lidars to establish a unified coordinate system.

[0027] C. Conduct a motion test on the robot to be tested within the test area.

[0028] D. During the motion test, real-time collect the point cloud data from at least three lidars and perform time synchronization.

[0029] E. Process the time-synchronized point cloud data, including identifying the point cloud clusters corresponding to the target module on the robot in the unified coordinate system.

[0030] F. Based on the identified point cloud clusters and the pre-stored point cloud template of the target module, calculate the real-time position and attitude of the target module.

[0031] G. Fuse the position and attitude calculation results of the same target module from multiple lidars.

[0032] H. Output the fused position and attitude as the real-time positioning test result of the robot.

[0033] Optionally, step A includes: creating a lidar model, importing the 3D CAD model of the test area, and setting the test area boundary; intercepting one or more working planes in the environment model; dividing the working plane into multiple small working planes; for the i-th small working plane, establish a point cloud density objective function where Q i is the point cloud density of the i-th small working plane, M i is the total number of pixel points of the i-th small working plane, and L i is the number of pixel points covered by the laser point cloud in the simulation; adjust the position and attitude of the lidar in the model until the point cloud density of all small working planes meets the preset requirements; record the position and attitude of the lidar that meets the requirements as the optimized deployment layout.

[0034] Optionally, the lidar parameters include horizontal field of view, vertical field of view, and angular resolution.

[0035] Optionally, the spatial registration in step B uses the ICP (Iterative Closest Point) algorithm for point cloud stitching.

[0036] Optionally, step B also includes sampling the point cloud data and performing point cloud density verification; and for areas with blind spots or sparse point cloud density, adding lidars for small-area blind spot compensation.

[0037] Optionally, the identification of the point cloud cluster corresponding to the target module in step E is achieved using a clustering algorithm.

[0038] Optionally, step F is achieved by using an improved PointPillars algorithm for model matching with a pre-stored target point cloud template.

[0039] Optionally, step G uses a Kalman filter to achieve data fusion and spatio-temporal alignment.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. Wide coverage and strong scalability: By using multi-lidar fusion perception, the measurement distance of a single lidar can reach more than 100m, and through optimized layout and point cloud stitching, the coverage range can be infinitely expanded. For the same number of lidars, the coverage range is 3-5 times that of the same number of vision camera systems, solving the problem of limited measurement range of existing systems.

[0042] 2. High-precision positioning: By using high-resolution 3D lidar and optimized point cloud processing, model matching, and data fusion algorithms, centimeter-level or even sub-centimeter-level positioning accuracy can be achieved. For example, the horizontal positioning accuracy reaches ±2mm at a distance of 100m.

[0043] 3. Continuous trajectory measurement and high-speed tracking: The lidar can collect point cloud data at high speed. Combining with efficient processing algorithms, it supports continuous tracking measurement of moving targets and can support high-speed moving target tracking at speeds of up to 30km / h.

[0044] 4. Strong anti-environmental interference ability: The lidar actively emits laser light and is insensitive to environmental illumination. It has stronger robustness compared to vision systems for reflective ground, etc. By identifying specific targets installed on the robot and shielding other object information in the environment, it effectively resists interference. The system is not affected by other objects in the environment (such as walls, furniture, other robots, etc.) and is insensitive to ground materials (such as reflective ground), and can be stably used in indoor and outdoor environments.

[0045] 5. Support multi-target testing: Install targets on each robot to be tested. The system can simultaneously identify, track, and measure the positions and postures of multiple different targets, achieving multi-target synchronous testing.

[0046] 6. Deployment optimization and dynamic blind spot filling: The proposed dynamic optimization layout algorithm can guide the efficient deployment of lidars, and further improve the system coverage quality and reliability through point cloud density verification and dynamic blind spot filling. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is the system block diagram of the large-scale robot positioning and testing system based on multi-lidars of the present invention.

[0048] Figure 2 is the point cloud map completed by stitching of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] It should be understood that in the embodiments, the functions / actions that appear may be different from the order in which the drawings appear. For example, depending on the functions / actions involved, they may actually be executed substantially concurrently, or sometimes the two continuously shown figures may be executed in the reverse order.

[0051] As Figure 1 and Figure 2 shown, the large-scale robot positioning and testing system based on multi-lidars of the present invention includes the following modules:

[0052] Perception module: An array is composed of n (n≥3) lidars with three-dimensional scanning capabilities. Preferably, automotive-grade lidars with a 1550nm wavelength band and multiple lines (e.g., 180 lines) are used. The 1550nm wavelength band is selected because it has a high eye safety threshold, allowing for a longer measurement distance (e.g., exceeding 100m) with a higher transmission power and strong anti-environmental light ability; the multi-line structure provides rich vertical direction point cloud information, which helps to build a fine three-dimensional environment or target model; the automotive-grade design ensures stability and reliability in complex environments and supports wide applications indoors and outdoors. The installation positions and postures of the lidar array are determined through pre-optimized calculations to ensure effective coverage of the entire test area and sufficient point cloud density.

[0053] Data Transmission Module: Configure an industrial-grade gigabit Ethernet switch. The lidar generates a large amount of point cloud data, and the gigabit Ethernet ensures the data transmission bandwidth requirement. The industrial-grade design enhances the environmental adaptability of the system. It supports the IEEE802.3at PoE (Power over Ethernet) power supply standard, simplifies the on-site wiring, and reduces the deployment complexity.

[0054] Processing Module: Equip with a high-performance computing platform, such as a GPU-accelerated computing server and a dedicated point cloud processing acceleration card. The processing of point cloud data, especially the registration, filtering, target recognition, and pose solution of three-dimensional point clouds, involves a huge amount of computation and requires powerful parallel computing capabilities. The GPU and the acceleration card can significantly improve the processing efficiency and meet the real-time requirements.

[0055] Timing Module: Support multiple high-precision time synchronization protocols, such as GPS, PTP (Precision Time Protocol, IEEE 1588), and gPTP (generalized Precision Time Protocol, IEEE 802.1AS). The data collected by multiple lidars need to be fused and processed under a unified time reference. High-precision time synchronization is the key to achieving accurate spatio-temporal registration and moving target tracking.

[0056] Target Module: Installed on the body of the robot to be measured. The target module includes a carefully designed polyhedron retroreflective target, and its surface is provided with specific feature structures (for example, specific reflective material patterns, prism combinations, or three-dimensional geometric shapes). The polyhedron structure is selected to increase the probability of being illuminated by at least one lidar at different angles. The reflective material or prism can enhance the detection effect of the lidar on the target, making it more easily recognizable in the point cloud. The feature structure provides rich geometric features for accurately extracting key points or shape information from the point cloud cluster for high-precision model matching and pose solution.

[0057] The core technical features and specific implementation methods of the present invention are mainly reflected in the following aspects:

[0058] (1) Multi-lidar fusion perception and point cloud stitching:

[0059] By deploying multiple lidars, the measurement range of a single lidar is expanded, and full coverage of the test area is ensured through optimized layout. To process multi-lidar data within a unified spatial framework, accurate spatio-temporal registration of point clouds is required. Spatial registration (i.e., point cloud stitching) is to transform the point clouds collected by different lidars in their own coordinate systems to the same global coordinate system. During the system installation and calibration phase, mature point cloud registration algorithms such as ICP (Iterative Closest Point) are usually adopted to achieve the initial stitching of the point clouds of multiple lidars and establish a globally unified static coordinate system. During the real-time positioning phase, the real-time point clouds from different lidars are first transformed to this unified coordinate system according to the calibrated transformation relationship for subsequent processing. Accurate time synchronization (through a timing module) is crucial for processing moving targets and fusing multi-lidar data, ensuring that the point clouds collected by different lidars at different times can be accurately aligned on the time axis.

[0060] (2) Dynamic optimization layout algorithm:

[0061] The present invention proposes an optimization method for determining the optimal installation position and attitude of lidars in a sensing module. This method first creates a mathematical model of the lidar (including parameters such as its field of view angle, ranging ability, angular resolution, etc.) and imports the 3D CAD model of the test area, and sets the boundaries of the test area. Considering that the robot mainly moves on the ground or within a certain height range, one or more "working planes" or "working voxel spaces" can be intercepted in the environmental model to represent the areas where the targets may appear. These working planes or voxel spaces are divided into smaller units (such as pixel point grids). For each unit (e.g., the i-th small working plane), a point cloud density objective function is defined where Q i is the point cloud density of the i-th small working plane, M i is the total number of pixel points of the i-th small working plane, and L i is the number of pixel points covered by the laser point cloud in the simulation. L i and M i can be obtained by setting different grayscales or colors for the test area and the laser points in the simulation, and then statistically calculating the pixel point colors by a computer. The optimization objective is to find a set of positions (x, y, z) and attitudes of the lidars such that the point cloud density Q i of all working planes or voxel units all meet the preset minimum requirements (e.g., sufficient to ensure stable detection of the target and pose solution in this area). Currently, this adjustment process can be carried out interactively in simulation software, or an automatic search algorithm based on genetic algorithms, particle swarm algorithms or other optimization techniques can be designed to quickly solve for the optimal layout. This optimization process ensures that the system can provide sufficient coverage and data quality with limited resources.

[0062] Where: x, y are the planar projection coordinates of the lidar in the test area; z is the height of the lidar from the ground of the test area; The heading angle, pitch angle, and roll angle of the lidar in space.

[0063] (3) Point cloud registration algorithm:

[0064] In addition to the initial point cloud stitching (such as ICP) used to establish a global unified coordinate system, during the real-time operation of the system or in specific calibration / maintenance phases, it may be necessary to further register the point clouds of a single lidar or part of the lidars. The KISS-Matcher mentioned in the technical disclosure may be used in specific scenarios. For example, after dynamic blind spot filling, the newly added lidar point cloud can be accurately registered into the existing coordinate system, or it can be used to optimize the initial calibration. The specific selection of the registration algorithm and the application order depend on the specific implementation details and the requirements for accuracy and efficiency. The key is to ensure that all real-time collected point clouds can be accurately transformed into a unified coordinate system for subsequent processing.

[0065] (4) Model matching positioning method:

[0066] This is the core of achieving high-precision positioning. During the real-time positioning phase, the system uses algorithms such as clustering to identify the point cloud clusters corresponding to the target modules on the robot from the point cloud data in the unified coordinate system. Then, a modified PointPillars algorithm or a similar deep learning / feature matching algorithm is used to match the real-time obtained target point cloud clusters with the precisely made and stored precise three-dimensional point cloud templates of the target modules. The modified PointPillars algorithm (or other point cloud feature learning and matching algorithms) can extract robust features from the point cloud data and, through matching with the template features, quickly and accurately calculate the real-time three-dimensional position and attitude (6 degrees of freedom: x, y, z, roll, pitch, yaw) of the target module in the unified coordinate system. Centimeter-level positioning accuracy mainly depends on high-resolution lidars, the fineness of target features, and precise model matching algorithms.

[0067] (5) Data fusion:

[0068] Since multiple lidars may simultaneously observe the same target module, or for the need to improve robustness and smooth the measurement results, it is necessary to fuse the positioning results from different lidars or at different times. The present invention uses a Kalman filter or other state estimation algorithms to achieve data fusion. The Kalman filter can effectively combine the measurement data of different sensors (here, the calculation results of different lidars) and consider the motion model of the system to output a more stable and accurate posterior estimate. At the same time, the Kalman filter or other filtering methods are also used to achieve the spatio-temporal alignment and smoothing processing of the measurement data at different times.

[0069] The large-scale positioning test method for a robot based on multiple lidars mainly includes the following three stages:

[0070] (1) System deployment stage:

[0071] Environmental modeling: Create a mathematical model of the lidar to be used, including key parameters such as its field of view angle, ranging range, angular resolution, etc. Import the 3D CAD model or other forms of 3D representation of the actual area to be tested, and set the precise boundaries of the test area in the model.

[0072] Lidar layout optimization: In the above environmental model, according to the type and movement range of the robot to be tested, intercept one or more "working planes" or "working voxel spaces" within a certain height range, representing the areas where the targets on the robot may appear. Divide these working planes / voxel spaces into small units (such as pixel grids). For each unit i, calculate the density of the laser point cloud coverage in this simulation layout where Q i is the point cloud density of the i-th small working plane, M i is the total number of pixel points of the i-th small working plane, and L i is the number of pixel points covered by the laser point cloud in the simulation. By adjusting the positions (x, y, z) and postures of each lidar in the model and recalculating Q i , until the Q i of all key areas (working planes / voxel spaces) meet the preset minimum point cloud density requirements. Record the positions and posture parameters of each lidar at this time as the basis for subsequent actual installation. This process is the key to ensuring effective system coverage and data quality. Among them: x, y: the planar projection coordinates of the lidar in the test area; z: the height of the lidar from the ground of the test area; the heading angle, pitch angle, and roll angle of the lidar in space.

[0073] Record the optimized layout parameters: Save the determined lidar position and posture parameters for guiding on-site installation.

[0074] (2) Installation and calibration stage:

[0075] Lidar installation and deployment: According to the precise position and posture parameters of the lidar calculated above, install and deploy each lidar of the perception module on-site in the actual test area. Ensure that the installation position and direction are as close as possible to the results of the optimized calculation.

[0076] Point cloud stitching (initial calibration): After the lidar is installed, static point cloud data covering the entire test area is collected. Using precise point cloud registration algorithms such as ICP (Iterative Closest Point), the rigid body transformation matrix from each lidar coordinate system to the unified global coordinate system is calculated. Through these transformation matrices, the point clouds from different lidars are transformed into the same coordinate system, completing the initial stitching of multi-lidar point clouds and establishing a globally unified reference coordinate system.

[0077] Density verification: After completing the initial point cloud stitching, the point cloud data in the unified coordinate system is sampled and analyzed to verify whether the point cloud density in the key areas (working plane / voxel space) meets the requirements set in the optimization layout stage.

[0078] Dynamic lidar blind spot filling: If density verification finds blind spots or point cloud density below the requirements in certain areas, additional lidars are added near these areas for local blind spot filling, and the steps of "point cloud stitching (initial calibration)" are repeated for point cloud stitching and calibration until the coverage and point cloud density of the entire test area meet the test requirements.

[0079] (3) Real-time positioning stage:

[0080] Robot motion test: The robot to be tested is placed in the test area and performs predefined test tasks (e.g., driving along a specified path, docking, etc.).

[0081] Real-time point cloud acquisition and synchronization: Each lidar of the perception module real-time collects the point cloud data of the test area and sends it to the processing module through the data transmission module. The timing module provides high-precision timestamps for all collected point cloud data to ensure the time synchronization of the data.

[0082] Target recognition: The processing module receives the multi-lidar point cloud data after time synchronization. In the unified coordinate system, using methods such as clustering algorithms, the point cloud clusters corresponding to the target modules installed on the robot are quickly and effectively recognized from the complex environmental point clouds. The clustering algorithm can group the points belonging to the same target according to the spatial proximity between points.

[0083] Pose solution: For the recognized target point cloud clusters, the improved PointPillars algorithm or other model matching algorithms are used to match them with the precise point cloud templates of the target modules prepared in advance. This process solves the three-dimensional position and pose (6 degrees of freedom) of the target module in the unified coordinate system at the current moment. The "improved" PointPillars may refer to being optimized for identifying specific artificial targets rather than general targets, or combining traditional feature matching methods to improve the accuracy and robustness of target pose solution.

[0084] E. Data fusion: Different lidars may simultaneously "see" and calculate the pose of the same target. To improve the positioning accuracy and robustness, data fusion algorithms such as the Kalman filter are used to fuse the pose calculation results from different lidars or consecutive frames, taking into account the confidence levels of the respective calculation results, and outputting a smoother and more accurate real-time positioning test result of the robot. The Kalman filter can also be used for spatio-temporal alignment and filtering of the measurement results.

[0085] F. Output test results: Output the fused real-time position and attitude data of the robot to the tester or recording system for analyzing the positioning performance of the robot.

[0086] Through the above system and method, the present invention provides an efficient, accurate, flexible and widely applicable solution for the robot positioning performance test, filling the gap in the prior art in the field of large-scale and high-precision robot positioning tests.

[0087] The present invention is not limited to the above optional embodiments, and any person can obtain other various forms of products under the inspiration of the present invention. However, no matter what changes are made in its shape or structure, as long as the technical solutions fall within the scope defined by the claims of the present invention, they are all within the protection scope of the present invention.

Claims

1. A large-scale positioning test system for a robot based on multiple lidars, characterized in that, Comprising: A perception module, which consists of an array of at least three lidars with 3D scanning capabilities, and the lidars are configured to cover a preset test area; A data transmission module, connected to the perception module, for collecting and transmitting the point cloud data obtained by the lidars; A processing module, connected to the data transmission module, for processing the point cloud data; A timing module, connected to the perception module and / or the processing module, for time synchronization of the point cloud data obtained by the lidars; A target module, installed on the robot to be tested, and the target module includes a polyhedral reflective target; The processing module is configured to: Receive the point cloud data of the lidars synchronized by the timing module from the data transmission module; perform spatial registration on the point cloud data of the lidars to establish a unified coordinate system; and in the unified coordinate system, identify the point cloud cluster corresponding to the target module from the point cloud data; Based on the identified point cloud cluster and the pre-stored point cloud template of the target module, calculate the real-time position and attitude of the target module; Fuse the position and attitude calculation results from multiple lidars, and output the positioning test result of the robot.

2. The large-scale positioning test system for a robot according to claim 1, wherein: The perception module consists of at least three 1550nm multi-line automotive-grade lidars with 3D scanning capabilities, and the measurement distance of the lidars is greater than 100m.

3. The robot large-scale positioning test system according to claim 1, wherein: The lidars in the perception module are deployed according to an optimized layout determined in advance based on the test area and the lidar parameters.

4. The robot large-scale positioning test system according to claim 1, characterized in that: The spatial registration uses the ICP algorithm to achieve point cloud stitching to establish the unified coordinate system.

5. The robot large-scale positioning test system according to claim 1, characterized in that: Identifying the point cloud cluster corresponding to the target module is achieved by using a clustering algorithm; calculating the real-time position and attitude of the target module is achieved by using an improved PointPillars algorithm for model matching with the pre-stored target point cloud template; and fusing the position and attitude calculation results from multiple lidars is achieved by using a Kalman filter for data fusion and spatio-temporal alignment.

6. A large-scale positioning test method for a robot based on multiple lidars, which is used to perform positioning tests on a robot equipped with a target module, is characterized in that, Including the steps: A. Determine the optimized deployment layout of at least three 3D lidars within the test area. The layout constructs a coverage model based on the 3D model of the test area and the lidar parameters, and determines the installation positions and attitudes of the lidars that meet the coverage requirements through an optimization algorithm or iterative adjustment; B. Install and calibrate the at least three 3D lidars according to the optimized deployment layout. The calibration includes performing spatial registration on the point cloud data from different lidars to establish a unified coordinate system; C. Conduct a motion test on the robot to be tested within the test area; D. During the motion test, collect the point cloud data from the at least three lidars in real time and perform time synchronization; E. Process the time-synchronized point cloud data, including identifying the point cloud cluster corresponding to the target module on the robot in the unified coordinate system; F. Based on the identified point cloud cluster and the pre-stored point cloud template of the target module, calculate the real-time position and attitude of the target module; G. Integrate the position and attitude solution results of the same target module from multiple lidars; H. Output the fused position and attitude as the real-time positioning test result of the robot.

7. The method for large-scale positioning test of a robot according to claim 6, characterized in that, The step A includes: Create the lidar model, import the 3D CAD model of the test area, and set the boundary of the test area; Intercept one or more working planes in the environmental model; Divide the working plane into multiple small working planes; For the i-th small working plane, establish a point cloud density objective function where Q i is the point cloud density of the i-th small working plane, M i is the total number of pixel points of the i-th small working plane, and L i is the number of pixel points covered by the laser point cloud in the simulation; Adjust the position and attitude of the lidar in the model until the point cloud density of all small working planes meets the preset requirements; Record the lidar position and attitude that meet the requirements as the optimized deployment layout.

8. The method for large-scale positioning test of a robot according to claim 7, characterized in that: The lidar parameters include horizontal field of view, vertical field of view, and angular resolution.

9. The method for large-scale positioning test of a robot according to claim 7, characterized in that: The step B further includes sampling the point cloud data and verifying the point cloud density; and for areas with blind spots or sparse point cloud density, adding lidars to fill in the small areas.

10. The method for large - scale positioning test of a robot according to claim 7, wherein: The spatial registration in the step B uses the ICP algorithm for point cloud stitching; the identification of the point cloud cluster corresponding to the target module in the step E is achieved by using a clustering algorithm; the step F is achieved by using an improved PointPillars algorithm for model matching with the pre-stored target point cloud template; the step G is achieved by using a Kalman filter for data fusion and spatio-temporal alignment.

Images