Fusion positioning method of unmanned container truck in simulation environment
By using the lidar module in the unmanned card to obtain point cloud data and combining laser odometer and inertial navigation information for fusion, the problems of insufficient positioning accuracy, poor robustness and insufficient real-time performance of the unmanned card in complex simulation environments are solved, and the positioning effect of high precision, robustness and real-time performance is achieved.
Patent Information
- Application Number
- CN202510247083.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-04
- Publication Date
- 2025-06-10
AI Technical Summary
The existing positioning technology of unmanned clusters in complex simulation environments has problems such as insufficient accuracy, poor robustness and insufficient real-time performance, which is difficult to meet the needs of high-precision navigation and fast response.
The lidar module is used to continuously scan and obtain point cloud data. Through denoising, filtering, downsampling and pre-processing, geometric features are extracted and matched using a normal distribution transformation algorithm, and global optimization is carried out in combination with the graph optimization algorithm to build a global environment map, and the laser odometer and inertial navigation information are fused to achieve positioning.
It improves the positioning accuracy and robustness of unmanned clusters in complex simulation environments, enhances real-time performance, and meets the application needs of large-scale scenarios such as ports.
Smart Images
Figure CN120121052A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent driving technology, and particularly to a fusion positioning method of an unmanned container truck in a simulation environment. Background Art
[0002] In the field of port logistics, the application of unmanned container trucks is becoming increasingly widespread, and their autonomous navigation and precise positioning are the keys to achieving efficient and safe transportation. However, in a complex simulation environment, the positioning technology of unmanned container trucks still faces many challenges. Currently, the positioning methods of unmanned container trucks mainly include GPS positioning, visual positioning, laser positioning, etc. GPS positioning has a good positioning effect in open areas, but in complex environments such as ports, due to the influence of obstacles such as buildings and containers, GPS signals are easily interfered, resulting in a decrease in positioning accuracy. Visual positioning methods rely on image information captured by cameras, but in cases of light changes, occlusion, reflection, etc., the image quality will be affected, thus affecting the positioning accuracy. Laser positioning methods obtain environmental information through laser scanners, with high accuracy and stability, but in dynamic environments, such as scenes where vehicles and personnel move frequently, the laser scanner may be interfered, resulting in inaccurate positioning. In addition, the existing positioning methods of unmanned container trucks also have the following problems: one is the insufficient positioning accuracy, which is difficult to meet the requirements of high-precision navigation; the second is the poor robustness, which is easily affected by environmental factors; the third is the insufficient real-time performance, which is difficult to meet the requirements of rapid response. Therefore, how to overcome the deficiencies of the existing technology and improve the positioning accuracy, robustness and real-time performance of unmanned container trucks in a complex simulation environment has become an urgent problem to be solved. In recent years, with the rapid development of sensor technology and computer vision technology, multi-sensor fusion positioning technology has gradually become a research hotspot. However, how to effectively fuse the data of multiple sensors to achieve high-precision and high-robustness positioning of unmanned container trucks in a complex simulation environment is still a difficult and hot issue in current research. Summary of the Invention
[0003] In view of this, the present invention provides a fusion positioning method of an unmanned container truck in a simulation environment to solve the technical problems such as insufficient positioning accuracy, poor robustness and insufficient real-time performance of the existing technology.
[0004] The present invention provides a fusion positioning method for an unmanned container truck in a simulation environment. The method includes: Step 1, continuously scanning the driving path of the unmanned container truck in the simulation environment by using a lidar module to obtain point cloud data; Step 2, performing denoising, filtering, and downsampling preprocessing on the collected point cloud data to improve data quality and the efficiency of subsequent processing; Step 3, extracting specified geometric features from the preprocessed point cloud data; Step 4, according to the extracted geometric features, matching the point cloud data of adjacent frames through the normal distribution transformation algorithm to calculate the relative pose change; Step 5, using a graph optimization algorithm to globally optimize the relative pose change obtained by matching; Step 6, applying the optimized relative pose change to the point cloud data, and then stitching adjacent frame point clouds to construct a global environment map; Step 7, fusing the relative pose change provided by the laser odometer in the simulation with the absolute position information provided by the inertial navigation; Step 8, matching the fused result with the global environment map to achieve positioning.
[0005] Further, Step 1 further includes recording the data of the inertial navigation system and other auxiliary sensors.
[0006] Further, the denoising preprocessing is to perform radius filtering; the filtering preprocessing is to perform Gaussian filtering and mean filtering; the preprocessing is to perform voxel gridization.
[0007] Further, the geometric features include edges, corners, and planes.
[0008] Further, Step 4 further includes using the nearest neighbor search based on a tree data structure and parallel computing based on a vision processor as acceleration strategies to improve the matching efficiency and accuracy.
[0009] Further, the method for global optimization includes: based on the spatial relationship and time continuity factors of the point cloud data, performing algorithm integration based on the Kalman filter and the smoothing and mapping algorithm to eliminate cumulative errors and improve the positioning accuracy.
[0010] Further, Step 6 further includes: during the stitching process, maintaining the continuity and consistency of the point cloud data, and adopting an incremental and hierarchical stitching strategy to improve the stitching efficiency and accuracy.
[0011] Further, the method for fusion includes: according to the data characteristics and error distributions provided by the laser odometer and inertial navigation in the simulation, using the Kalman filter algorithm, Bayesian network algorithm, and particle filter algorithm for fusion, thereby improving the positioning accuracy and robustness.
[0012] Further, the method for matching the fused result with the global environmental map includes: according to the accuracy and resolution of the map, using methods based on point features and line features, as well as methods based on image matching and image stitching for precise navigation and positioning of the driverless container truck.
[0013] The present invention provides a fusion positioning method for a driverless container truck in a simulation environment. This fusion positioning method combines point cloud stitching, laser odometry, and inertial navigation information. By fusing data from multiple sensors, it achieves complementary advantages, improves positioning accuracy and robustness, addresses the deficiencies of existing driverless container truck positioning technologies, and improves the positioning accuracy, robustness, and real-time performance of driverless container trucks in complex simulation environments, meeting the application requirements in large-scale scenarios such as ports. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 is a schematic flow chart of a fusion positioning method for a driverless container truck in a simulation environment provided by the present invention; Figure 2 is a schematic diagram of a simulation architecture provided by the present invention; Figure 3 is a schematic diagram of sensor simulation modeling provided by the present invention; Figure 4 is a schematic flow chart of point cloud data processing provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all 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.
[0016] Embodiment 1: The present invention provides a fusion positioning method for a driverless container truck in a simulation environment. As Figure 1 shown, the method includes the following steps.
[0017] Step 1, use a lidar module to continuously scan the driving path of the driverless container truck in the simulation environment to obtain point cloud data; Step 2, perform preprocessing such as denoising, filtering, and downsampling on the collected point cloud data to improve data quality and the efficiency of subsequent processing; Step 3, extract specified geometric features from the preprocessed point cloud data; Step 4, according to the extracted geometric features, match the point cloud data of adjacent frames through the normal distribution transformation algorithm to calculate the relative pose change; Step 5: Use the graph optimization algorithm to globally optimize the relative pose changes obtained by matching. Step 6: Apply the optimized relative pose changes to the point cloud data, and then splice the adjacent frame point clouds to construct a global environmental map. Step 7: Integrate the relative pose changes provided by the laser odometer in the simulation with the absolute position information provided by the inertial navigation. Step 8: Match the integrated result with the global environmental map to achieve positioning.
[0018] The present invention provides a fusion positioning method for an unmanned container truck in a simulation environment. This fusion positioning method combines point cloud splicing, laser odometer, and inertial navigation information. By integrating data from multiple sensors, it achieves complementary advantages, improves positioning accuracy and robustness, addresses the deficiencies of existing unmanned container truck positioning technologies, and improves the positioning accuracy, robustness, and real-time performance of unmanned container trucks in complex simulation environments, meeting the application requirements in large-scale scenarios such as ports.
[0019] Embodiment 2: The present invention provides a fusion positioning method for an unmanned container truck in a simulation environment, as Figure 1 shown. The method includes the following steps.
[0020] Step 1: Use the lidar module to continuously scan the driving path of the unmanned container truck in the simulation environment to obtain point cloud data. The simulation environment refers to the virtual environment provided by the simulation platform for the research and testing of unmanned container truck autonomous driving technology. As Figure 2 and Figure 3As shown, the simulation platform is a software system that highly simulates the real port operation environment and can accurately simulate the driving path of the driverless container truck, the distribution of obstacles, weather conditions, etc. In the scenario simulation software, the corresponding sensor output modules include ultrasonic radar module, millimeter-wave radar module, lidar module, camera module, and satellite positioning module; among them, the ultrasonic radar module and the millimeter-wave radar module collect the true port road condition information in the simulation software, and encapsulate and output the true value information to the real-time simulator through the sensor module; after processing the received encapsulated sensor information, the real-time simulator forms an overall environmental perception target list and sends the target list to the actual domain controller. The point cloud data refers to a large set of three-dimensional spatial coordinate points collected by sensors such as laser scanners. These points record geometric information such as the shape, position, and size of objects and are the basic data for constructing an environmental map and realizing object recognition and positioning. Step 1 also includes recording the data of the inertial navigation system and other auxiliary sensors. The operation mode of the inertial navigation system is that the satellite module positioning in the scenario simulation software converts the true position of the container truck in the virtual scenario into a Gaussian coordinate position through relevant protocols and sends it to the real-time simulator. After processing the position data, the real-time simulator sends it to the domain controller, and after being driven by the MCU in the domain controller, the positioning data input stream is obtained. Other auxiliary sensors include wheel odometer, IMU (inertial measurement unit), etc., which are used to provide additional motion information and enhance the robustness of the positioning system.
[0021] Step 2, perform denoising, filtering, and downsampling preprocessing on the collected point cloud data to improve the data quality and the efficiency of subsequent processing; The denoising preprocessing is to perform radius filtering; the filtering preprocessing is to perform Gaussian filtering and mean filtering; the preprocessing is to perform voxel gridification, as Figure 4 shown.
[0022] Step 3, extract specified geometric features from the preprocessed point cloud data; The geometric features include edges, corners, and planes. These geometric features are unique, stable, and their distinguishability is stable.
[0023] Step 4, according to the extracted geometric features, match the point cloud data of adjacent frames through the normal distribution transformation algorithm to calculate the relative pose change; The normal distribution transformation algorithm (NDT algorithm) is a method for point cloud data registration. By calculating the probability distribution of the point cloud data and performing matching, the precise alignment of the point cloud is achieved, as Figure 4 shown. Step 4 also includes using the nearest neighbor search based on the tree data structure (KD tree) and parallel computing based on the visual processor (GPU) as acceleration strategies to improve the matching efficiency and accuracy. Of course, other acceleration strategies can also be adopted according to actual needs.
[0024] Step 5, globally optimize the relative pose change obtained by matching using a graph optimization algorithm; Step 6, apply the optimized relative pose change to the point cloud data, and then splice adjacent frame point clouds to construct a global environment map; Step 7, fuse the relative pose change provided by the laser odometer in the simulation with the absolute position information provided by the inertial navigation; Step 8, match the fused result with the global environment map to achieve positioning.
[0025] The present invention provides a fusion positioning method for an unmanned container truck in a simulation environment. This fusion positioning method combines point cloud splicing, laser odometer, and inertial navigation information. By fusing data from multiple sensors, it achieves complementary advantages, improves positioning accuracy and robustness. Aiming at the deficiencies of existing unmanned container truck positioning technologies, it improves the positioning accuracy, robustness, and real-time performance of unmanned container trucks in complex simulation environments, meeting the application requirements in large-scale scenarios such as ports.
[0026] Embodiment 3: The present invention provides a fusion positioning method for an unmanned container truck in a simulation environment, as Figure 1 shown, the method includes the following steps.
[0027] Step 1, use a lidar module to continuously scan the driving path of the unmanned container truck in the simulation environment to obtain point cloud data; Step 2, perform preprocessing such as denoising, filtering, and downsampling on the collected point cloud data to improve data quality and the efficiency of subsequent processing; Step 3, extract specified geometric features from the preprocessed point cloud data; Step 4, according to the extracted geometric features, match the point cloud data of adjacent frames through the normal distribution transformation algorithm to calculate the relative pose change; Step 5, globally optimize the relative pose change obtained by matching using a graph optimization algorithm; The method of the global optimization includes: based on the spatial relationship and time continuity factors of the point cloud data, perform algorithm integration based on the Kalman filter and the smoothing and mapping algorithm to eliminate cumulative errors and improve positioning accuracy.
[0028] Step 6, apply the optimized relative pose change to the point cloud data, and then splice adjacent frame point clouds to construct a global environment map; The step 6 further includes: during the splicing process, maintain the continuity and consistency of the point cloud data, and adopt an incremental and hierarchical splicing strategy to improve the splicing efficiency and accuracy.
[0029] Step 7: Fuse the relative pose changes provided by the laser odometer in the simulation with the absolute position information provided by the inertial navigation system. The fusion method includes: According to the data characteristics and error distributions provided by the laser odometer and inertial navigation system in the simulation, use the Kalman filter algorithm, Bayesian network algorithm, and particle filter algorithm for fusion to improve the positioning accuracy and robustness. The laser odometer is a sensor system that calculates the vehicle's movement distance and direction changes based on laser scan data. By continuously scanning the environment and comparing the changes in the scan results, the vehicle's movement trajectory is deduced. Inertial navigation, that is, the full name inertial navigation system, is an autonomous navigation system that uses sensors such as accelerometers and gyroscopes to measure the carrier's acceleration and angular velocity, and then deduces the carrier's position, speed, and attitude.
[0030] Step 8: Match the fused result with the global environment map to achieve positioning.
[0031] The method of matching the fused result with the global environment map includes: According to the accuracy and resolution of the map, use methods based on point features and line features, as well as methods based on image matching and image stitching for precise navigation and positioning of the unmanned container truck.
[0032] The present invention provides a fusion positioning method for an unmanned container truck in a simulation environment. This fusion positioning method combines point cloud stitching, laser odometer, and inertial navigation information. By fusing the data of multiple sensors, it realizes complementary advantages, improves the positioning accuracy and robustness. Aiming at the deficiencies of the existing unmanned container truck positioning technology, it improves the positioning accuracy, robustness, and real-time performance of the unmanned container truck in a complex simulation environment, meeting the application requirements in large-scale scenarios such as ports.
[0033] In summary, the fusion positioning method for an unmanned container truck provided by the embodiment of the present invention can significantly improve the positioning accuracy of the unmanned container truck in a complex simulation environment by combining point cloud stitching, laser odometer, and inertial navigation information. This technical solution adopts multi-sensor fusion technology, which can make full use of the advantages of various sensors, complement each other's deficiencies. Even when some sensors fail or the data is abnormal, it can still rely on the information provided by other sensors to maintain normal operation, thereby enhancing the robustness and reliability of the system. This technical solution ensures that the unmanned container truck can quickly respond to environmental changes, achieve real-time positioning and navigation, and helps to reduce waiting time and improve the overall operation efficiency through means such as optimizing algorithm implementation and hardware acceleration. Since the present invention adopts multi-sensor fusion technology, it reduces the dependence on a single sensor, so it reduces the downtime and maintenance costs caused by sensor failures. At the same time, through precise positioning and navigation, it also reduces vehicle damage and cargo losses caused by inaccurate positioning, further reducing the operation costs.
[0034] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.
Claims
1. A fusion positioning method for unmanned container trucks in a simulation environment, characterized in that: The method comprises: Step 1: Use the laser radar module to continuously scan the driving path of the unmanned container truck in a simulation environment to obtain point cloud data; Step 2: pre-process the collected point cloud data by denoising, filtering, and downsampling to improve data quality and the efficiency of subsequent processing; Step 3, extracting specified geometric features from the preprocessed point cloud data; Step 4: According to the extracted geometric features, the point cloud data of adjacent frames are matched by a normal distribution transformation algorithm to calculate the relative pose change; Step 5, using a graph optimization algorithm to globally optimize the relative pose changes obtained by matching; Step 6: Apply the optimized relative pose change to the point cloud data, and then stitch the point clouds of adjacent frames to construct a global environment map; Step 7, integrating the relative posture change provided by the laser odometer in the simulation with the absolute position information provided by the inertial navigation system; Step 8: Match the fused result with the global environment map to achieve positioning.
2. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The step 1 also includes recording data from the inertial navigation system and other auxiliary sensors.
3. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The denoising preprocessing is radius filtering; the filtering preprocessing is Gaussian filtering and mean filtering; the preprocessing is voxel gridding.
4. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The geometric features include edges, corners, and planes.
5. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The step 4 also includes using the nearest neighbor search based on the tree data structure and the parallel computing based on the visual processor as an acceleration strategy to improve the matching efficiency and accuracy.
6. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The global optimization method includes: performing algorithm integration based on Kalman filtering and smoothing and mapping algorithms according to the spatial relationship and time continuity factors of point cloud data to eliminate cumulative errors and improve positioning accuracy.
7. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The step 6 also includes: during the stitching process, maintaining the continuity and consistency of the point cloud data, and adopting incremental and layered stitching strategies to improve the stitching efficiency and accuracy.
8. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The fusion method includes: according to the data characteristics and error distribution provided by the laser odometer and the inertial navigation in the simulation, a Kalman filter algorithm, a Bayesian network algorithm, and a particle filter algorithm are used for fusion, so as to improve the positioning accuracy and robustness.
9. According to claim 1, a fusion positioning method for unmanned container trucks in a simulation environment is characterized in that: The method for matching the fused result with the global environment map includes: according to the accuracy and resolution of the map, using a method based on point features and line features, as well as a method based on image matching and image stitching to perform accurate navigation and positioning of the unmanned container truck.