Fusion navigation method and device of unmanned diet transport vehicle

Abstract

The invention discloses a fusion navigation method and device of an unmanned diet transport vehicle. The method comprises the following steps: acquiring position information of a target to be arrived; the to-be-arrived target position information is the to-be-arrived position information of the unmanned diet transport vehicle; generating preset path information based on the target position information; and driving the unmanned diet transport vehicle to run according to the preset path information to reach the target position.

Description

Technical Field

[0001] This invention relates to the fields of autonomous driving and target tracking, and specifically to a fusion navigation method and device for an unmanned food delivery vehicle. Background Technology

[0002] With the development of intelligent manufacturing, the Internet of Things, and new energy power, "intelligent food transport vehicles," including automated guided vehicles (AGVs / AMRs) used in factories / canteens, small electric vans for last-mile delivery in cities, and containerized mobile kitchens for field or emergency scenarios, are showing a rapid growth trend in industrial applications. Their core driving technology stems from the progress in the field of autonomous driving. Through multi-sensor fusion (LiDAR, vision, millimeter-wave radar) and SLAM technology, they can achieve L4-level autonomous driving navigation and high-precision parking in semi-structured environments such as parks and factories.

[0003] However, the navigation performance of unmanned food delivery vehicles directly affects their application effectiveness. In existing technologies, navigation methods using a single sensor have drawbacks such as limited sensing range and poor environmental adaptability. For example, when relying solely on cameras for navigation, recognition errors are prone to occur in poor lighting conditions or complex environments; when using only LiDAR, the accuracy in recognizing dynamic obstacles is insufficient; and although millimeter-wave radar performs stably in adverse weather conditions, its ability to identify details of targets is weak.

[0004] Meanwhile, in terms of path planning and dynamic obstacle avoidance, traditional methods are mostly based on static path planning using pre-set maps, which is difficult to cope with dynamic obstacles that suddenly appear during transportation (such as pedestrians, other mobile devices, etc.), easily leading to transportation interruptions or collision risks. In addition, the fusion processing of multi-sensor data is also a technical challenge in unmanned navigation. Differences in the time synchronization, spatial coordinate consistency, and data reliability of different sensors can affect the accuracy of target detection, thereby reducing the reliability of the navigation system.

[0005] In conclusion, a single navigation system is insufficient for accurate navigation and path planning of unmanned food delivery vehicles in complex environments. How to effectively integrate multiple types of sensors is an urgent problem to be solved. Summary of the Invention

[0006] This invention primarily addresses the problem that single-system navigation methods are insufficient for achieving accurate navigation and path planning for unmanned food delivery vehicles in complex environments. This invention discloses a fusion navigation method and device for unmanned food delivery vehicles.

[0007] In a first aspect, this application discloses a fusion navigation method for an unmanned food delivery vehicle, comprising: S1, Obtain the location information of the target to be reached; the location information of the target to be reached is the location information of the unmanned food delivery vehicle to be reached; S2, Based on the target location information, generate preset path information; S3, drive the unmanned food delivery vehicle to travel according to the preset path information and reach the target location; The target location is the location corresponding to the target location information.

[0008] The step of generating preset path information based on target location information includes: S21, Perform SLAM modeling on the movement area of ​​the unmanned food delivery vehicle to obtain a set of path information for the movement area; the set of path information includes the path information of each point in the movement area; S22, Based on the target location information, query the path information of the corresponding point in the path information set; S23, confirm that the path information obtained from the query is the preset path information.

[0009] The unmanned food delivery vehicle travels according to the preset path information to reach the target location, including: S31, drive the unmanned food delivery vehicle to travel according to the preset path information; S32, while driving, detects targets in the surrounding environment and obtains target information; S33, Based on the target information, the preset path information is corrected to obtain updated preset path information; S34 drives the unmanned food delivery vehicle to travel according to the updated preset route information and reach the target location.

[0010] The target detection of the surrounding environment to obtain target information includes: S321 performs registration processing on the cameras, lidar, and millimeter-wave radar mounted on the unmanned food delivery vehicle; S322, using the camera, lidar and millimeter-wave radar to detect targets in the surrounding environment and obtain target information.

[0011] The method of using the camera, lidar, and millimeter-wave radar to detect targets in the surrounding environment and obtain target information includes: S3221 uses a camera to acquire images of the target area; S3222 uses lidar to collect point cloud data of the target area; S3223, Perform joint target detection on the target region image and the target region point cloud data to obtain target presence information; S3224, If the target information is true, use millimeter-wave radar to detect and locate the target to obtain the target location information; confirm the target information as target location information; If the target information is negative, the target information is confirmed to be non-existent.

[0012] The joint target detection of the target region image and target region point cloud data to obtain target presence information includes: S32231, Perform image detection processing on the target region image to obtain a first target location information sequence; S32232, Perform target point cloud matching processing on the target area point cloud data to obtain a second target location information sequence; the location information sequence is a sequence composed of location information obtained by detecting the target area image or target area point cloud data at several times; S32233, Perform coordinate transformation on the first target position information sequence and the second target position information sequence respectively to obtain a first three-dimensional position information sequence and a second three-dimensional position information sequence; S32234, Perform joint discrimination calculation on the first three-dimensional position information sequence and the second three-dimensional position information sequence to obtain fused position information and statistical discrimination value; S32235, determine whether the statistical discrimination value is less than a preset discrimination threshold to obtain a first discrimination result; if the first discrimination result is less than, determine that the target information exists; if the first discrimination result is not less than, determine that the target information exists.

[0013] The step of jointly discriminating the first three-dimensional location information sequence and the second three-dimensional location information sequence to obtain fused location information and statistical discriminant values ​​includes: Statistical distribution modeling is performed on the first three-dimensional location information sequence to obtain the first mean. First variance and the set of first probability distribution functions; Perform hypothesis testing on the first set of probability distribution functions to obtain the corresponding first test power value. ; Statistical distribution modeling is performed on the second three-dimensional location information sequence to obtain the second mean. Second variance The set of second probability distribution functions; Perform hypothesis testing on the second set of probability distribution functions to obtain the corresponding second test power value. ; The first mean, first variance, and first test power value are combined and calculated to obtain the statistical discriminant value. The expression for the statistical discriminant value is: , , in, is the scaling factor, F is the quantile function of the standard normal distribution, and t is the statistical discriminant value; The first three-dimensional position information sequence and the second three-dimensional position information sequence are fused to obtain fused position information.

[0014] The beneficial effects of this invention are as follows: This invention achieves omnidirectional detection of the surrounding environment by fusing data from multiple sensors, including cameras, lidar, and millimeter-wave radar, effectively compensating for the limitations of single sensors in environmental adaptability and detection accuracy. Through systematic error registration, temporal registration, and spatial registration of the multiple sensors, consistency of data from different sensors in both time and space is ensured, providing a reliable data foundation for subsequent target detection and path correction, and improving the accuracy and stability of target detection.

[0015] This invention employs SLAM modeling to construct a set of path information for the movement area and combines it with target location information to generate a preset path. Simultaneously, during operation, the path is dynamically corrected through real-time target detection, achieving an organic combination of static path planning and dynamic obstacle avoidance. When an obstacle is detected, local obstacle avoidance is performed using artificial potential field methods or dynamic window methods, generating a dynamic path to adjust the driving trajectory. This effectively avoids the risk of collisions with dynamic obstacles during transportation, improving the autonomous navigation safety and path adaptability of the unmanned food transport vehicle.

[0016] In the target detection process, this invention performs joint target detection by combining images captured by a camera and point cloud data collected by a lidar. It combines statistical distribution modeling and hypothesis testing to determine the existence of the target and further uses millimeter-wave radar to accurately detect the target location, forming a multi-level target recognition mechanism. This significantly improves the accuracy and reliability of target detection and provides accurate decision-making basis for path correction. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention. Detailed Implementation

[0018] To better understand the content of this invention, an embodiment is provided here.

[0019] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0020] To address the challenge of achieving stable tracking of target vehicles in unstructured environments using single-system target tracking methods, and to solve the problem of effectively fusing multiple types of sensors, this invention discloses a fusion navigation method and device for unmanned food transport vehicles.

[0021] In a first aspect, this application discloses a fusion navigation method for an unmanned food delivery vehicle, comprising: S1, Obtain the location information of the target to be reached; The target location information to be reached is the location information to be reached by the unmanned food delivery vehicle; S2, Based on the target location information, generate preset path information; S3, drive the unmanned food delivery vehicle to travel according to the preset path information and reach the target location; The target location is the location corresponding to the target location information.

[0022] The step of generating preset path information based on target location information includes: S21, Perform SLAM modeling on the movement area of ​​the unmanned food delivery vehicle to obtain a set of path information for the movement area; the set of path information includes the path information of each point in the movement area; S22, Based on the target location information, query the path information of the corresponding point in the path information set; S23, confirm that the path information obtained from the query is the preset path information.

[0023] The unmanned food delivery vehicle travels according to the preset path information to reach the target location, including: S31, drive the unmanned food delivery vehicle to travel according to the preset path information; S32, while driving, detects targets in the surrounding environment and obtains target information; S33, Based on the target information, the preset path information is corrected to obtain updated preset path information; S34 drives the unmanned food delivery vehicle to travel according to the updated preset route information and reach the target location.

[0024] The step of correcting the preset path information based on the target information to obtain updated preset path information includes: If the target information is target location information, local obstacle avoidance processing is performed on the target location information to obtain a dynamic path. The preset path information is then corrected using the dynamic path to obtain updated preset path information.

[0025] The local obstacle avoidance process can employ methods such as artificial potential field method or dynamic window method.

[0026] The target detection of the surrounding environment to obtain target information includes: S321 performs registration processing for cameras, lidar, and millimeter-wave radar; S322, using the camera, lidar and millimeter-wave radar to detect targets in the surrounding environment and obtain target information.

[0027] The registration process for the camera, lidar, and millimeter-wave radar includes: S3211, Perform system error registration processing on the camera, lidar and millimeter-wave radar; S3212, perform time registration processing and spatial registration processing on the camera, lidar and millimeter-wave radar.

[0028] The method of using the camera, lidar, and millimeter-wave radar to detect targets in the surrounding environment and obtain target information includes: S3221 uses a camera to acquire images of the target area; S3222 uses lidar to collect point cloud data of the target area; S3223, Perform joint target detection on the target region image and the target region point cloud data to obtain target presence information; S3224, If the target information is true, use millimeter-wave radar to detect and locate the target to obtain the target location information; confirm the target information as target location information; If the target information is negative, the target information is confirmed to be non-existent. The joint target detection of the target region image and target region point cloud data to obtain target presence information includes: S32231, Perform image detection processing on the target region image to obtain a first target location information sequence; S32232, Perform target point cloud matching processing on the target area point cloud data to obtain a second target location information sequence; the location information sequence is a sequence of location information obtained at several measurement times; S32233, Perform coordinate transformation on the first target position information sequence and the second target position information sequence respectively to obtain a first three-dimensional position information sequence and a second three-dimensional position information sequence; S32234, Perform joint discrimination calculation on the first three-dimensional position information sequence and the second three-dimensional position information sequence to obtain fused position information and statistical discrimination value; S32235, determine whether the statistical discrimination value is less than a preset discrimination threshold to obtain a first discrimination result; if the first discrimination result is less than, determine that the target information exists; if the first discrimination result is not less than, determine that the target information exists.

[0029] The step of jointly discriminating the first three-dimensional location information sequence and the second three-dimensional location information sequence to obtain fused location information and statistical discriminant values ​​includes: Statistical distribution modeling is performed on the first three-dimensional location information sequence to obtain the first mean. First variance and the set of first probability distribution functions; Perform hypothesis testing on the first set of probability distribution functions to obtain the corresponding first test power value. ; Statistical distribution modeling is performed on the second three-dimensional location information sequence to obtain the second mean. Second variance The set of second probability distribution functions; Perform hypothesis testing on the second set of probability distribution functions to obtain the corresponding second test power value. ; The first mean, first variance, and first test power value are combined and calculated to obtain the statistical discriminant value. The expression for the statistical discriminant value is: , , in, is the scaling factor, F is the quantile function of the standard normal distribution, and t is the statistical discriminant value; The first three-dimensional position information sequence and the second three-dimensional position information sequence are fused to obtain fused position information.

[0030] The expression for the statistical discriminant value, by integrating the first test power value, the second test power value, and parameters such as the first mean, the second mean, the first variance, and the second variance, achieves a comprehensive discrimination between the first and second three-dimensional position information sequences. Introducing sine and exponential functions for nonlinear processing of the parameters amplifies the differences between different sequences, making the discrimination of target presence more sensitive. Simultaneously, by combining the quantile function F of the standard normal distribution, the scaling factor is mapped to a reasonable statistical distribution interval, ensuring the scientific validity and reliability of the statistical discriminant value, effectively improving the accuracy of target presence information judgment, and providing a precise decision-making basis for whether to subsequently use millimeter-wave radar for precise positioning.

[0031] The expression for the fusion location estimation process includes: , , , in, To fuse location information, M is the length of the location information sequence. and These are the j-th elements of the first three-dimensional position information sequence and the second three-dimensional position information sequence, respectively. , and These are the mean values ​​of the x-axis, y-axis, and z-axis coordinates of the first three-dimensional position information sequence, respectively. , and These are the variances of the x-axis, y-axis, and z-axis coordinates of the first three-dimensional position information sequence, respectively. , and These are the mean values ​​of the x-axis, y-axis, and z-axis coordinates of the second three-dimensional position information sequence, respectively. , and These are the variances of the x-axis, y-axis, and z-axis coordinates of the second three-dimensional position information sequence, respectively.

[0032] The expression for the fusion location estimation process introduces an arcsine function to process each location information element, which can suppress outliers that deviate significantly from the mean and reduce the impact of abnormal data on the fusion result. Simultaneously, by using a weight allocation of 1 / (2M), the contribution of each element in the two sequences is balanced, avoiding fusion bias caused by an excessively high proportion of data from a single sequence. This fusion method combines the advantages of both camera and LiDAR, preserving the detail recognition capabilities of image data while incorporating the spatial positioning accuracy of point cloud data. This results in more accurate and stable fused location information, providing reliable guidance for subsequent millimeter-wave radar detection and positioning.

[0033] The first mean and the first variance are obtained by calculating the mean and variance of the coordinate data of the x-axis, y-axis and z-axis of the first three-dimensional position information sequence, and then calculating the mean of each.

[0034] The second mean and the second variance are obtained by calculating the mean and variance of the x-axis, y-axis and z-axis coordinate data of the second three-dimensional position information sequence, and then calculating the mean of each.

[0035] The statistical distribution modeling assumes that the three-dimensional location information sequence follows a normal distribution, performs statistical distribution modeling on it to obtain a probability distribution function, then performs hypothesis testing on the probability distribution functions of the set of probability distribution functions to obtain the corresponding test power value, and calculates the mean of all test power values ​​to obtain the first test power value or the second test power value. The method of using millimeter-wave radar to detect and locate the target, and obtain the target location information, includes: Using fused location information as guidance information, millimeter-wave radar is used to detect the target and obtain initial location information. ; The target location information is obtained by performing fusion location calculation on the fused location information and the initial location information.

[0036] The expression for calculating the fusion position is: , , , The expression for the target location information is: .

[0037] The expression for calculating the fused location uses an exponential function as the weighting coefficient, dynamically allocating weights based on the magnitude of the two location information values. When the value of a certain location information is larger, its corresponding weight will decrease accordingly, thereby reducing the impact of data with larger errors on the result. This dynamic weighted fusion method fully combines the fused location information from the joint detection of cameras and lidar with the stable detection capability of millimeter-wave radar in harsh environments, realizing the complementary advantages of multi-sensor data. This further improves the accuracy and robustness of target location information, providing a high-precision target location reference for the path correction of unmanned food transport vehicles, and ensuring the accuracy of obstacle avoidance and navigation.

[0038] The coordinate transformation is to convert the coordinates of the position information in the measurement coordinate system into the coordinates in the three-dimensional coordinate system of the unmanned food transport vehicle. The system error registration process involves acquiring the measurement system errors of the camera, lidar, and millimeter-wave radar respectively, and using the measurement system errors of each measuring device to perform error calibration on the collected data.

[0039] The time registration process is to unify the measurement data from different measuring devices onto the same time reference; the time registration process can employ methods such as extrapolation / extrapolation and Lagrange three-point interpolation. The spatial registration process involves transforming the coordinates of measurement data from different measuring devices to unify them under the same coordinate system.

[0040] The image detection processing can be based on a template image of the target, performing template matching processing on the target region image information to obtain target location region information and target location center information; the template matching processing can be the MAD algorithm or the SSD algorithm.

[0041] The target point cloud matching process can employ point cloud feature extraction algorithms from LiDAR or deep learning algorithms for 3D point cloud target detection. Specifically, point cloud template data of the target can be acquired first, and then a global feature extraction algorithm based on three-dimensional Hough transform can be used to match the point cloud template data to obtain the second target location region information.

[0042] The target tracked in this method can be a vehicle, a building, or an obstacle; The camera, lidar, and millimeter-wave radar are all mounted on the same vehicle platform.

[0043] In all embodiments of the present invention, the variables involved in all computational expressions or mathematical functions have been dimensionlessized before computation.

[0044] In all embodiments of the present invention, the values ​​of the independent variables in the input of all computational expressions or mathematical functions meet the reasonable requirements of the input range of the computational expressions or mathematical functions, and can ensure that the computational expressions or mathematical functions can be calculated smoothly without violating physical laws or mathematical rules.

[0045] This invention employs a unique fusion location calculation method to fuse location information acquired by multiple sensors. By combining statistical parameters such as mean, variance, and test power value to construct a fusion model, it not only preserves the effective characteristics of each sensor's data but also reduces the impact of single data errors on the results through weighted fusion. This makes the calculation of target location information more accurate and provides a high-precision location reference for path correction and navigation control of unmanned food transport vehicles.

[0046] The fusion navigation method of this invention can adapt to complex and ever-changing transportation environments. Whether it is a change in lighting, dynamic movement of obstacles, or multi-device collaborative scenarios, it can maintain stable navigation performance, ensuring that unmanned food transport vehicles arrive at the target location efficiently and safely. It effectively improves the automation level and operational efficiency of food transportation, reduces the need for manual intervention, and has strong practical application value.

[0047] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A fusion navigation method for an unmanned food delivery vehicle, characterized in that, include: S1, Obtain the location information of the target to be reached; The target location information to be reached is the location information to be reached by the unmanned food delivery vehicle; S2, Based on the target location information, generate preset path information; S3, drive the unmanned food delivery vehicle to travel according to the preset path information and reach the target location.

2. The fusion navigation method for unmanned food delivery vehicles as described in claim 1, characterized in that, The step of generating preset path information based on target location information includes: S21, Perform SLAM modeling on the movement area of ​​the unmanned food delivery vehicle to obtain a set of path information for the movement area; the set of path information includes the path information of each point in the movement area; S22, Based on the target location information, query the path information of the corresponding point in the path information set; S23, confirm that the path information obtained from the query is the preset path information.

3. The fusion navigation method for unmanned food transport vehicles as described in claim 2, characterized in that, The unmanned food delivery vehicle travels according to the preset path information to reach the target location, including: S31, drive the unmanned food delivery vehicle to travel according to the preset path information; S32, while driving, detects targets in the surrounding environment and obtains target information; S33, Based on the target information, the preset path information is corrected to obtain updated preset path information; S34 drives the unmanned food delivery vehicle to travel according to the updated preset route information and reach the target location.

4. The fusion navigation method for unmanned food transport vehicles as described in claim 3, characterized in that, The target detection of the surrounding environment to obtain target information includes: S321 performs registration processing on the cameras, lidar, and millimeter-wave radar mounted on the unmanned food delivery vehicle; S322, using the camera, lidar and millimeter-wave radar to detect targets in the surrounding environment and obtain target information.

5. The fusion navigation method for unmanned food transport vehicles as described in claim 4, characterized in that, The method of using the camera, lidar, and millimeter-wave radar to detect targets in the surrounding environment and obtain target information includes: S3221 uses a camera to acquire images of the target area; S3222 uses lidar to collect point cloud data of the target area; S3223, Perform joint target detection on the target region image and the target region point cloud data to obtain target presence information; S3224, If the target information is true, use millimeter-wave radar to detect and locate the target to obtain the target location information; confirm the target information as target location information; If the target information is negative, the target information is confirmed to be non-existent.

6. The fusion navigation method for unmanned food transport vehicles as described in claim 5, characterized in that, The joint target detection of the target region image and target region point cloud data to obtain target presence information includes: S32231, Perform image detection processing on the target region image to obtain a first target location information sequence; S32232, Perform target point cloud matching processing on the target area point cloud data to obtain a second target location information sequence; the location information sequence is a sequence composed of location information obtained by detecting the target area image or target area point cloud data at several times; S32233, Perform coordinate transformation on the first target position information sequence and the second target position information sequence respectively to obtain a first three-dimensional position information sequence and a second three-dimensional position information sequence; S32234, Perform joint discrimination calculation on the first three-dimensional position information sequence and the second three-dimensional position information sequence to obtain fused position information and statistical discrimination value; S32235, determine whether the statistical discrimination value is less than a preset discrimination threshold to obtain a first discrimination result; if the first discrimination result is less than, determine that the target information exists; if the first discrimination result is not less than, determine that the target information exists.

7. The fusion navigation method for unmanned food transport vehicles as described in claim 6, characterized in that, The step of jointly discriminating the first three-dimensional location information sequence and the second three-dimensional location information sequence to obtain fused location information and statistical discriminant values ​​includes: Statistical distribution modeling is performed on the first three-dimensional location information sequence to obtain the first mean. First variance and the set of first probability distribution functions; Perform hypothesis testing on the first set of probability distribution functions to obtain the corresponding first test power value. ; Statistical distribution modeling is performed on the second three-dimensional location information sequence to obtain the second mean. Second variance The set of second probability distribution functions; Perform hypothesis testing on the second set of probability distribution functions to obtain the corresponding second test power value. ; The first mean, first variance, and first test power value are combined and calculated to obtain the statistical discriminant value. The expression for the statistical discriminant value is: ， ， in, is the scaling factor, F is the quantile function of the standard normal distribution, and t is the statistical discriminant value; The first three-dimensional position information sequence and the second three-dimensional position information sequence are fused to obtain fused position information.

8. A fusion navigation device for an unmanned food delivery vehicle, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the fusion navigation method of the unmanned food transport vehicle as described in any one of claims 1 to 7.

9. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by the computer, are used to execute the fusion navigation method of the unmanned food transport vehicle as described in any one of claims 1 to 7.

10. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the fusion navigation method for the unmanned food transport vehicle as described in any one of claims 1 to 7.

Images