A multi-radar high-redundancy alignment method for port unmanned IGV empty container area

Abstract

The application discloses a kind of port unmanned IGV empty container area multi-radar high redundancy alignment methods, it is related to unmanned technical field, the method includes the following specific steps: task receiving and rough positioning travel: unmanned IGV vehicle receives target empty container area task with target latitude and longitude information of target berth point, according to latitude and longitude information and target latitude and longitude information of target berth point, independently travel to target empty container area destination range;The application calculates the parking reference distance by intelligently selecting the way of detecting container or detecting stacker according to the actual operation environment of empty container area, effectively deals with the problem of conventional algorithm failure caused by the lack of obvious plane features of stacker, in addition, the influence of equipment failure on perception system is reduced by multi-radar redundancy design, avoids the operation interruption caused by single sensor failure, enhances the scene adaptability and operation continuity of system, provides stable and reliable technical support for port empty container loading and unloading operation.

Description

Technical Field

[0001] This invention relates to the field of unmanned driving technology, specifically to a method for high-redundancy multi-radar alignment in the empty container area of ​​an unmanned IGV (Inertial Vehicle) in a port. Background Technology

[0002] In the port container handling system, the efficient operation of the empty container area is a key link to ensure the smooth flow of port logistics. Among them, the positioning operation of the Intelligent Guided Vehicle (IGV) in the empty container area is particularly important, as it directly connects the loading and unloading process of empty containers. The accuracy of positioning not only affects the continuity of the operation process, but also the safety of the operation. With the continuous improvement of port automation and intelligence, higher requirements are placed on the positioning technology of IGV vehicles in the empty container area. It needs to have higher precision, stronger redundancy and better scenario adaptability to cope with the complex and ever-changing port operation environment.

[0003] Traditional port IGV vehicle positioning technology mainly relies on a single sensor for positioning and sensing functions. GPS positioning is a commonly used method. However, the complex port environment, with its numerous buildings and equipment, easily obstructs and interferes with GPS signals, leading to significant positioning errors and failing to meet the precise positioning requirements for empty container loading and unloading operations. Another common technology, single-laser radar sensing, while providing some environmental awareness, has significant blind spots, and equipment malfunctions directly cause sensing failure, severely impacting the stability and reliability of positioning. (Publication number CN11274843) The invention of 7A discloses a method for precise alignment of a single container in a container yard. It measures the deviation of a single container from a preset alignment baseline by a single detection device set on an unmanned truck, thereby guiding the vehicle to perform alignment operations. However, this method is only designed for single containers and cannot effectively handle complex scenarios with two or more containers. Furthermore, the sensor configuration has low redundancy, uses only a single detection device, and does not adopt a multi-radar cooperative perception strategy. Under abnormal conditions such as obstruction, strong light, or sensor failure, it is prone to data loss or false detection, reducing system reliability. Therefore, a multi-radar high-redundancy alignment method for empty container areas of unmanned IGVs in ports is to be developed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a multi-radar high-redundancy alignment method for empty container areas in unmanned IGV (Independent Container Vehicle) systems at ports. This method involves strategically deploying two first lidars and four second lidars on the unmanned IGV vehicle and using an onboard terminal to fuse multi-radar data. After receiving the task, the vehicle first performs coarse positioning and moves to the vicinity of the target empty container area. Then, through multi-radar data processing and a scenario-specific reference acquisition strategy, it accurately calculates the parking reference distance V1. Simultaneously, it uses the two first lidars to determine the vehicle-mounted container offset distance V2, thereby obtaining the target parking distance. Finally, it controls the vehicle to complete precise parking. This method improves system redundancy through multi-radar fusion technology and enhances scenario adaptability through a scenario-specific reference acquisition strategy. It effectively solves the problems of insufficient alignment accuracy and poor scenario adaptability in empty container areas, providing stable and reliable alignment technology support for IGV empty container loading and unloading operations at ports.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-radar high-redundancy alignment method for empty container areas of unmanned IGVs in ports, the method comprising the following specific steps: Task reception and coarse positioning and navigation: The unmanned IGV vehicle receives the task of the target empty container area carrying the latitude and longitude information of the target location, and autonomously travels to the target empty container area destination area based on its own latitude and longitude information and the latitude and longitude information of the target location. Multi-radar data processing and parking reference distance calculation: Raw laser data from the lidar is synchronously acquired through the vehicle-mounted terminal. After coordinate system transformation, data fusion, and invalid point removal, a cluster of points to be processed is obtained through direct filtering. Based on the placement of containers in the target empty container area, either the container itself or the forklift is detected to calculate the parking reference distance. ; Target parking distance determination: The vehicle-mounted terminal uses two first-stage lidar sensors to detect the distance between the front and rear ends of the container and the vehicle, thus determining the offset distance of the vehicle-mounted container. Then calculate the reference distance for parking. Distance from vehicle body The difference is used to obtain the target stopping distance; Autonomous parking control: The unmanned IGV vehicle autonomously controls the accelerator to drive the vehicle to the target position according to the target parking distance, and controls the brake to complete the parking.

[0006] Furthermore, in the multi-radar data processing and parking reference distance calculation steps, the vehicle terminal is communicatively connected to two first lidars and four second lidars respectively; wherein the two first lidars are set on the top of the unmanned IGV vehicle and are located at the front right and rear left positions respectively; the four second lidars are respectively set around the unmanned IGV vehicle.

[0007] Furthermore, in the multi-radar data processing and parking reference distance calculation steps, the filtering range is set as follows: the X-axis is 8 meters before and after the target bay, which is dynamically adjusted according to the relative position of the vehicle and the target bay; the Y-axis is 1.8 to 7.0 or -1.8 to -7.0; and the Z-axis is 0 to 2m. After filtering, only the cluster of points to be processed containing the box body is obtained.

[0008] Furthermore, in the multi-radar data processing and positioning reference distance calculation steps, the point cloud to be processed is projected into two-dimensional data (Z-value stripped) and then projected onto the X-axis (Y-value stripped), and arranged in ascending order of X-values ​​to form point cloud segments; the segments are divided into three key regions: the region within 20cm after the starting point of the 16-meter fixed segment region, the region within 10cm before and after the center of the 16-meter fixed segment region, and the region within 20cm before the ending point of the 16-meter fixed segment region; the number of point clouds in each region is counted, and if ≥15, a corresponding marker position is established. Array elements are set to 1 if they are not 1, and 0 otherwise. The array contains three elements, each corresponding to a region 20cm after the starting point. 10cm area before and after the center 20cm before the finish line Its possible value combinations include , , , , , , .

[0009] Furthermore, in the multi-radar data processing and parking reference distance calculation steps, if array is , , or That is, the target location has a box: the point with the maximum X value in the 10cm area before and after the center. Search in the direction of increasing X value; if the distance between two adjacent points is >30cm, record the previous point as... Otherwise, record the last point of the search as The point with the minimum X value in this region Search in the direction of decreasing X value; if the distance between two adjacent points is >30cm, record the previous point as... Otherwise, record the last point of the search as Through formula calculate .

[0010] Furthermore, in the multi-radar data processing and parking reference distance calculation steps, if array is or That is, the target location has no box but there is a box in front of it: the point with the maximum X value in the area 20cm in front of the endpoint. Search in the direction of decreasing X value; if the distance between two adjacent points is >30cm, record the previous point as... The second point is Through formula calculate If no points with an interval > 30cm are found, record the last point searched as [missing information]. Through formula calculate in, The gap between containers is fixed at 0.45m. This refers to the width of the container.

[0011] Furthermore, in the multi-radar data processing and parking reference distance calculation steps, if array is That is, the target location has no box but there is a box behind it: the point with the minimum X value in the 20cm area behind the starting point. Search in the direction of increasing X value to find the point with the maximum X value in the region. Through formula calculate ,in, The gap between containers is fixed at 0.45m. This refers to the width of the container.

[0012] Furthermore, the width of the container Determined based on the type of container in the target bay: 20-foot container. It is a 6.096m, 40-foot container It is 12.192m.

[0013] Furthermore, in the multi-radar data processing and parking reference distance calculation steps, if array is The raw laser data from four second lidars were acquired. Based on the pre-completed laser calibration results, the coordinate systems of each lidar were transformed to the vehicle coordinate system and then fused to form a unified laser dataset. Invalid points in the unified laser dataset were removed, and a second pass-through filter was used to filter and obtain a point cloud containing only the forklift. The filtering range of the second pass-through filter was: X-axis 2m before and after the target location, Y-axis (-8.0 to -1.8) or (1.8 to 8.0), and Z-axis (0 to 3)m. The point cloud was projected into two-dimensional data, and a Euclidean clustering algorithm based on KD-Tree nearest neighbor query was used to remove noise points and extract the point cloud cluster with the most laser points. The last two-thirds of this point cloud cluster was extracted, and it was assumed that this part contained n point clouds, with the X values ​​of each point cloud being as follows. , … Through formula Calculate the mean of X values; this mean is... .

[0014] Furthermore, in the target parking distance determination step, the vehicle-mounted box offset distance... The calculation method is determined based on the specific delivery scenario of the task type: Send front box: ; Send front box: ; Delivery to middle box: ; Send back box: ; Send back box: ; The calculation method for the target parking distance is determined based on the task type and sub-type. 20cm is the manually controlled distance between the two containers to prevent them from being too close and hindering operations. Three calibration parameters are defined: , ; ; For the small box before closing: target stopping distance ; Rear box: Target parking distance ; Receiving medium and large boxes: target parking distance ; For the delivery box: target stopping distance Delivery box: Target parking distance ; Delivery of medium to large boxes (including 40-foot and other sizes): Target parking distance .

[0015] Compared with existing technologies, this method for high-redundancy multi-radar alignment in empty container areas of unmanned IGVs in ports has the following advantages: I. This invention calculates the stopping reference distance by intelligently selecting the detection method of the container or the forklift based on the actual working environment of the empty container area. This effectively addresses the problem of conventional algorithms failing due to the lack of obvious planar features of the forklift. In addition, the multi-radar redundancy design reduces the impact of equipment failure on the sensing system, avoids operation interruption due to the failure of a single sensor, enhances the system's scene adaptability and operation continuity, and provides stable and reliable technical support for port empty container loading and unloading operations.

[0016] Second, this invention solves the problems of large positioning errors, blind spots, and susceptibility to environmental interference caused by single sensors by combining data from two first lidars and four second lidars through multi-radar fusion technology. The high redundancy design of the multi-radar system ensures the reliability and integrity of the data, while the scenario-based benchmark acquisition strategy designs differentiated alignment methods for different operating scenarios, significantly improving the alignment accuracy and stability of IGV vehicles in empty container areas at ports.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 A flowchart of a multi-radar high-redundancy alignment method for an unmanned IGV empty container area in a port; Figure 2 This is a schematic diagram of the installation position of the vehicle-mounted radar in a multi-radar high-redundancy alignment method for an unmanned IGV empty container area in a port. Figure 3 This is a flowchart illustrating the steps of multi-radar data processing and parking reference distance calculation for a multi-radar high-redundancy alignment method for an unmanned IGV empty container area in a port. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] Example 1 At bay number 3 of the empty container storage area of ​​a coastal port, a transfer and unloading task of 20-foot empty containers needs to be completed (task type: container delivery to a central location). Three rows of 20-foot empty containers (0.45m spacing) are already arranged closely around this bay. If... Figure 2 As shown, the unmanned IGV vehicle is equipped with one first lidar unit (scanning frequency 10Hz, ranging accuracy ±2cm) containing right front and left rear units, and four second lidar units distributed around the front, rear, left, and right sides of the vehicle (scanning angle 360°, ranging range 0.1-200m). Figure 1 As shown, the specific alignment process is as follows: The vehicle dispatch system sends tasks to IGV vehicles via the 5G network. The task data packet contains the latitude and longitude coordinates of position 3 (38.56°N, 121.23°E) and the target container type (20-foot empty container). During the vehicle's journey, it obtains its own position in real time through the onboard GPS. When the vehicle detects a latitude and longitude deviation of less than 5m from position 3, it determines that it has entered the vicinity of the destination area and then reduces its speed to 2km / h to prepare for the alignment process. At this time, the GPS is only used for preliminary positioning. Due to the positioning error of ±0.5m, it does not directly participate in the parking control.

[0022] like Figure 3 As shown, four secondary lidars simultaneously collect raw point cloud data of the surrounding environment. Each lidar generates 10 sets of data per second. The vehicle terminal, based on pre-calibrated lidar extrinsic parameters (including translation and rotation angle parameters), uniformly transforms the point cloud data from each lidar coordinate system to the vehicle coordinate system (X-axis parallel to the vehicle's driving direction, positive in front of the vehicle center and negative behind; Y-axis parallel to the ground and perpendicular to the X-axis, positive to the left of the vehicle center and negative to the right; Z-axis perpendicular to the ground, positive upwards). Then, data deduplication technology is used to fuse them into a complete lidar dataset to ensure data consistency.

[0023] The vehicle-mounted terminal preprocesses the fused dataset. First, it removes invalid points that are at infinity (distance greater than 200m) and noisy (isolated single points without adjacent points) by using a threshold. Then, it uses the first direct-pass filter to filter the point cloud to be processed. The X-axis range is set to 8m before and after the 3rd bay position (dynamically calculated and adjusted according to the real-time vehicle position and bay position coordinates to ensure coverage of the target area container). The Y-axis range is selected from the roadside (1.8~7.0m, adapting to the width of the container placed on the roadside). The Z-axis is set to 0~2m (covering the height of the container body and avoiding interference from ground debris). Finally, only the point cloud related to the container is retained.

[0024] The filtered point cloud data was projected into two-dimensional data (Z-axis height information was removed), and then further projected onto the X-axis and arranged in ascending order of X-values ​​to form continuous point cloud segments. The number of point cloud segments in three key regions—20cm after the start of the segment, 10cm before and after the center of the segment, and 20cm before the end of the segment—was counted. The number of point cloud segments in each region exceeded 15 (meeting the calculation accuracy requirements), therefore... The array is set to {1, 1, 1}, and the target bit is determined to exist in the box.

[0025] Maximum X value in the central region Searching for the interval between adjacent points in the direction of increasing X value, no points with an interval greater than 30cm were found (indicating that the box is continuous and without gaps). The last point searched was recorded as... Similarly, the minimum X value for the central region. Searching in the direction of decreasing X value, no points with intervals greater than 30cm were found. The last point searched was recorded as... Finally, according to the formula Calculate the reference distance for parking. ,make sure Corresponding to the center position of the box.

[0026] The front right and rear left units of the first lidar scan the front and rear boxes of the vehicle, respectively, and use point cloud matching algorithms to detect the distance between the rear of the box and the center of the vehicle. The distance is 3.2m, which is the distance between the front end of the container and the center of the vehicle. The length is 9.3m. Since the task type is a delivery box, according to the formula... Calculate the offset distance of the vehicle body ( The distance from the center of the vehicle-mounted container to the center of the vehicle is 3.2m + (9.3m - 3.2m) / 2 = 6.25m. Based on the task type (delivery container), the definition is... According to the formula, the target stopping distance Substitute (Assuming the calculated value is 7.75m) , The target stopping distance is 7.75m - (6.25m + 0) = 1.5m.

[0027] The vehicle terminal sends the target parking distance to the vehicle control system. The system automatically adjusts the throttle opening based on this distance (gradually reducing it to 0). At the same time, it controls the braking force based on the real-time vehicle speed (obtained through wheel speed sensors). It first lightly applies the brakes to slow down, and when the vehicle is only 0.3m away from the target position, it slowly applies the brakes until the vehicle comes to a complete stop. The final parking error is controlled within ±5cm, which meets the alignment accuracy requirements for loading and unloading 20-foot empty containers.

[0028] Example 2 Bay number 8 in the empty container area of ​​an inland port is a newly built area and currently has no containers. Unmanned IGV vehicles are needed to transport 20-foot empty containers to this bay and work in conjunction with a forklift to complete the unloading operation. The unmanned IGV vehicle is equipped with one primary lidar unit (scanning frequency 10Hz, ranging accuracy ±2cm) with units on the right front and left rear, and four secondary lidar units distributed around the vehicle's front, rear, left, and right sides (scanning angle 360°, ranging range 0.1-200m). The forklift has been pre-positioned near the center of bay number 8 (cab facing the vehicle's direction of travel). The specific alignment process is as follows: The vehicle receives instructions containing the latitude and longitude of position 8 (30.12°N, 114.35°E) and the target task (unloading a 20-foot empty box). During the journey, it uses GPS for real-time positioning. When the deviation from the latitude and longitude of position 8 is less than 3m, it is determined that it has entered the vicinity of the destination area, and the vehicle speed is reduced to 1.5km / h to prepare for alignment.

[0029] Data collected by four secondary lidar sensors was converted and fused into the vehicle coordinate system. After being filtered by a first direct-pass filter (8m for X-axis, 1.8-7.0m for Y-axis, and 0-2m for Z-axis), the point cloud was projected onto the X-axis and sorted. The areas 20cm after the start of the line segment, 10cm before and after the center, and 20cm before the end of the line segment were examined. The number of point clouds in each area was less than 15 (no box-shaped point clouds). If the array is set to {0, 0, 0}, the container detection mode is deemed invalid, and the system automatically switches to the forklift detection mode. .

[0030] Four secondary lidars re-collect raw data of the surrounding environment, with each lidar generating 10 sets of data per second. The vehicle terminal, based on pre-calibrated lidar extrinsic parameters (including translation and rotation angle parameters), uniformly transforms the point cloud data in each lidar coordinate system to the vehicle coordinate system (X-axis parallel to the vehicle's driving direction, positive in front of the vehicle center and negative behind; Y-axis parallel to the ground and perpendicular to the X-axis, positive to the left of the vehicle center and negative to the right; Z-axis perpendicular to the ground, positive upwards). The data is then unified to the vehicle coordinate system and fused to remove duplicates, resulting in a point cloud dataset containing the forklift and its surrounding environment.

[0031] The fused dataset is preprocessed to remove invalid points at infinity and those with noise. Then, a second pass-through filter is used to filter the point cloud. The X-axis is set to 2m before and after position 8 (focusing on the area where the forklift is located), the Y-axis is selected from the seaside range (-8.0 to -1.8m, to match the parking position of the forklift on the seaside), and the Z-axis is set to 0 to 3m (covering the height of the forklift cab to avoid missing point clouds of key parts). Point clouds related to the forklift are retained.

[0032] The filtered point cloud was projected as two-dimensional data (Z-axis information was removed). Based on KD-Tree nearest neighbor query acceleration, a clustering distance threshold of 0.5m was set to classify the two-dimensional point cloud, and noisy points such as surrounding cones (few point cloud items and small cluster size) were removed. The point cluster with the most laser points (i.e., the forklift point cloud) was selected. Through point cloud distribution analysis, it was found that the first 1 / 3 of the point cluster (forklift lift) had sparse point cloud and a large X value range (error fluctuation ±0.3m), while the latter 2 / 3 (forklift cockpit) had dense and evenly distributed point cloud. Therefore, the average X value of the latter 2 / 3 point cloud was taken to obtain the parking reference distance V1. According to field tests, the deviation of this average value from the actual center position of the forklift was within ±15cm, which met the positioning accuracy requirements of the forklift.

[0033] The first lidar scanner scanned the 20-foot empty container carried by the vehicle and detected the distance between the rear of the container and the center of the vehicle. It is 2.8m, the distance between the front end of the box and the center of the vehicle. The length is 8.9m. Since the task type is delivering a front small box and there is no rear box, according to the formula... Calculate the offset distance of the vehicle body ( The distance from the center of the container to the center of the vehicle is 2.8m + (8.9m - 2.8m) / 2 = 5.85m. The width of a 20-foot container is defined as... , (i.e., -0.2m - 6.09m / 2) -3.245m), according to the formula, the target stopping distance = Substitute 20cm = 0.2m , , The calculated target stopping distance is 2.15m.

[0034] Based on the target parking distance and information from the onboard millimeter-wave radar (which assists in detecting the real-time distance to the forklift), the vehicle control system autonomously controls the throttle and brakes: it first slowly approaches the target position at a speed of 1.5 km / h. When the difference between the real-time distance and the target parking distance is less than 0.5 m, the throttle is turned off, and the vehicle coasts by inertia. When the difference is less than 0.2 m, the brakes are lightly applied until the vehicle comes to a complete stop. The final parking position is within ±8 cm of the forklift, allowing the forklift to successfully complete the empty container unloading operation.

[0035] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations 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 scope of the present invention.

Claims

1. A method for high-redundancy multi-radar alignment in the empty container area of ​​an unmanned IGV (Inertial Vehicle) in a port, characterized in that, The method includes the following specific steps: Task reception and coarse positioning and navigation: The unmanned IGV vehicle receives the task of the target empty container area carrying the latitude and longitude information of the target location, and autonomously travels to the target empty container area destination area based on its own latitude and longitude information and the latitude and longitude information of the target location. Multi-radar data processing and parking reference distance calculation: Raw laser data from the lidar is synchronously acquired through the vehicle-mounted terminal. After coordinate system transformation, data fusion, and invalid point removal, a cluster of points to be processed is obtained through direct filtering. Based on the placement of containers in the target empty container area, either the container itself or the forklift is detected to calculate the parking reference distance. ; Target parking distance determination: The vehicle-mounted terminal uses two first-stage lidar sensors to detect the distance between the front and rear ends of the container and the vehicle, thus determining the offset distance of the vehicle-mounted container. Then calculate the reference distance for parking. Distance from vehicle body The difference is used to obtain the target stopping distance; Autonomous parking control: The unmanned IGV vehicle autonomously controls the accelerator to drive the vehicle to the target position according to the target parking distance, and controls the brake to complete the parking.

2. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 1, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, the vehicle terminal is communicatively connected to two first lidars and four second lidars respectively; the two first lidars are set on the top of the unmanned IGV vehicle and are located at the front right and rear left positions respectively; the four second lidars are respectively set around the unmanned IGV vehicle.

3. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 1, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, the filtering range is set as follows: the X-axis is 8 meters before and after the target bay, which is dynamically adjusted according to the relative position of the vehicle and the target bay; the Y-axis is 1.8 to 7.0 or -1.8 to -7.0; and the Z-axis is 0 to 2m. After filtering, only the cluster of points to be processed containing the box body is obtained.

4. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 1, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, the point cloud to be processed is projected into two-dimensional data and then projected onto the X-axis, and the point cloud segments are arranged in ascending order of X value. The three key regions for dividing the line segment are: the area within 20cm after the starting point of the 16-meter fixed line segment; the area within 10cm before and after the center of the 16-meter fixed line segment; and the area within 20cm before the ending point of the 16-meter fixed line segment. The number of point clouds in each region is counted; if ≥15, a corresponding marker is established. Array elements are set to 1 if they are not 1, and 0 otherwise. The array contains three elements, each corresponding to a region 20cm after the starting point. 10cm area before and after the center 20cm before the finish line Its possible value combinations include , , , , , , .

5. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 4, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, if array is , , or That is, the target location has a box: the point with the maximum X value in the 10cm area before and after the center. Search in the direction of increasing X value; if the distance between two adjacent points is >30cm, record the previous point as... Otherwise, record the last point of the search as The point with the minimum X value in this region Search in the direction of decreasing X value; if the distance between two adjacent points is >30cm, record the previous point as... Otherwise, record the last point of the search as Through formula calculate .

6. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 4, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, if array is or That is, the target location has no box but there is a box in front of it: the point with the maximum X value in the area 20cm in front of the endpoint. Search in the direction of decreasing X value; if the distance between two adjacent points is >30cm, record the previous point as... The second point is Through formula calculate If no point with an interval > 30cm is found, record the last point searched as [missing information]. Through formula calculate in, The gap between containers is fixed at 0.45m. This refers to the width of the container.

7. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 4, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, if array is That is, the target location has no box but there is a box behind it: the point with the minimum X value in the 20cm area behind the starting point. Search in the direction of increasing X value to find the point with the maximum X value in the region. Through formula calculate ,in, The gap between containers is fixed at 0.45m. This refers to the width of the container.

8. The method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 7, is characterized in that... The width of the container Determined based on the type of container in the target bay: 20-foot container. It is a 6.096m, 40-foot container It is 12.192m.

9. A method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 4, is characterized in that... In the multi-radar data processing and parking reference distance calculation steps, if array is The raw laser data from four second lidars were acquired. Based on the pre-completed laser calibration results, the coordinate systems of each lidar were transformed to the vehicle coordinate system and then fused to form a unified laser dataset. Invalid points in the unified laser dataset were removed, and a second pass-through filter was used to filter and obtain a point cloud containing only the forklift. The filtering range of the second pass-through filter was: X-axis 2m before and after the target location, Y-axis (-8.0 to -1.8) or (1.8 to 8.0), and Z-axis (0 to 3)m. The point cloud was projected into two-dimensional data, and a Euclidean clustering algorithm based on KD-Tree nearest neighbor query was used to remove noise points and extract the point cloud cluster with the most laser points. The last two-thirds of this point cloud cluster was extracted, and it was assumed that this part contained n point clouds, with the X values ​​of each point cloud being as follows. , … Through formula Calculate the mean of X values; this mean is... .

10. A method for high-redundancy multi-radar alignment in an empty container area of ​​an unmanned IGV in a port, as described in claim 1, is characterized in that... In the target parking distance determination step, the vehicle-mounted box offset distance The calculation method is determined based on the specific delivery scenario of the task type: For items in the front box, and for items in the rear box: ; For boxes without a rear box: ; Delivery to middle box: ; For items stored in the rear compartment, or in the front compartment: ; Delivery to the rear box; for cases where the front box is not available: ; The calculation method for the target parking distance is determined based on the task type and sub-type. 20cm is the manually controlled distance between the two containers to prevent them from being too close and hindering operations. Three calibration parameters are defined: 、 ； ； For the small box before closing: target stopping distance ; Rear box: Target parking distance ; Receiving medium and large boxes: target parking distance ; For the delivery box: target stopping distance Delivery box: Target parking distance ; Delivery of medium to large boxes (including 40-foot and other sizes): Target parking distance .

Images