An AGV path planning method based on two-dimensional and three-dimensional data

Abstract

The present invention discloses an AGV path planning method based on two-dimensional and three-dimensional data to solve the problems that the traditional DWA algorithm cannot handle moving obstacles and will fall into local optimality; specifically, it includes: establishing the working space of the AGV, adding information about the starting point, target point, and known obstacles; using the RRT* algorithm to plan the global path of the AGV; training an obstacle recognition model to process the obstacles input by the industrial camera and determine the accurate position of the obstacles relative to the AGV during movement; selecting the speed combination of the AGV based on the obstacle distribution, and then combining the distance and speed information of the obstacles measured by the lidar to perform local path optimization through an improved DWA algorithm, so that the AGV reaches the specified position. The method proposed by the present invention realizes real-time local obstacle avoidance of the AGV and improves the automation ability of AGV path planning.

Description

Technical Field

[0001] The present invention relates to the technical field of AGV path planning, and particularly to an AGV path planning method based on two-dimensional and three-dimensional data. Background Art

[0002] With the maturity of AGV technology, AGVs are used more and more frequently. And the path planning of AGVs, as a key technology, determines the working efficiency and safety of AGVs.

[0003] The path planning algorithms of AGVs include global planning algorithms such as Dijkstra, A*, RRT*, D*, etc., and local planning algorithms such as DWA, TEB, MPC, etc. The global planning algorithm can make the AGV move from the starting point to the end point with only one planning, but it can only handle the situation where the size and position of obstacles are known, and due to the design of the AGV structure, etc., there may be an error between the theoretical position and the actual position. The local planning algorithm can detect dynamic obstacles and has a relatively small error; however, lacking the guidance of the global path may cause the problem of a long path and increase the calculation amount at the same time. Currently, many algorithms combining global planning and local planning have also emerged, but the processing effect of these algorithms on dynamic obstacles is not ideal, and the calculation amount is also relatively large. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention proposes an AGV path planning method based on two-dimensional and three-dimensional data. By combining an industrial camera and a lidar, and adopting an algorithm combining RRT* and improved DWA, the obstacle avoidance function of the AGV for dynamic obstacles is improved, and while reducing the calculation amount, the rationality of speed combination selection is improved.

[0005] To achieve the above technical objectives, the present invention provides the following technical solutions:

[0006] An AGV path planning method based on two-dimensional and three-dimensional data, which specifically includes the following steps:

[0007] S1. Establish a two-dimensional motion space of the AGV, and add the starting position, target position and known obstacle information of the AGV in the two-dimensional motion space;

[0008] S2. Use the RRT* algorithm to plan a global path in the two-dimensional working space of the AGV;

[0009] S3. Install an industrial camera on the AGV, control the AGV to move in the three-dimensional working space, and collect obstacle images during the movement through the industrial camera; use the obstacle images as input to train an obstacle recognition model; use the trained model to determine the position of the obstacle relative to the AGV;

[0010] S4. The AGV moves along the global path planned by the RRT* algorithm, determines the distribution of surrounding obstacles according to the trained obstacle recognition model, selects the speed combination of the AGV based on the obstacle distribution, and then combines the distance and speed information of the obstacles measured by the lidar to optimize the local path through the improved DWA algorithm, so that the AGV reaches the specified position.

[0011] Further, step S2 specifically includes:

[0012] S21. Create a search tree directory Search for the AGV, define the starting point of the AGV as the first point of the search tree, and define the exploration step size of the algorithm as Step;

[0013] S22. Use the RRT* algorithm to plan the global path of the AGV, record the coordinates of each node on the path in the AGV working space, and create a path node set J at the same time. J contains the serial number of each node on the path, the coordinates of each node, and the serial number of the parent node of each node.

[0014] Further, step S3 specifically includes the following steps:

[0015] S31. Install an industrial camera in front of the AGV, control the AGV to move in the three-dimensional working space with obstacles, and obtain a large number of images during the movement of the AGV;

[0016] S32. Process the images obtained by the industrial camera, mark the range of the obstacles and the position of the obstacles relative to the AGV on the images, and classify the obstacles according to the relative position;

[0017] S33. Input the marked image data into the YOLOv5 model for training to obtain a model for accurately identifying the position of obstacles;

[0018] S34. During the movement of the AGV, the industrial camera acquires images of the surrounding environment at a fixed sampling frequency and inputs them into the trained model for recognition to obtain the accurate position and distribution of the obstacles around the AGV during the movement.

[0019] More specifically, the classification of obstacles in step S32 is specifically:

[0020] According to the position of the obstacles relative to the AGV, the obstacles are divided into 4 types: left, right, middle, and full, which respectively represent the 4 situations where the obstacles are located on the left side, right side, directly in front of the AGV, and the obstacles cover the entire field of view of the industrial camera.

[0021] Further, step S4 specifically includes:

[0022] S41. Determine the speed value range and angular velocity value range of the AGV according to the AGV's own structure limit, motor limit, and obstacle distance limit;

[0023] S42. Set the sampling interval to group and combine the speed values and angular velocity values, and obtain the combined speed values ; then select the groups with the angular velocity exceeding the threshold to form a new set of combined speed values ;

[0024] S43. Based on the distribution of surrounding obstacles, select the speed combination; and calculate the next position parameter of the AGV according to the selected speed combination combined with the current position parameter of the AGV;

[0025] S44. Use lidar sensors installed one in front and one behind the AGV to measure the distance from the surrounding obstacles to the AGV and the speed and angular velocity information of the obstacles; calculate the movement trajectory of the obstacles after time according to the current position, speed, and angular velocity of each obstacle;

[0026] S45. Calculate the movement trajectory of the AGV after time under each group of speed combinations, and calculate the score of each group of speed combinations ; the score calculation formula is:

[0027] ;

[0028] where, is the selected speed combination, , respectively represent the selected speed and angular velocity; under this speed combination, is the azimuth evaluation function, is the obstacle distance evaluation function, is the speed evaluation function, is the global path distance evaluation function; is the weight coefficient of the evaluation function; represents the normalization process;

[0029] Select the group of speed and angular velocity with the highest score among all speed combinations as the speed combination for the AGV to move from the current position to the next position;

[0030] S46. After the AGV moves into place, repeat steps S43 - S45 until the AGV enters the range with the target point as the center and a radius of R. Then, select the speed combination with the highest score with the target point as the end point as the speed and angular velocity for the last step to achieve the AGV reaching the target point.

[0031] More specifically, step S41 is specifically as follows:

[0032] S411. Limit the speed and angular velocity range according to the structural limitations of the AGV The values of

[0033] are as follows:

[0034] Among them, and are the minimum speed and maximum speed that the AGV can reach during movement, and are the minimum angular velocity and maximum angular velocity that the AGV can reach during movement;

[0035] S412. Limit the acceleration and angular acceleration during the movement of the AGV according to the motor limitations of the AGV, and obtain the speed and angular velocity range under this limitation The values of

[0036] are as follows:

[0037] Among them, and are the current speed and angular velocity of the AGV, and are the maximum acceleration and maximum angular acceleration of the AGV, is the time required from the current position to the next position;

[0038] S413. Then consider the obstacle limitation, that is, the movement path of the AGV from deceleration to stop is less than the shortest distance from the obstacle, and obtain the speed and angular velocity range The values of

[0039] are as follows:

[0040] Among them, is the shortest distance between the movement path of the AGV and the obstacle under the current speed and angular velocity combination;

[0041] S414. According to the value range of , the value range of the final speed and angular velocity is the intersection of, that is:

[0042] ;

[0043] ;

[0044] Among them, and are the minimum and maximum values within the speed value range, respectively. and are the minimum and maximum values within the angular velocity value range, respectively.

[0045] More specifically, step S42 is specifically as follows:

[0046] Set the speed sampling interval and the angular velocity sampling interval , then the number of speed values is groups, and the number of angular velocity values is groups; expressed by the formula:

[0047] ;

[0048] Among them, and are the minimum and maximum values within the speed value range, respectively, and are the minimum and maximum values within the angular velocity value range, respectively;

[0049] In this way, a total of groups of combined speed values of speed combinations are obtained ; then select the groups where the angular velocity is greater than the threshold to form a new set of combined speed values .

[0050] More specifically, in step S43, based on the distribution of surrounding obstacles, the selection of speed combinations is specifically as follows:

[0051] Assume that at the current position, the AGV captures p images through an industrial camera, and each image has obstacles, i represents the i th image, then the total number of obstacles around the current position B is:

[0052] ;

[0053] Calculate the proportion of obstacles located directly in front of the AGV and covering the entire field of view of the industrial camera in the total number , and the calculation formula is:

[0054] ;

[0055] Among them, , respectively represent the number of obstacles located directly in front of the AGV and covering the entire field of view of the industrial camera;

[0056] If Greater than the set threshold , then a speed combination is selected from the set of resultant speed values , otherwise a speed combination is selected from the set of resultant speed values .

[0057] More specifically, each evaluation function in step S45 is specifically as follows:

[0058] Azimuth evaluation function ;

[0059] Among them, is the angle error between the movement direction of the AGV from the current position to the next position and the line connecting to the target point under the selected speed combination; the smaller the angle error, the higher the azimuth evaluation;

[0060] Obstacle distance evaluation function ;

[0061] Among them, j is the total number of obstacles detected by the lidar, is the i th minimum distance between the movement trajectory of the obstacle and the movement trajectory of the AGV;

[0062] Speed evaluation function , the greater the speed, the higher the evaluation;

[0063] Global path distance evaluation function, which represents the maximum distance between the end point of the AGV movement path optimized by the improved DWA algorithm and the global path planned by the RRT* algorithm; the formula is expressed as:

[0064] ;

[0065] Among them, is the maximum distance from the end point of the AGV movement path to the global path under the i th group of speed combinations, is the set distance threshold, k is a constant for controlling the descending speed; when the maximum distance is within the distance threshold, is set to a fixed real number N ; when the maximum distance is outside the distance threshold, the farther the distance, the lower the evaluation.

[0066] Based on the above technical solutions, the method proposed by the present invention has at least the following beneficial effects:

[0067] 1. The present invention addresses the problem that the DWA algorithm lacks global planning guidance, which can lead to long paths. A method combining the RRT* algorithm and an improved DWA algorithm is proposed to reduce the path length. The obstacle distance evaluation formula of the DWA algorithm is improved, and a global path distance evaluation formula is added to improve the rationality of the DWA algorithm.

[0068] 2. The present invention combines the information of an industrial camera and a lidar to obtain the information of obstacles around the AGV in real time, and adjusts the speed selection range of the DWA algorithm according to the obstacle information, improving the efficiency of path planning, reducing the calculation amount, and being able to better handle dynamic obstacles to ensure the safety of the AGV. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments and descriptions thereof are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0070] Figure 1 is a flowchart of the AGV path planning method proposed by the present invention;

[0071] Figure 2 is a schematic installation diagram of the industrial camera and lidar involved in the method proposed by the present invention;

[0072] Figure 3 is a flowchart of the improved DWA algorithm designed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0073] To make the above objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments. Thereby, a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects can be obtained and implemented accordingly.

[0074] Please refer to Figures 1 - 3 , which shows a specific embodiment of this embodiment. In this embodiment, the working space of the AGV is established in a two-dimensional space, and information such as the starting point, target point, and obstacles is added to the working space. The RRT* algorithm is used to plan the global path of the AGV. The YOLOv5 model is trained to process the image data input by the industrial camera so that the model can identify the position of the obstacle relative to the AGV. Combining the data output by the lidar and the YOLOv5 model, the improved DWA algorithm is used for the local path planning of the AGV. The specific improvement points are: different speed selection intervals are selected according to the position of the obstacle relative to the AGV, the evaluation of the distance to the global path is added to the speed combination evaluation, and the evaluation formula of the distance to the obstacle is improved. Finally, the AGV can safely reach the specified position.

[0075] Please refer to Figure 1 , this embodiment shows an AGV path planning method based on two-dimensional and three-dimensional data, which specifically includes the following steps:

[0076] S1. Establish the two-dimensional motion space of the AGV. In this embodiment, it is stipulated that the starting position of the AGV is the coordinate origin, and the x-axis and y-axis directions of the AGV motion space are determined; and the coordinates of the target position of the AGV and the position and size information of the known obstacles are added to the motion space;

[0077] S2. Use the RRT* algorithm to plan a global path in the working space of the AGV;

[0078] As a preferred implementation manner, step S2 specifically includes:

[0079] S21. Create a search tree directory Search for the AGV, define the starting point of the AGV as the first point of the search tree, and define the exploration step size of the algorithm as Step;

[0080] S22. Use the RRT* algorithm to plan the global path of the AGV, record the coordinates of each node on the path in the working space of the AGV, and create a path node set J at the same time; J contains the serial number of each node on the path, the coordinates of each node, and the serial number of the parent node of each node.

[0081] In this embodiment, after planning the global path, the mathematical expression of the path is calculated to calculate the distance between the end point of the speed combination in the DWA algorithm and the global path, that is, after step S22:

[0082] Segment the global path according to nodes, and find the expression of the path between every two adjacent nodes .

[0083] S3. As Figure 2 shown, install an industrial camera on the AGV, control the AGV to move in the three-dimensional working space, and collect obstacle images during the movement through the industrial camera; use the obstacle images as input to train an obstacle recognition model; use the trained model to determine the position of the obstacle relative to the AGV;

[0084] As a preferred implementation manner, step S3 specifically includes:

[0085] S31. Install an industrial camera in front of the AGV, control the AGV to move into the three-dimensional working space with obstacles, and obtain a large number of images during the movement of the AGV;

[0086] S32. Perform data processing on the images obtained by the industrial camera, mark the range of the obstacles and the position of the obstacles relative to the AGV on the images, and classify the obstacles according to the relative positions;

[0087] S33. Input the marked image data into the YOLOv5 model for training to obtain a model for accurately identifying the positions of obstacles.

[0088] S34. During the movement of the AGV, the industrial camera acquires images of the surrounding environment at a fixed sampling frequency and inputs them into the trained model for identification to obtain the accurate positions and distribution of obstacles around the AGV during movement.

[0089] S4. The AGV moves along the global path planned by the RRT* algorithm, determines the distribution of surrounding obstacles based on the trained obstacle recognition model; selects the speed combination of the AGV based on the obstacle distribution, and then combines the distance and speed information of the obstacles measured by the lidar, and performs local path optimization through the improved DWA algorithm as shown in Figure 3 to enable the AGV to reach the specified position.

[0090] As a preferred embodiment, step S4 specifically includes:

[0091] S41. Determine the speed value range and angular speed value range of the AGV according to the AGV's own structural limitations, motor limitations, and obstacle distance limitations.

[0092] More specifically, step S41 specifically includes:

[0093] S411. According to the AGV's own structural limitations, limit its speed and angular speed ranges and the values are as follows:

[0094] ;

[0095] where and are the minimum and maximum speeds that the AGV can reach during movement, and are the minimum and maximum angular speeds that the AGV can reach during movement;

[0096] S412. According to the AGV motor limitations, limit its acceleration and angular acceleration during movement, and obtain the speed and angular speed ranges under this limitation and the values are as follows:

[0097] ;

[0098] where and are the current speed and angular speed of the AGV, and is the maximum acceleration and maximum angular acceleration of the AGV, is the time required from the current position to the next position;

[0099] S413. Then consider the obstacle limitation, that is, the moving path of the AGV from deceleration to stop is less than the shortest distance from the obstacle, and obtain the speed and angular velocity ranges The value of is as follows:

[0100] ;

[0101] Among them, is the shortest distance between the moving path of the AGV and the obstacle under the current speed and angular velocity combination;

[0102] S414. According to the value range of, the value range of the final speed and angular velocity is the intersection of, that is:

[0103] ;

[0104] ;

[0105] Among them, and are the minimum and maximum values within the speed value range respectively, and are the minimum and maximum values within the angular velocity value range respectively.

[0106] S42. Set the sampling interval to group and combine the speed values and angular velocity values, and obtain the combined speed value ; Then select the groups with the angular velocity exceeding the threshold to form a new set of combined speed values ;

[0107] In this embodiment, step S42 is specifically:

[0108] Set the speed sampling interval and the angular velocity sampling interval , then there are a total of groups of speed values, and a total of groups of angular velocity values; The formula is expressed as:

[0109] ;

[0110] Among them, and are the minimum and maximum values within the speed value range respectively, and They are the minimum and maximum values within the angular velocity range respectively;

[0111] In this way, a total of sets of resultant velocity values of the velocity combinations are obtained ; Then, select the groups with angular velocity greater than the threshold to form a new set of resultant velocity values ; Such grouping allows for directly selecting velocity combinations in the set when large-angle rotation is required, thereby reducing the meaningless scoring calculation for combinations with relatively small angular velocities, that is, reducing the overall computational load.

[0112] S43. Select a velocity combination based on the distribution of surrounding obstacles; In this embodiment, assume that at the current position, the AGV captures p images through an industrial camera, and each image has obstacles, i represents the i th image, then the total number of obstacles around the current position B is:

[0113] ;

[0114] Calculate the proportion of obstacles located directly in front of the AGV and covering the entire field of view of the industrial camera in the total number , and the calculation formula is:

[0115] ;

[0116] Among them, , respectively represent the number of obstacles located directly in front of the AGV and covering the entire field of view of the industrial camera;

[0117] If is greater than the set threshold , then select a velocity combination from the set of resultant velocity values , otherwise select a velocity combination from the set of resultant velocity values ; According to the above selection rules, the selected velocity combination is combined with the current position parameters of the AGV to calculate the next position parameters of the AGV; Specifically:

[0118] Assume that the current position parameters of the AGV are , then according to the selected velocity and angular velocity, the next position parameters of the AGV can be calculated:

[0119] ;

[0120] And calculate the movement route of the AGV at this velocity and angular velocity.

[0121] S44. Use lidar sensors installed one in front and one at the rear of the AGV to measure the distances from surrounding obstacles to the AGV, as well as the speed and angular velocity information of the obstacles; calculate the movement trajectories of the obstacles after a certain time based on the current positions, speeds, and angular velocities of the respective obstacles;

[0122] S45. Calculate the movement trajectories of the AGV after a certain time for each set of speed combinations, and calculate the score for each set of speed combinations ; the score calculation formula is:

[0123] ;

[0124] where is the selected speed combination, , respectively represent the selected speed and angular velocity; under this speed combination, is the azimuth evaluation function, is the obstacle distance evaluation function, is the speed evaluation function, is the global path distance evaluation function; in this embodiment, the specific forms of each evaluation function are:

[0125] Azimuth evaluation function ;

[0126] where is the angle error between the movement direction of the AGV from the current position to the next position and the line connecting to the target point under the selected speed combination; the smaller the angle error, the higher the azimuth evaluation;

[0127] Obstacle distance evaluation function ;

[0128] where j is the total number of obstacles detected by the lidar sensor, is the minimum distance between the movement trajectory of the i th obstacle and the movement trajectory of the AGV;

[0129] During the movement of the AGV, it may encounter multiple obstacles. When the distances from these obstacles to the AGV are relatively large, the score will be relatively high; conversely, it will be small. However, to prevent the situation where although some obstacles are far away, one of them is very close and may still cause danger, the present application designs the above-mentioned obstacle distance evaluation function. When one of the obstacles is very close to the AGV, it will have a great impact on the score, significantly reducing the score and better ensuring the safety of the movement path;

[0130] Speed evaluation function , the greater the speed, the higher the evaluation;

[0131] Global path distance evaluation function, which characterizes the maximum distance between the end point of the AGV movement path optimized by the improved DWA algorithm and the global path planned by the RRT* algorithm; The formula is expressed as:

[0132] ;

[0133] Among them, is the maximum distance from the end point of the AGV movement path under the i th group of speed combinations to the global path, is the set distance threshold, k is a constant for controlling the descending speed; when the maximum distance is within the distance threshold, is set as the fixed real number N ; when the maximum distance is outside the distance threshold, the farther the distance, the lower the evaluation;

[0134] Setting the global path evaluation function can ensure that when using the DWA algorithm to optimize the local path, the score is fixed when deviating from the global path not too much, and the farther the distance exceeds this threshold, the lower the score; This can ensure that the evaluation function can play a role as long as it is within a certain range of the global path planning, remove paths with too far distances, and use other evaluation functions as the criteria for judgment within a reasonable range.

[0135] The present invention improves the rationality of the DWA algorithm for optimizing the local path by improving the obstacle distance evaluation formula of the DWA algorithm and adding the global path distance evaluation formula;

[0136] is the weight coefficient of the evaluation function; represents normalization processing;

[0137] The normalization processing formula is as follows:

[0138] ;

[0139] ;

[0140] ;

[0141] ;

[0142] Among them, m represents the total number of speed values in the speed value set, e represents the e th group of values;

[0143] Through normalization processing, it is ensured that each term in the evaluation function can function properly without the influence of other terms being eliminated due to one term being too large.

[0144] Select the set of speed and angular velocity with the highest score among all speed combinations as the speed combination for the AGV to move from the current position to the next position.

[0145] S46. After the AGV moves into place, repeat steps S43 - S45 until the AGV enters the range with a radius of R centered on the target point. Then, select the speed combination with the highest score with the target point as the end point as the speed and angular velocity for the last step to enable the AGV to reach the target point.

[0146] The AGV path planning method proposed in this embodiment is based on the data of industrial cameras and lidar. Aiming at the problem that the DWA algorithm may cause a long path, a method combining the RRT* algorithm and the improved DWA algorithm is proposed. And it combines the data of industrial cameras and lidar to avoid dynamic obstacles and changes the scoring calculation formula of the DWA algorithm to enable the AGV to select a more reasonable speed combination during the movement.

[0147] Those of ordinary skill in the art can understand that all or part of the steps in implementing the above embodiment methods can be completed by instructing relevant hardware through a program. Therefore, this application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0148] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the above embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0149] The above embodiments have introduced the present invention in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, there will be changes in the specific implementation manner and application scope according to the idea of the present invention. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An AGV path planning method based on two-dimensional and three-dimensional data, characterized in that, Specifically, it includes the following steps: S1. Establish a two-dimensional motion space for the AGV, and add the starting position, target position, and known obstacle information of the AGV in the two-dimensional motion space; S2. Use the RRT* algorithm to plan a global path in the two-dimensional working space of the AGV; S3. Install an industrial camera on the AGV, control the AGV to move in the three-dimensional working space, and collect obstacle images during the movement through the industrial camera; train an obstacle recognition model with the obstacle images as input; use the trained model to determine the position of the obstacle relative to the AGV; S4. The AGV moves according to the global path planned by the RRT* algorithm, determines the surrounding obstacle distribution based on the trained obstacle recognition model; selects the speed combination of the AGV based on the obstacle distribution, and then combines the lidar to measure the distance and speed information of the obstacle, and performs local path optimization through the improved DWA algorithm to make the AGV reach the specified position; Step S4 specifically includes: S41. Determine the speed value range and angular velocity value range of the AGV according to the AGV's own structure limit, motor limit, and obstacle distance limit; S42. Set the sampling interval to group and combine the speed values and angular velocity values, and obtain the combined speed value set V x ; Then take the groups where the angular velocity exceeds the threshold w d to form a new combined speed value set V y ; Step S42 is specifically as follows: Set the speed sampling interval C v and the angular velocity sampling interval C w , then there are n1 groups of speed values and n2 groups of angular velocity values; it is expressed by the formula as: Among them, v fmin and v fmax are respectively the minimum and maximum values within the speed value range, and w fmin and w fmax are respectively the minimum and maximum values within the angular speed value range; In this way, a set V of resultant velocity values with a total of n1·n2 groups of velocity combinations is obtained x ; then select the groups with angular velocity greater than the threshold w d to form a new set V of resultant velocity values y ; S43. Based on the distribution of surrounding obstacles, select a speed combination; and calculate the next position parameter of the AGV according to the selected speed combination and the current position parameter of the AGV; Based on the distribution of surrounding obstacles, the selection of the speed combination is specifically: Assume that at the current position, the AGV captures p images through an industrial camera, and each image contains b i obstacles. Let i represent the i-th image. Then the total number B of obstacles around the current position is: Calculate the proportion of obstacles located directly in front of the AGV and covering the entire field of view of the industrial camera in the total number The calculation formula is as follows: Among them, b 中 , b 全 respectively represent the number of obstacles located directly in front of the AGV and covering the entire field of view of the industrial camera; If is greater than the set threshold then a velocity combination is selected from the resultant velocity value set V y otherwise, a velocity combination is selected from the resultant velocity value set V x ; S44. Use lidars installed at the front and rear of the AGV respectively to measure the distance from the surrounding obstacles to the AGV and the speed and angular velocity information of the obstacles; calculate the movement trajectory of the obstacles after Δt time according to the current position, speed, and angular velocity of each obstacle; S45. Calculate the movement trajectory of the AGV after Δt time under each group of speed combinations, and calculate the score G(v, w) of each group of speed combinations; the score calculation formula is: G(v, w) = σ[μ·h(v, w) + σ[β·d o (v, w) + σ[γ·v(v, w) + σ[ε·d c (v, w)]; Among them, (v, w) is the selected speed combination, where v and w respectively represent the selected speed and angular velocity; under this speed combination, h(v, w) is the azimuth evaluation function, d o (v, w) is the obstacle distance evaluation function, v(v, w) is the speed evaluation function, d c (v, w) is the global path distance evaluation function; μ, β, γ, and ε are the weight coefficients of the evaluation functions; σ represents normalization processing; the specific forms of each evaluation function are as follows: Azimuth evaluation function h(v, w) = π - Δθ; Where, Δθ is the angle error between the movement direction of the AGV from the current position to the next position and the line connecting to the target point under the selected (v, w) speed combination; the smaller the angle error, the higher the azimuth evaluation; Obstacle distance evaluation function where j is the total number of obstacles detected by the lidar, and d i is the minimum distance between the movement trajectory of the i-th obstacle and the movement trajectory of the AGV; Speed evaluation function v(v, w) = |v|, the greater the speed, the higher the evaluation; Global path distance evaluation function, which represents the maximum distance between the end point of the AGV movement path optimized by the improved DWA algorithm and the global path planned by the RRT* algorithm; the formula is expressed as: Among them, d ci is the maximum distance from the end point of the AGV moving path to the global path under the i-th group of speed combinations, and d c0 is the set distance threshold, and k is a constant for controlling the descent speed; when the maximum distance d ci is within the distance threshold, set d c (v, w) as the fixed real number N; when the maximum distance d ci is outside the distance threshold, the farther the distance, the lower the evaluation; Select the group of speed and angular velocity with the highest score among all speed combinations as the speed combination for the AGV to move from the current position to the next position; S46. After the AGV moves into place, repeat steps S43 - S45 until when the AGV enters the range with the target point as the center and radius R, select the speed combination with the highest score as the speed and angular velocity for the last step to achieve the AGV reaching the target point.

2. The AGV path planning method based on two-dimensional and three-dimensional data according to claim 1, wherein Step S2 specifically includes the following steps: S21. Create a search tree directory Search for the AGV, define the starting point of the AGV as the first point of the search tree, and define the exploration step size of the algorithm as Step; S22. Use the RRT* algorithm to plan the global path of the AGV, record the coordinates of each node on the path in the AGV's working space, and create a path node set J at the same time. J contains the serial number of each node on the path, the coordinates of each node, and the serial number of the parent node of each node.

3. A method for AGV path planning based on two-dimensional and three-dimensional data according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31. Install an industrial camera in front of the AGV, control the AGV to move in a three-dimensional working space with obstacles, and obtain a large number of images during the movement of the AGV. S32. Process the data of the images obtained by the industrial camera, mark the range of the obstacles and the position of the obstacles relative to the AGV on the images, and classify the obstacles according to the relative position. S33. Input the marked image data into the YOLOv5 model for training to obtain a model for identifying the accurate position of the obstacles. S34. During the movement of the AGV, the industrial camera obtains images of the surrounding environment at a fixed sampling frequency f and inputs them into the trained model for identification to obtain the accurate position and distribution of the obstacles around the AGV during the movement.

4. A method for AGV path planning based on two-dimensional and three-dimensional data according to claim 3, characterized in that, The classification of the obstacles in step S32 is specifically as follows: According to the position of the obstacles relative to the AGV, the obstacles are divided into 4 types: left, right, middle, and full, which respectively represent the 4 situations where the obstacles are located on the left side, right side, directly in front of the AGV, and the obstacles cover the entire field of view of the industrial camera.

5. A method for AGV path planning based on two-dimensional and three-dimensional data according to claim 1, characterized in that, Step S41 is specifically as follows: S411. According to the structural limitations of the AGV itself, the value range of its speed and angular velocity V1 is restricted as follows: where v min and v max are the minimum and maximum speeds that the AGV can achieve during movement, and w min and w max are the minimum and maximum angular speeds that the AGV can achieve during movement; S412. According to the motor limitations of the AGV, restrict its acceleration and angular acceleration during the movement, and obtain the value range of the speed and angular velocity V2 under this restriction as follows: Among them, v c and w c are the current linear velocity and angular velocity of the AGV, a max and α max are the maximum linear acceleration and maximum angular acceleration of the AGV, and Δt is the time required from the current position to the next position; S413. Then consider the obstacle limitation, that is, the moving path of the AGV from decelerating to stopping is less than the shortest distance to the obstacle, and obtain the value range of the speed and angular velocity V3 as follows: where d is the shortest distance between the moving path of the AGV and the obstacle under the current speed and angular velocity combination. S414. According to the value ranges of V1, V2, and V3, the value range V of the final speed and angular velocity F is the intersection of V1, V2, and V3, that is: V F = V1 ∩ V2 ∩ V3; Among them, v fmin and v fmax are respectively the minimum and maximum values within the speed value range, and w fmin and w fmax are respectively the minimum and maximum values within the angular velocity value range.

Images