Forward-Reach Forklift Navigation System and Method Based on Neural Network Model

Abstract

The present invention belongs to the technical field of stacker trucks. A reach stacker truck navigation system and method based on a neural network model are provided. The method includes: obtaining first dynamic and static information of ground obstacles and second dynamic and static information of goods in a shelf area from a video image of a front target area captured by a camera; using the neural network model to perform a collision probability analysis on the second dynamic and static information, matching the number of trajectories according to the analyzed collision probability, and planning multiple forward trajectories corresponding to the number of trajectories in the front target area according to the first dynamic and static information and the second dynamic and static information; when the reach stacker truck is about to reach the front target area, screening out one of the forward trajectories as the target forward trajectory, and controlling the reach stacker truck to travel according to the target forward trajectory. The present invention plans the trajectory of the reach stacker truck based on both ground obstacles and shelf goods, and has higher navigation safety.

Description

Technical Field

[0001] The present invention relates to the technical field of stacker trucks, and more particularly to a reach stacker navigation system and method based on a neural network model. Background Art

[0002] In the field of modern logistics warehousing, reach stacker trucks, as an important material handling equipment, are widely used in operation scenarios such as warehouse goods handling and stacking.

[0003] Previously, the navigation methods of reach stacker trucks mainly relied on traditional track navigation, magnetic strip navigation, etc. Track navigation requires laying special tracks on the ground, and the stacker truck travels along the tracks. Although this method has high stability, the cost of laying tracks is high, and when adjusting the warehouse layout later, it is difficult and costly to modify the tracks, lacking flexibility. Magnetic strip navigation is to paste magnetic strips on the ground, and the stacker truck uses sensors to identify the magnetic strips for navigation. However, magnetic strips are easily affected by the environment. For example, ground dust, moisture, etc. will interfere with the sensor's identification of the magnetic strips, resulting in a decrease in navigation accuracy. And in a complex warehouse environment, the route planning of magnetic strip navigation is relatively single and difficult to meet the diverse operation requirements. With the increasing complexity of the warehousing environment and the continuous improvement of the requirements for logistics operation efficiency, these existing navigation methods have gradually become difficult to meet the requirements of reach stacker trucks in terms of precise positioning, efficient path planning, and flexible adaptation to different operation scenarios.

[0004] Currently, a navigation method based on image recognition technology has been applied to reach stacker trucks to some extent. That is, a camera is arranged at a suitable position in the front of the reach stacker truck, and based on the video image captured by the camera, the distribution of the objects in front to be traveled is identified, and then a forward path is planned for the reach stacker truck. The reach stacker truck walks along this forward path without collision.

[0005] The above-mentioned navigation methods based on video images mainly focus on the distribution of ground objects. That is, the planned forward path can only ensure that the reach stacker truck does not collide with ground objects. However, in the warehousing area, some goods on the shelves may protrude outside the shelves due to improper placement, and these goods may collide with the lifting frame of the reach stacker truck with a certain height. The prior art lacks consideration of the goods on the shelves. Therefore, it is necessary to improve the video navigation method of the reach stacker truck. Summary of the Invention

[0006] In view of this, the present invention provides a reach stacker navigation method, system, computer storage medium, and computer program product based on a neural network model to solve the above technical problems.

[0007] The present invention discloses a navigation method for a reach truck based on a neural network model. The method includes the following steps: receiving a video image of a front target area captured by a camera deployed in front of the reach truck, extracting first dynamic and static information of obstacles located in the ground area and second dynamic and static information of goods located in the shelf area from the video image; using the neural network model to perform a collision probability analysis on the second dynamic and static information, matching the number of trajectories according to the collision probability obtained from the analysis, and planning multiple forward trajectories corresponding to the number of trajectories in the front target area according to the first dynamic and static information and the second dynamic and static information; when the reach truck is about to reach the front target area, screening out one of the forward trajectories as the target forward trajectory, and controlling the reach truck to travel according to the target forward trajectory.

[0008] Preferably, the using the neural network model to perform a collision probability analysis on the second dynamic and static information and matching the number of trajectories according to the collision probability obtained from the analysis includes: judging whether there is goods protruding from the edge of the shelf in the front target area according to the second dynamic and static information; if not, setting the collision probability to be lower than the probability threshold, and at this time, the number of trajectories obtained by matching is the first number of trajectories; if so, setting the collision probability to be higher than the probability threshold, and further: judging the protruding size and movement information of the target goods belonging to the goods protruding from the edge of the shelf according to the second dynamic and static information; if the movement information indicates that the target goods are in a static state, calculating the second number of trajectories according to the first protruding size of the target goods and the driving width of the front target area; if the movement information indicates that the target goods are in a shaking state, using the neural network model to perform a prediction analysis on the movement information to obtain the second protruding size of the target goods, and calculating the third number of trajectories according to the second protruding size of the target goods and the driving width of the front target area; wherein, the first number of trajectories is lower than the second number of trajectories and the third number of trajectories.

[0009] Preferably, when the collision probability is higher than the probability threshold, it further includes: determining the protruding size and distribution position of each target good, calculating the median value of each protruding size, and taking it as the target protruding size; and evaluating the distribution concentration degree of the target goods in the front target area according to each distribution position; obtaining a corresponding goods inclination coefficient according to the size of the target protruding size and the distribution concentration degree, and adjusting the second number of trajectories to the fourth number of trajectories according to the goods inclination coefficient, or adjusting the third number of trajectories to the fifth number of trajectories.

[0010] Further, according to the above-mentioned cargo tilt coefficient, the obtained second trajectory quantity can be adjusted to a slightly larger fourth trajectory quantity, or the third trajectory quantity can be adjusted to a slightly larger fifth trajectory quantity. Of course, when the distribution concentration degree of each target cargo is low, it indicates that the placement of these target cargos is not standardized and belongs to individual cases, not the overall non-standard placement situation mentioned above, and there will be no overall tilt situation. Therefore, the cargo tilt coefficient can be set to 1, that is, the obtained second trajectory quantity and third trajectory quantity are not adjusted.

[0011] Preferably, the distribution concentration degree is obtained by an evaluation model based on a classification algorithm.

[0012] Preferably, when the reach truck is about to reach the front target area, one of the forward trajectories is selected as the target forward trajectory, including: during the process of the reach truck approaching the front target area, the third dynamic and static information of the cargo located in the shelf area is obtained multiple times, and the deviation degree between the latest third dynamic and static information and the second dynamic and static information is evaluated; when the deviation degree is greater than the deviation threshold, it is set that when the reach truck is at a first distance from the front target area, one of the forward trajectories is selected as the target forward trajectory; when the deviation degree is less than the deviation threshold, it is set that when the reach truck is at a second distance from the front target area, one of the forward trajectories is selected as the target forward trajectory; wherein, the first distance is less than the second distance.

[0013] The present invention also discloses a reach truck navigation system based on a neural network model. The system includes a navigation controller and a camera deployed in front of the reach truck. The navigation controller is used to implement the following steps: receiving the video image of the front target area captured by the camera, extracting the first dynamic and static information of the obstacles located in the ground area from the video image, and extracting the second dynamic and static information of the cargo located in the shelf area; using the neural network model to perform collision probability analysis on the second dynamic and static information, matching the trajectory quantity according to the analyzed collision probability, and planning multiple forward trajectories corresponding to the trajectory quantity in the front target area according to the first dynamic and static information and the second dynamic and static information; when the reach truck is about to reach the front target area, one of the forward trajectories is selected as the target forward trajectory, and the reach truck is controlled to travel according to the target forward trajectory.

[0014] Preferably, when using the neural network model to analyze the collision probability of the second dynamic and static information and matching the number of trajectories according to the analyzed collision probability, it includes: judging whether there are goods protruding from the edge of the shelf in the front target area according to the second dynamic and static information; if not, setting the collision probability to be lower than the probability threshold, and at this time, the number of trajectories obtained by matching is the first number of trajectories; if so, setting the collision probability to be higher than the probability threshold, and further: judging the protruding size and movement information of the target goods protruding from the edge of the shelf according to the second dynamic and static information; if the movement information indicates that the target goods are in a static state, calculating the second number of trajectories according to the first protruding size of the target goods and the driving width of the front target area; if the movement information indicates that the target goods are in a swaying state, using the neural network model to predict and analyze the movement information to obtain the second protruding size of the target goods, and calculating the third number of trajectories according to the second protruding size of the target goods and the driving width of the front target area; wherein, the first number of trajectories is lower than the second number of trajectories and the third number of trajectories.

[0015] Preferably, when the collision probability is higher than the probability threshold, it further includes: determining the protruding size and distribution position of each target good, calculating the median value of each protruding size, and taking it as the target protruding size; and evaluating the distribution concentration degree of the target goods in the front target area according to each distribution position; obtaining the corresponding goods inclination coefficient according to the size of the target protruding size and the distribution concentration degree, and adjusting the second number of trajectories to the fourth number of trajectories according to the goods inclination coefficient, or adjusting the third number of trajectories to the fifth number of trajectories.

[0016] Further, according to the above-mentioned goods inclination coefficient, the previously obtained second number of trajectories can be adjusted to a slightly larger fourth number of trajectories, or the third number of trajectories can be adjusted to a slightly larger fifth number of trajectories. Of course, when the distribution concentration degree of each target good is low, it means that the placement of these target goods is not standardized and belongs to individual cases, not the above-mentioned overall non-standard placement situation, and there will be no overall inclination situation. Therefore, the goods inclination coefficient can be set to 1, that is, the previously obtained second number of trajectories and third number of trajectories are not adjusted.

[0017] Preferably, the distribution concentration degree is obtained by an evaluation model based on a classification algorithm.

[0018] Preferably, when the forward-stacking truck is about to reach the front target area, one of the forward trajectories is selected as the target forward trajectory, including: during the process of the forward-stacking truck approaching the front target area, the third dynamic and static information of the goods located in the shelf area is obtained multiple times, and the deviation degree between the latest third dynamic and static information and the second dynamic and static information is evaluated; when the deviation degree is greater than the deviation threshold, it is set that when the forward-stacking truck is at a first distance from the front target area, one of the forward trajectories is selected as the target forward trajectory; when the deviation degree is less than the deviation threshold, it is set that when the forward-stacking truck is at a second distance from the front target area, one of the forward trajectories is selected as the target forward trajectory; wherein, the first distance is less than the second distance.

[0019] The present invention also discloses an electronic device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and the computer program, when run by the processor, implements the method as described in any one of the preceding items.

[0020] The present invention also discloses a computer storage medium, which stores a computer program, and the computer program, when run by the processor, implements the method as described in any one of the preceding items.

[0021] The present invention also discloses a computer program product, which packages computer program codes, and the computer program codes, when run by the processor, implement the method as described in any one of the preceding items.

[0022] The beneficial effect of the present invention is that: the solution of the present invention simultaneously considers the influence of ground obstacles and irregularly placed goods on the forward movement of the forward-stacking truck, thereby planning a more reasonable forward trajectory, and thus can improve the driving safety of the forward-stacking truck in the storage area. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 It is a schematic flowchart of a navigation method for a forward-stacking truck based on a neural network model disclosed in an embodiment of the present invention.

[0025] Figure 2It is a schematic structural diagram of a forward-reaching stacker navigation system based on a neural network model disclosed in an embodiment of the present invention. Specific Embodiments

[0026] The following specific embodiments illustrate the implementation manners of the present application. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0027] In addition, the technical features involved in different implementation manners of the present application described below can be combined with each other as long as they do not conflict with each other.

[0028] The forward-reaching stacker in the present invention has the function of self-movement, that is, it can at least realize the path self-planning and walking execution in the warehousing area in cooperation with a camera and a navigation controller. Since it belongs to the prior art, other functional components will not be elaborated.

[0029] As Figure 1 shown, an embodiment of the present invention discloses a navigation method for a forward-reaching stacker based on a neural network model. The method includes the following steps: receiving a video image of a front target area captured by a camera deployed in front of the forward-reaching stacker, extracting first dynamic and static information of obstacles located in the ground area from the video image, and extracting second dynamic and static information of goods located in the shelf area.

[0030] In this step, the present invention deploys a camera in front of the forward-reaching stacker (for example, on the top of the lifting rack), and its function is to capture in real time the video image of the front target area that the forward-reaching stacker needs to drive through. From the captured video image, through specific image recognition algorithms and technologies, the first dynamic and static information of obstacles located in the ground area is extracted. The ground obstacles here may include sundries, goods temporarily placed on the ground, other forward-reaching stackers, etc. The first dynamic and static information includes whether the ground obstacles are stationary or in a moving state, as well as their position, shape, size and other characteristics, which helps to plan a safe driving path subsequently.

[0031] At the same time, the second dynamic and static information of the goods located in the shelf area is extracted from the video image. The second dynamic and static information includes whether there is a situation where the goods protrude from the shelf, the size of the goods protruding from the shelf, whether there are dynamic signs such as shaking of the goods on the shelf, etc. As described in the background technology, if the goods on the shelf protrude from the shelf, there is a possibility of collision with the forward-reaching stacker. Therefore, obtaining this part of information is of great significance for ensuring the driving safety of the stacker.

[0032] Use a neural network model to analyze the collision probability of the second dynamic and static information, match the number of trajectories based on the analyzed collision probability, and plan multiple forward trajectories corresponding to the number of trajectories in the front target area according to the first dynamic and static information and the second dynamic and static information.

[0033] In this step, the present invention pre-constructs a neural network model, which has been trained with a large amount of data and has powerful pattern recognition and data analysis capabilities. It can accurately calculate the probability of a reach truck colliding with the goods on the shelf under different driving conditions according to the characteristic information of these goods. For example, when the neural network model detects that the goods on a certain shelf protrude significantly, it will calculate the probability of the reach truck colliding with it according to information such as the size and angle of the protrusion and the size of the reach truck itself.

[0034] Before the reach truck reaches the front target area, multiple forward trajectories are planned according to the above first dynamic and static information of the ground obstacles, the second dynamic and static information of the shelf goods, and path planning algorithms (such as Dijkstra algorithm, Rapidly-exploring Random Tree (RRT) algorithm, A* algorithm, etc.) for the reach truck to further screen.

[0035] At the same time, the present invention also uses the pre-constructed neural network model to analyze the collision probability of the second dynamic and static information of the goods on the shelf, and determines the number of the above forward trajectories based on the analysis result of the collision probability. When the collision probability between the reach truck and the goods on the shelf in the front target area is higher, in order to ensure the safe driving of the reach truck, more alternative trajectories need to be planned, such as 7 - 10, to increase the possibility of finding a safe path in the follow-up; on the contrary, if the collision probability is lower, the number of trajectories can be appropriately reduced, such as 3 - 5, to improve the efficiency of path planning.

[0036] It should be noted that when planning, the planned forward trajectories should not only avoid ground obstacles but also meet the driving requirements of the reach truck as much as possible, such as the smoothness of driving and the rationality of the turning radius.

[0037] When the reach truck is about to reach the front target area, select one from each of the forward trajectories as the target forward trajectory, and control the reach truck to drive according to the target forward trajectory.

[0038] In this step, when the reach truck is about to reach the front target area, the reach truck needs to select an optimal trajectory from the multiple forward trajectories planned above according to the actual situation, and connect the selected target forward trajectory with the currently executing trajectory, so as to smoothly enter the front target area.

[0039] The solution of the present invention simultaneously considers the influence of ground obstacles and irregularly placed goods on the shelf on the movement of the reach truck, thereby planning a more reasonable forward trajectory, and thus improving the driving safety of the reach truck in the storage area.

[0040] Preferably, the neural network model is used to analyze the collision probability of the second dynamic and static information, and the number of trajectories is matched according to the analyzed collision probability, including: judging whether there are goods protruding from the shelf edge on the shelf in the front target area according to the second dynamic and static information; if not, setting the collision probability to be lower than the probability threshold, and at this time, the number of the matched trajectories is the first number of trajectories; if so, setting the collision probability to be higher than the probability threshold, and further: judging the protruding size and movement information of the target goods protruding from the shelf edge according to the second dynamic and static information; if the movement information indicates that the target goods are in a static state, calculating the second number of trajectories according to the first protruding size of the target goods and the driving width of the front target area; if the movement information indicates that the target goods are in a shaking state, using the neural network model to perform predictive analysis on the movement information to obtain the second protruding size of the target goods, and calculating the third number of trajectories according to the second protruding size of the target goods and the driving width of the front target area; wherein, the first number of trajectories is lower than the second number of trajectories and the third number of trajectories.

[0041] In the embodiment of the present invention, the present invention sets different trajectory number determination strategies according to different types represented by the second dynamic and static information, specifically as follows: the second dynamic and static information includes whether there are target goods protruding from the shelf edge on the shelf, the size of these target goods protruding from the shelf edge, and also includes the movement information of these target goods, that is, whether they are in a static state or in a moving state. The above information can be used to analyze the collision risk of the reach truck driving in the front target area, and then determine the number of alternative forward trajectories to be planned subsequently.

[0042] 1) The goods do not protrude from the shelf, that is, there are no target goods: When the judgment result is that there are no goods protruding from the shelf edge on the shelf, since the possibility of the reach truck colliding with the goods on the shelf is extremely low in this case, the collision probability can be set to be lower than the probability threshold. The probability threshold here is a value set according to actual experience and safety standards, and is used to define the level of collision risk. In this low-risk case, in order to simplify the path planning process and improve the planning efficiency, the number of matched trajectories is set to the smallest first number of trajectories. The trajectory planned in this case is only a relatively direct and efficient path planned based on factors such as ground obstacle information and the driving destination of the reach truck. The first number of trajectories is, for example, 1 or 2.

[0043] 2) When there is a target cargo protruding from the shelf, i.e., the target cargo exists: If it is determined that there is a target cargo protruding from the edge of the shelf on the shelf, set the collision probability higher than the probability threshold. Specifically, the second dynamic and static information also includes the motion state of the target cargo. If the motion state indicates that the target cargo is in a stationary state, at this time, the size of the target cargo protruding from the edge of the shelf, i.e., the first protruding size, is known and fixed. The second trajectory quantity is directly calculated based on the first protruding size of the target cargo and the driving width of the front target area. For example, , where is a constant greater than 1.

[0044] Among them, when the first protruding size of the target cargo is larger, the present invention sets and plans more alternative trajectories. In this way, when it is found that the actual protruding size is larger than the first protruding size when arriving at the area where the target cargo is located, for example, other alternative trajectories can be quickly switched without the need for temporary planning; and when the driving width of the front target area is larger, since the driving space is very sufficient, correspondingly fewer alternative trajectories are set and planned (the distance between each alternative trajectory is larger), so that the reach truck can be controlled to walk on a trajectory farther from the target cargo.

[0045] In addition, the motion state may also indicate that the target cargo is in a shaking state. For example, when the reach truck behind the shelf is sorting goods, the target cargo may be affected by the sorting and shake. The final protruding size of the target cargo in the shaking state cannot be determined at present. Therefore, the present invention further uses the pre-constructed neural network model to deeply analyze the motion information of the target cargo in the shaking state, that is, based on the motion information such as the shaking law, starting position, shaking speed and direction of the target cargo, predict the possible second protruding size of the target cargo in the next period of time. Then, use the foregoing method to calculate the corresponding third trajectory quantity: , where is a constant greater than 1.

[0046] Preferably, when the collision probability is higher than the probability threshold, it further includes: determining the protruding size and distribution position of each target cargo, calculating the median value of each protruding size, and using it as the target protruding size; and evaluating the distribution concentration degree of the target cargo in the front target area according to each distribution position; obtaining the corresponding cargo inclination coefficient according to the size of the target protruding size and the distribution concentration degree, and adjusting the second trajectory quantity to the fourth trajectory quantity according to the cargo inclination coefficient, or adjusting the third trajectory quantity to the fifth trajectory quantity.

[0047] In the embodiments of the present invention, the above-mentioned embodiments only consider the prominent dimensions of each target good itself to determine the number of planned alternative trajectories. However, in fact, when the prominent dimensions of the target goods protruding from the edge of the shelf are generally large and these target goods are concentrated in distribution, it indicates that these target goods are not placed in a standardized manner as a whole. In this case, these target goods are more likely to tilt as a whole, that is, gradually tilt outwards under the action of gravity. When the reach truck has not reached the forward target driving area, the overall tilt degree of these target goods is not high. At this time, the above-mentioned second trajectory number and third trajectory number can be determined according to the median value of the prominent dimensions of these target goods. When the reach truck reaches the forward target driving area, due to the driving vibration of the reach truck, these target goods will accelerate to tilt, resulting in the median value of the latest prominent dimension becoming much larger than before, and thus there is a greater probability of collision with the reach truck.

[0048] Therefore, in view of the above situation, the present invention sets to evaluate the distribution concentration degree of each target good according to its distribution position, and then obtains the corresponding goods tilt coefficient according to the above-mentioned target prominent dimension and the size of the evaluated distribution concentration degree. The goods tilt coefficient represents the difference between the actual prominent dimension of these target goods when the reach truck reaches the forward target area (specifically, the distribution area of these target goods) and the initial prominent dimension of these target goods when the reach truck has not reached the forward target area. In other words, the larger the goods tilt coefficient (for example, 1.2), the larger the above-mentioned difference; the smaller the goods tilt coefficient (for example, 1.0, 1.1), the smaller the above-mentioned difference.

[0049] It should be noted that the corresponding relationship between the goods tilt coefficient and the distribution concentration degree and the target prominent dimension can also be determined in the form of a comparison table, and this corresponding relationship is obviously positively correlated, which will not be elaborated here. An example of the comparison table is as follows:

[0050] Furthermore, according to the above-mentioned goods tilt coefficient, the previously obtained second trajectory number can be adjusted to a slightly larger fourth trajectory number, or the third trajectory number can be adjusted to a slightly larger fifth trajectory number. Of course, when the distribution concentration degree of each target good is relatively low, it indicates that the non-standard placement of these target goods is an individual case and does not belong to the above-mentioned overall non-standard placement situation, and there will be no overall tilt situation. Therefore, the goods tilt coefficient can be set to 1, that is, the previously obtained second trajectory number and third trajectory number are not adjusted. The adjustment here is, for example, a multiplication operation.

[0051] Preferably, the distribution concentration degree is obtained by an evaluation model based on a classification algorithm.

[0052] In an embodiment of the present invention, a model dedicated to evaluating the above-mentioned concentration degree of the distribution is pre-constructed. Such models can be constructed and trained based on algorithms such as SVM (Support Vector Machine), decision tree, and random forest. The specific process will not be elaborated here.

[0053] Preferably, when the reach truck is about to reach the front target area, one of the forward trajectories is selected as the target forward trajectory, including: during the process of the reach truck approaching the front target area, the third dynamic and static information of the goods located in the shelf area is obtained multiple times, and the deviation degree between the latest third dynamic and static information and the second dynamic and static information is evaluated; when the deviation degree is greater than the deviation threshold, it is set that when the reach truck is at a first distance from the front target area, one of the forward trajectories is selected as the target forward trajectory; when the deviation degree is less than the deviation threshold, it is set that when the reach truck is at a second distance from the front target area, one of the forward trajectories is selected as the target forward trajectory; wherein, the first distance is less than the second distance.

[0054] In an embodiment of the present invention, since the multiple forward trajectories obtained by the foregoing planning are planned when the reach truck is at a certain distance (such as 5 m) from the front target area, during the process of its traveling within this certain distance, the dynamic and static information of the goods located in the shelf area may have changed greatly. Therefore, it is necessary to optimize the screening timing of the target forward trajectory.

[0055] Specifically, in the present invention, during the process of the reach truck approaching the edge of the front target area, the third dynamic and static information of the goods located in the shelf area is obtained multiple times, and the deviation degree between it and the initial second dynamic and static information is evaluated. The evaluation of the deviation degree is at least based on the above-mentioned protrusion size and motion information. For example, when the protrusion size of the target goods becomes larger and / or the motion information becomes stronger (the speed increases, the shaking amplitude becomes larger, etc.) at the front and rear moments, the corresponding deviation degree is larger; conversely, when the protrusion size of the target goods becomes smaller and / or the motion information becomes weaker (the speed slows down, the shaking amplitude becomes smaller, etc.) at the front and rear moments, the corresponding deviation degree is also larger. In the reverse case of the above situation, the corresponding deviation degree is smaller.

[0056] When the above deviation degree is greater than the deviation threshold, the present invention sets to screen out one from each forward trajectory as the target forward trajectory when the reach truck is closer to the target area ahead, so as to improve the passing safety of the screened target forward trajectory; while when the above deviation degree is less than the deviation threshold, the present invention sets to screen out one from each forward trajectory as the target forward trajectory when the reach truck is farther from the target area ahead, so as to determine the target forward trajectory as early as possible, thereby reserving more time for the connection between the target forward trajectory and the current driving trajectory, and reducing the probability of the reach truck stopping and waiting (calculating the connection trajectory).

[0057] As Figure 2 shown, an embodiment of the present invention also discloses a reach truck navigation system based on a neural network model. The system includes a navigation controller and a camera deployed in front of the reach truck. The navigation controller is used to implement the following steps: receiving a video image of the target area ahead captured by the camera, extracting first dynamic and static information of obstacles located in the ground area and second dynamic and static information of goods located in the shelf area from the video image; using the neural network model to perform a collision probability analysis on the second dynamic and static information, matching the number of trajectories according to the analyzed collision probability, and planning multiple forward trajectories corresponding to the number of trajectories in the target area ahead according to the first dynamic and static information and the second dynamic and static information; when the reach truck is about to reach the target area ahead, screening out one from each of the forward trajectories as the target forward trajectory, and controlling the reach truck to drive according to the target forward trajectory.

[0058] An embodiment of the present invention also discloses an electronic device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor. When the computer program is run by the processor, it implements the method described in any one of the previous items.

[0059] An embodiment of the present invention also discloses a computer storage medium, which stores a computer program. When the computer program is run by the processor, it implements the method described in any one of the previous items.

[0060] An embodiment of the present invention also discloses a computer program product, which packages computer program code. When the computer program code is run by the processor, it implements the method described in any one of the previous items.

[0061] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A navigation method for reach trucks based on a neural network model, characterized in that: The method includes the following steps: Receiving a video image of a front target area captured by a camera deployed in front of a reach truck, obtaining first dynamic and static information of obstacles located in the ground area from the video image, and obtaining second dynamic and static information of goods located in the shelf area; Using a neural network model to perform a collision probability analysis on the second dynamic and static information, matching the number of trajectories according to the analyzed collision probability, and planning multiple forward trajectories corresponding to the number of trajectories in the front target area based on the first dynamic and static information and the second dynamic and static information; When the reach truck is about to reach the front target area, screening out one of the forward trajectories as the target forward trajectory, and controlling the reach truck to travel according to the target forward trajectory; Using a neural network model to perform a collision probability analysis on the second dynamic and static information, and matching the number of trajectories according to the analyzed collision probability, including: Judging whether there are goods protruding from the shelf edge on the shelf in the front target area according to the second dynamic and static information; if not, setting the collision probability lower than the probability threshold, and at this time, the number of trajectories matched is the first number of trajectories; If so, setting the collision probability higher than the probability threshold, and further: Judging the protruding size and movement information of the target goods that protrude from the shelf edge according to the second dynamic and static information; If the movement information indicates that the target goods are in a static state, calculating a second number of trajectories according to the first protruding size of the target goods and the driving width of the front target area; If the movement information indicates that the target goods are in a shaking state, using a neural network model to perform a prediction analysis on the movement information to obtain a second protruding size of the target goods, and calculating a third number of trajectories according to the second protruding size of the target goods and the driving width of the front target area; Wherein, the first number of trajectories is lower than the second number of trajectories and the third number of trajectories; When the collision probability is higher than the probability threshold, it further includes: Determining the protruding size and distribution position of each target good, calculating the median value of each protruding size, and taking it as the target protruding size; and evaluating the distribution concentration degree of the target goods in the front target area according to each distribution position; Obtaining a corresponding goods inclination coefficient according to the size of the target protruding size and the distribution concentration degree, and adjusting the second number of trajectories to a fourth number of trajectories according to the goods inclination coefficient, or adjusting the third number of trajectories to a fifth number of trajectories.

2. The navigation method of the reach truck based on the neural network model according to claim 1, characterized in that: The distribution concentration degree is obtained by an evaluation model based on a classification algorithm.

3. The navigation method of the reach truck based on the neural network model according to claim 1, characterized in that: When the reach truck is about to reach the front target area, screening out one of the forward trajectories as the target forward trajectory, including: During the process of the reach truck approaching the front target area, obtaining the third dynamic and static information of the goods located in the shelf area multiple times, and evaluating the deviation degree between the latest third dynamic and static information and the second dynamic and static information; When the degree of deviation is greater than the deviation threshold, it is set that when the reach truck is at a first distance from the front target area, one is selected from each of the forward trajectories as the target forward trajectory; When the degree of deviation is less than the deviation threshold, it is set that when the reach truck is at a second distance from the front target area, one is selected from each of the forward trajectories as the target forward trajectory; Wherein, the first distance is less than the second distance.

4. A forward-reaching stacker navigation system based on a neural network model, characterized in that: The system includes a navigation controller and a camera deployed in front of the reach truck, and the navigation controller is used to implement the following steps: Receive the video image of the front target area captured by the camera, extract the first dynamic and static information of the obstacles located in the ground area and the second dynamic and static information of the goods located in the shelf area from the video image; Use a neural network model to perform a collision probability analysis on the second dynamic and static information, match the number of trajectories according to the analyzed collision probability, and plan multiple forward trajectories corresponding to the number of trajectories in the front target area according to the first dynamic and static information and the second dynamic and static information; When the reach truck is about to reach the front target area, one is selected from each of the forward trajectories as the target forward trajectory, and the reach truck is controlled to travel according to the target forward trajectory; The using a neural network model to perform a collision probability analysis on the second dynamic and static information and matching the number of trajectories according to the analyzed collision probability includes: Judge whether there are goods protruding from the shelf edge on the shelf in the front target area according to the second dynamic and static information; if not, set the collision probability to be lower than the probability threshold, and at this time the number of trajectories matched is the first number of trajectories; If so, set the collision probability to be higher than the probability threshold, and further: Judge the protruding size and movement information of the target goods that protrude from the shelf edge according to the second dynamic and static information; If the movement information indicates that the target goods are in a static state, calculate the second number of trajectories according to the first protruding size of the target goods and the driving width of the front target area; If the movement information indicates that the target goods are in a shaking state, use a neural network model to perform a prediction analysis on the movement information to obtain the second protruding size of the target goods, and calculate the third number of trajectories according to the second protruding size of the target goods and the driving width of the front target area; Wherein, the first number of trajectories is lower than the second number of trajectories and the third number of trajectories; When the collision probability is higher than the probability threshold, it further includes: Determine the protruding size and distribution position of each of the target goods, calculate the median value of each of the protruding sizes, and use it as the target protruding size; and evaluate the distribution concentration degree of the target goods in the front target area according to each of the distribution positions; Obtain a corresponding cargo tilt coefficient based on the magnitude of the target protrusion dimension and the degree of distribution concentration, and adjust the second track quantity to a fourth track quantity according to the cargo tilt coefficient, or adjust the third track quantity to a fifth track quantity.

5. An electronic device, characterized in that: The electronic device includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and the computer program, when run by the processor, implements the method according to any one of claims 1-3.

6. A computer storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program, when run by the processor, implements the method according to any one of claims 1-3.

7. A computer program product, characterized in that: The computer program product is pre-packaged with computer program code, and the computer program code, when run by the processor, implements the method according to any one of claims 1-3.

Images