A Local Mapping and Navigation Method for a Campus Logistics Cart

Abstract

A method for local mapping and navigation of a logistics cart in a park. The logistics cart uses a 16-line lidar as a perception processor, adopts lidar slam positioning technology, pre-establishes a high-precision map and a navigation vector map of the park, and uses the point cloud information collected in real time to determine the current pose of the logistics cart by using the NDT registration algorithm. Global path navigation planning is carried out according to the input destination and the current location, and a new type of local cost map is designed. The efficiency of logistics work is improved, and automation, intelligence, and unmanned operation of the logistics park are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and particularly relates to a method for local mapping and navigation of a logistics vehicle in a park. Background Art

[0002] According to incomplete statistics, there are 478 national-level economic development zones, export processing zones, bonded areas, etc. in China, 1,170 provincial-level development zones of various types, and more than 22,000 industrial parks and manufacturing parks of various types. There are countless university campuses, residential parks, etc. In the social environment of "cost reduction, efficiency improvement, energy conservation and environmental protection", the logistics industry in closed parks is also facing many problems such as high operating costs and difficulty in recruiting drivers. Therefore, it is necessary to improve the navigation of the logistics vehicles in the existing logistics parks to improve the efficiency of logistics operations such as loading, transportation, receiving, and warehousing, and to achieve high-quality transformation and upgrading development. The automation, intelligence, and unmanned operation of logistics parks have gradually become a development trend. Summary of the Invention

[0003] In order to overcome the above-mentioned deficiencies in technology, the present invention provides a method for local mapping and navigation of a logistics vehicle in a park to improve the efficiency of logistics operations.

[0004] The technical solution adopted by the present invention to overcome its technical problems is as follows:

[0005] A method for local mapping and navigation of a logistics vehicle in a park includes the following steps:

[0006] a) The logistics vehicle uses a lidar as a perception processor;

[0007] b) The logistics vehicle uses the lidar to establish a high-precision map and a navigation vector map of the logistics park;

[0008] c) The logistics vehicle determines the current pose of the logistics vehicle by using the point cloud information collected in real time;

[0009] d) The logistics vehicle performs global path navigation planning according to the input destination and the current position, and calculates a local planning path with the optimal path cost according to the global path navigation planning;

[0010] e) The sampling speed and angular velocity of the local planning path with the optimal path cost are sent to the chassis of the logistics vehicle to complete the movement of the logistics vehicle.

[0011] Preferably, in step a), the logistics vehicle uses a 16-line lidar.

[0012] Preferably, in step b), the lidar slam positioning algorithm is used to establish a high-precision map and a navigation vector map of the logistics park.

[0013] Preferably, in step c), the logistics trolley uses the real-time collected point cloud information and adopts the NDT registration algorithm to determine the current pose of the logistics trolley.

[0014] Further, step d) includes the following steps:

[0015] d-1) Load the real-time 16-line laser point cloud and read the road drivable area flag of the navigation vector map;

[0016] d-2) Screen out the clustered targets within the road drivable area flag;

[0017] d-3) Load the predicted trajectory of the clustered target, set the prediction time coefficient K, and generate a multi-layer local cost map; d-4) The global path navigation obtains the global path according to the read navigation vector map, in accordance with the starting point and the ending point;

[0018] d-5) Intercept the global path, set the size of the local cost map according to the speed of the logistics trolley, and intercept the part of the global path in the local cost map as the reference path for local path navigation;

[0019] d-6) Use the dwa algorithm according to the reference path and the local cost map to obtain multiple drivable paths;

[0020] d-7) Calculate the optimal path cost using the optimal path cost = maximum speed of the logistics trolley * logistics trolley speed cost + coincidence degree with the global reference path * coincidence degree cost + degree of road centering * centering cost, and select the local planning path with the optimal path cost.

[0021] Further, in step d-3), the multi-layer local cost map is composed of a wayarea layer, a clustered obstacle layer, an obstacle predicted trajectory layer, and a laser point cloud layer.

[0022] Further, the above-mentioned wayarea layer is to preset the road drivable area according to the navigation vector map.

[0023] Further, the above-mentioned clustered obstacle layer projects the 3d obstacle onto the 2d plane, specifies the height screening rule, uniformly sets the height coordinate in the xyz coordinates of all obstacles to the height coordinate of the local cost map, formulates the inflation coefficient, and inflates the existing obstacle grids according to the inflation coefficient ratio to leave a safety space.

[0024] Further, the grid size of the above-mentioned obstacle predicted trajectory layer is set according to different obstacle types, a calibration time threshold is set for the predicted trajectory, and calibration is performed according to the actual working conditions using the calibration time threshold, and an inflation coefficient is formulated to inflate the grids of the obstacle predicted trajectory layer.

[0025] Further, the above laser point cloud projects all point clouds onto the cost map height plane.

[0026] The beneficial effects of the present invention are as follows: The logistics vehicle uses a 16-line lidar as a perception processor, adopts the lidar slam positioning technology, pre-establishes a high-precision map and a navigation vector map of the park, utilizes the point cloud information collected in real time, and adopts the NDT registration algorithm to determine the current pose of the logistics vehicle. Global path navigation planning is carried out according to the input destination and the current position. The efficiency of logistics work is improved, and automation, intelligence, and unmanned operation of the logistics park are realized. Specific embodiments

[0027] The following further describes the present invention.

[0028] A local mapping and navigation method for a logistics vehicle in a park includes the following steps:

[0029] a) The logistics vehicle uses a lidar as a perception processor.

[0030] b) The logistics vehicle uses the lidar to establish a high-precision map and a navigation vector map of the logistics park.

[0031] c) The logistics vehicle determines the current pose of the logistics vehicle by using the point cloud information collected in real time.

[0032] d) The logistics vehicle performs global path navigation planning according to the input destination and the current position, and calculates a local planning path with the optimal path cost according to the global path navigation planning.

[0033] e) The sampling speed and angular velocity of the local planning path with the optimal path cost are sent to the chassis of the logistics vehicle to complete the movement of the logistics vehicle.

[0034] Embodiment 1:

[0035] In an embodiment of the present invention, in step a), the logistics vehicle uses a 16-line lidar. In step b), the lidar slam positioning algorithm is used to establish a high-precision map and a navigation vector map of the logistics park. In step c), the logistics vehicle uses the point cloud information collected in real time and adopts the NDT registration algorithm to determine the current pose of the logistics vehicle. The GPS and IMU fusion technology can also be used for repositioning and supplementing

[0036] Embodiment 2:

[0037] Step d) includes the following steps:

[0038] d-1) Load the real-time 16-line laser point cloud, and read the road drivable area flag (wayarea) of the navigation vector map;

[0039] d-2) Screen out the clustered targets within the road drivable area sign (plus a height screening rule: height 0 - 1.5 meters);

[0040] d-3) Load the predicted trajectories of the clustered targets, set the prediction time coefficient K, and generate a multi-layer local cost map; d-4) The global path navigation obtains the global path according to the read navigation vector map, based on the starting point and the ending point;

[0041] d-5) Intercept the global path, set the local cost map size according to the speed of the logistics cart (default 10m * 10m), and intercept the part of the global path in the local cost map as the reference path for local path navigation; d-6) Use the dwa algorithm based on the reference path and the local cost map to obtain multiple drivable paths;

[0042] d-7) Calculate the optimal path cost using the formula: optimal path cost = maximum speed of the logistics cart * speed cost of the logistics cart + coincidence degree with the global reference path * coincidence cost + degree of road centralization * centralization cost, and select the local planning path with the optimal path cost.

[0043] Example 3:

[0044] In step d-3), the multi-layer local cost map consists of a wayarea layer, a clustered obstacle layer, an obstacle predicted trajectory layer, and a laser point cloud layer.

[0045] The above-mentioned wayarea layer pre-sets the road drivable area according to the navigation vector map, and it is a basic layer.

[0046] Since the local cost map is a 2D map, the above-mentioned clustered obstacle layer projects 3D obstacles onto the 2D plane, specifying a height screening rule: clustered obstacles within 0 - 1.5 meters. The height coordinates in the xyz coordinates of all obstacles are uniformly set to the height coordinate of the local cost map, and an inflation coefficient is formulated to inflate the existing obstacle grids according to the inflation coefficient ratio, leaving a safety space.

[0047] The grid size of the above-mentioned obstacle predicted trajectory layer is set according to different obstacle types, that is, the grid sizes for people and vehicles are different. A calibration time threshold is set for the predicted trajectory, and it is calibrated according to the actual working conditions using the calibration time threshold. An inflation coefficient is formulated. Similar to the obstacle layer, it is a top view projection, and the grids of the obstacle predicted trajectory layer are inflated.

[0048] For each layer of the cost map, it can be selectively used. Configuration options are set, and at most they can be superimposed, and finally a complete local cost map is generated.

[0049] Further, the above-mentioned laser point cloud projects all point clouds onto the cost map height plane. Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for local map building and navigation of a campus logistics cart, characterized in that, It includes the following steps: a) The logistics vehicle uses a lidar as the perception processor; b) The logistics vehicle uses the lidar to build a high-precision map and a navigation vector map of the logistics park; c) The logistics vehicle determines its current pose using the point cloud information collected in real time; d) The logistics vehicle conducts global path navigation planning based on the input destination and current location, and calculates a local planning path with the optimal path cost according to the global path navigation planning; e) The sampling speed and angular velocity of the local planning path with the optimal path cost are sent to the chassis of the logistics vehicle to complete the movement of the logistics vehicle; Step d) includes the following steps: d-1) Load the real-time 16-line laser point cloud and read the road drivable area markers of the navigation vector map; d-2) Screen out the clustered targets within the road drivable area markers; d-3) Load the predicted trajectories of the clustered targets, set the prediction time coefficient K, and generate a multi-layer local cost map; d-4) The global path navigation reads the navigation vector map and obtains the global path according to the starting point and the ending point; d-5) Intercept the global path, set the size of the local cost map according to the speed of the logistics vehicle, and intercept the part of the global path in the local cost map as the reference path for local path navigation; d-6) Use the dwa algorithm to obtain multiple drivable paths based on the reference path and the local cost map; d-7) Calculate the optimal path cost using the formula: optimal path cost = maximum speed of the logistics vehicle * speed cost of the logistics vehicle + degree of coincidence with the global reference path * coincidence cost + degree of road centering * centering cost, and select the local planning path with the optimal path cost; In step d-3), the multi-layer local cost map consists of a wayarea layer, a clustered obstacle layer, an obstacle predicted trajectory layer, and a laser point cloud layer; The said wayarea layer pre-sets the road drivable area according to the navigation vector map.

2. The local map building and navigation method for the park logistics vehicle according to claim 1, characterized in that: In step a), the logistics vehicle uses a 16-line lidar.

3. The method for local mapping and navigation of a campus logistics vehicle according to claim 1, wherein: In step b), the lidar slam positioning algorithm is adopted to build a high-precision map and a navigation vector map of the logistics park.

4. The local mapping and navigation method for the park logistics vehicle according to claim 2, characterized in that: In step c), the logistics vehicle uses the point cloud information collected in real time and the NDT registration algorithm to determine its current pose.

5. The local map building and navigation method for the park logistics vehicle according to claim 1, characterized in that: The said clustered obstacle layer projects the 3d obstacles onto a 2d plane, specifies the height screening rule, uniformly sets the height coordinates in the xyz coordinates of all obstacles to the height coordinates of the local cost map, formulates the inflation coefficient, and inflates the existing obstacle grids according to the inflation coefficient ratio to leave a safety space.

6. The method for local map building and navigation of the park logistics vehicle according to claim 1, wherein: The grid size of the said obstacle predicted trajectory layer is set according to different obstacle types, a calibration time threshold is set for the predicted trajectory, and it is calibrated according to the actual working conditions using the calibration time threshold, and the inflation coefficient is formulated to inflate the grids of the obstacle predicted trajectory layer.

7. The local map building and navigation method for the park logistics vehicle according to claim 1, characterized in that: The said laser point cloud layer projects all the point clouds onto the cost map height plane.