Agv path optimization method based on vector field and dynamic prediction

Abstract

The application discloses an agv path optimization method based on a vector field and dynamic prediction, relates to the technical field of agv control, and comprises the following steps: acquiring interactive point data, carrying data and a task map of an agv, and generating a task area layer according to the interactive point data and the task map, taking the task area layer as a space carrier to construct a vector field for the agv to run; acquiring dynamic obstacle data and performing trajectory analysis to generate a trajectory point set and position prediction data; constructing a heat map based on the trajectory point set and the position prediction data, comparing heat values of each cell in the heat map with a preset heat threshold value to acquire a high-heat area and output a heat mask; and generating an obstacle avoidance vector for each cell based on the heat map and the high-heat mask, integrating the obstacle avoidance vector and an optimized guide vector to generate a final guide vector and guide the agv. The method realizes dynamic prediction of obstacles in a task map of the agv and path guidance of the AGV in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of AGV control technology, and specifically to an AGV path optimization method based on vector fields and dynamic prediction. Background Technology

[0002] With the development of intelligent manufacturing and smart logistics, automated guided vehicles (AGVs), as core equipment for flexible material handling, are widely used in various operational scenarios such as warehousing, distribution, and production. One of the core capabilities of AGV systems is path planning, which provides AGVs with safe, feasible, and efficient movement routes in complex and dynamic environments, thereby ensuring the efficient completion of tasks and the stable operation of the system.

[0003] Existing AGV path planning technologies mainly rely on static environment mapping and traditional navigation algorithms, such as the A* algorithm, Dijkstra's algorithm, or topology-based path search strategies. These methods mostly take the global shortest path as the optimization objective, making it difficult to effectively handle dynamic obstacle interference and path conflicts caused by multi-task scheduling. Some studies have attempted to introduce local obstacle avoidance algorithms (such as DWA, VFH, etc.) to deal with obstacle avoidance problems in complex environments. However, such methods are often limited to single-point obstacle avoidance logic and are difficult to balance the continuity of the overall path and environmental adaptability.

[0004] In real-world scenarios, AGVs not only face the problem of frequent dynamic obstacles entering the task area, but also need to handle different requirements for path stability and safety under various cargo states. Existing methods generally fail to organically integrate the cargo attributes of AGVs, the functional zoning of the task area, and the real-time environmental state, resulting in insufficient robustness of path planning schemes when the environment becomes more dynamic, and risks such as path oscillation, local entrapment, or even task failure. At the same time, most current path guidance strategies only adjust paths based on grid costs or simplified speed weights, lacking path adjustment mechanisms based on physical semantics and prediction mechanisms. Their ability to predict the future state of dynamic targets is limited, making it difficult to achieve high-precision and high-response obstacle avoidance decision control. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an AGV path optimization method based on vector fields and dynamic prediction, comprising: Acquire the interaction point data, transport data and task map of the AGV, and generate a task area layer based on the interaction point data and task map. Use the task area layer as a spatial carrier to construct the vector field for the AGV to run. Acquire dynamic obstacle data and perform trajectory analysis to generate a set of trajectory points and location prediction data; A heat map is constructed based on the set of trajectory points and location prediction data. The heat value of each cell in the heat map is compared with the preset heat threshold to obtain high heat areas and output heat masks. Based on heatmaps and high-heat masks, obstacle avoidance vectors are generated for each cell. The obstacle avoidance vectors and optimized guidance vectors are integrated to generate the final guidance vector and guide the AGV.

[0006] Furthermore, the transport data includes AGV cargo attributes and AGV operating speed. The cargo attributes include no load, small cargo load, and large cargo load. The interaction point data includes the cargo attributes and regional constraint attributes of the interaction points.

[0007] Furthermore, the logic for generating the task area layer is as follows: The task map of AGV is divided into m×n cells, each cell is labeled D(m,n), where m is the horizontal index of the cell and n is the vertical index of the cell; Define the attributes of each cell as follows: D(m, n) = {S(m, n), A(m, n)}, where m and n are both integers greater than 0; Where S(m,n) is the occupancy attribute of the cell in the m-th row and n-th column, and A(m,n) is the region attribute vector of the cell in the m-th row and n-th column; S(m,n)∈{0,1}, when S(m,n)=0, the corresponding cell is a passable cell, when S(m,n)=1, the corresponding cell is an obstacle cell; Define the region attribute vector for the corresponding cell based on the interaction point data, A(m, n) = [a1 m,n a2 m,n ]; Among them, a1 m,n For the cargo attributes in rows m and n, a1 m,n ∈{0, 1, 2}; When a1 m,n When =0, the cells in the m-th and n-th rows cannot hold goods; When a1 m,n When =1, the corresponding cell can hold small goods; When a1 m,n When the value is 2, the corresponding cell can hold large goods; a2 m,n For the region constraint attributes of the m-th and n-th rows, a2 m,n ∈{0,1}; a2 m,n When =0, the cells in the m-th and n-th rows have no area constraints; a2 m,nWhen =1, the corresponding cell has a range constraint condition; The output is a task map defined by the attribute definitions and region attribute vectors of each cell, and is output as a task region layer.

[0008] Furthermore, the construction logic of the vector field is as follows: In the task area layer, set the task target cell G, G=(m g n g ), m g n g Let the cell in the m-th row and n-th column be labeled with the index g, where g is an integer greater than 0. Initialize the vector field value V(m) in the target cell G. g n g ), making the vector value V(m) g n g ) is the zero vector, i.e., V(m) g n g )= ; Calculate the backpropagation guiding vector based on the task target cell to obtain the guiding vector for each cell: V(m, n) = =unit(G-(m,n); Based on the attribute definitions of each cell, the guiding vector V(m,n) of the cell is modified to generate an optimized guiding vector Vx(m,n), thus completing the construction of the vector field.

[0009] Furthermore, the logic for correcting the guiding vector V(m, n) of the cell is as follows: Iterate through each cell D(m, n). If the occupancy attribute S(m, n) of a cell D(m, n) is 1, it is determined to be an obstacle cell. The guiding vector V(m, n) is then corrected to a zero vector, and the optimized guiding vector Vx(m, n) is output. ; If the occupancy attribute S(m,n) of cell D(m,n) is 0, obtain the correction coefficient β and correct the guiding vector V(m,n) of the cell. The correction formula for V(m,n) is: Vx(m,n) = β × V(m,n). The logic for obtaining the correction coefficient β is as follows: Define the cargo load attribute of the AGV as L, where L∈{0,1,2}. When L=0, the AGV is unloaded; when L=1, the AGV is loaded with small goods; and when L=2, the AGV is loaded with large goods. Then, compare the AGV cargo load attribute L with the cargo attribute a1 of the cell. m,n If the comparison is performed, and the cargo attribute L of agv is less than the cargo attribute a1, then... m,n The output is β=1; If the cargo attribute L of AGV is greater than or equal to the cargo attribute a1 m,n The output β = 1 - γ × (L - a1) m,n ), where γ is the attenuation factor, and γ∈(0,1).

[0010] Furthermore, the dynamic obstacle data includes target identifier, timestamp, location information, and speed information.

[0011] Furthermore, the logic of trajectory analysis is as follows: The identifier of the j-th dynamic obstacle target entering the mission map is denoted as ID(j), where j is an integer greater than 0; Collect the current timestamp Ti of the dynamic obstacle ID(j), where i is a constant greater than or equal to 0, and the location information Pj(Ti) = (m j (Ti), n j (Ti), velocity information Uj(Ti) = (um j (Ti), un j (Ti)); The location information of dynamic obstacle ID(j) is integrated to construct a set of trajectory points: δj={Pj(T0),Pj(T1),……,Pj(Tk)},0<k≤i; Let the current timestamp of the dynamic obstacle ID(j) be T. Predict the position state of each obstacle at time Z in the future using linear extrapolation to generate position prediction data. The prediction formula is: P^(T+f)=Pj(T)+Uj(T)×f,f∈[0,Z].

[0012] Furthermore, the logic for constructing a heatmap based on the trajectory point set and location prediction data is as follows: The position prediction data P^(T+f) of each dynamic obstacle ID(j) is plotted into the corresponding cell D(m,n) of the task area layer; The heat value HT(m,n) is generated by statistically analyzing the frequency of occurrence of trajectory points in all trajectory point combinations and the location prediction data in each cell. The formula for calculating the heat value HT(m,n) is as follows: HT(m,n)= ; In the formula, To determine whether the location prediction data P^(T+f) falls into cell D(m,n), Wt is a preset weighting factor; Obtain the heat values ​​HT(m,n) of each cell D(m,n) to construct a heatmap.

[0013] Furthermore, the logic of comparing the heat value of each cell in the heat map with the preset heat threshold to obtain high-heat areas and output the heat mask is as follows: Compare the heat value HT(m, n) of each cell with the preset heat threshold η; If HT(m,n)≥η, mark the corresponding cell as a high-heat region and output the thermal mask Lh(m,n)=1; If HT(m,n) < η, mark the corresponding cell as a normal area and output the thermal mask Lh(m,n) = 0.

[0014] Furthermore, the final logic for generating the guiding vector is as follows: Extract the optimized guiding vector Vx(m,n) from cell D(m,n) in the task area layer, and simultaneously extract the corresponding heat value HT(m,n) and heat mask Lh(m,n). The cells with thermal mask Lh(m,n)=1 retain the optimized guiding vector as the final guiding vector for output. The cell with thermal mask Lh(m,n)=0 is designated as the cell to be corrected; For the cell to be corrected, extend one cell in the task area layer to form a 3×3 region centered on the current cell D(m,n), and calculate the thermal gradient ∇HT(m,n) within the adjacent region of the current cell: ∇HT(m,n)=( , ); The avoidance vector V is calculated based on the thermal gradient. avoid Avoid vector V avoid The calculation formula is: V avoid (m,n)=- ; Among them, gradient norm =( 2 +( 2 ; If the gradient norm When the gradient is less than or equal to the preset minimum gradient, the output V is... avoid (m, n) = 0, the thermal value is stable, and there is no need for obstacle avoidance; If the gradient norm When the gradient exceeds the preset minimum gradient, the thermal value becomes unstable, requiring obstacle avoidance guidance. The original guidance vector Vx(m,n) and the obstacle avoidance vector V of the cell that needs obstacle avoidance guidance are... avoidThe vectors (m, n) are integrated to generate the final guiding vector Ve(m, n). The formula for calculating the final guiding vector Ve(m, n) is as follows: Ve(m,n)=(1-y×Vx(m,n+y×V avoid (m, n), where y is a preset optimization factor, y∈(0,1).

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention combines the cargo-carrying attributes of AGVs with the cargo-carrying attributes and regional constraint attributes of each cell in the task map to achieve multi-dimensional attribute modeling of the task area layer. It also introduces an adjustable vector correction coefficient to dynamically adapt the path guidance under different load conditions, which significantly improves the safety, rationality and environmental adaptability of the path planning results. At the same time, it clearly distinguishes the vector semantics of target cells and obstacle cells, enhances the robustness of the guidance vector field in the path control process, and effectively supports the autonomous operation and task adaptation of AGVs in complex working environments. Furthermore, this invention extracts high-heat areas by comparing thermal values ​​with thermal thresholds, effectively identifying potential collision risk areas, and calculates avoidance vectors using thermal gradients to dynamically correct the original guidance path. Finally, by fusing and optimizing the guidance vectors and obstacle avoidance vectors to generate the final guidance path, it achieves dynamic prediction of obstacles in the AGV task map and path guidance for the AGV in complex environments. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 The flowchart shows an AGV path optimization method based on vector field and dynamic prediction provided in Embodiment 1 of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1 Please see Figure 1 As shown, this embodiment discloses an agv path optimization method based on vector fields and dynamic prediction, the method including: S100: Acquire the interaction point data, transport data and task map of the AGV, and generate a task area layer based on the interaction point data and task map. Use the task area layer as a spatial carrier to construct the vector field for the AGV to run. Specifically, the transport data includes AGV cargo attributes and AGV operating speed. The cargo attributes include no load, small cargo load, and large cargo load. The interaction point data includes the cargo attributes and regional constraint attributes of the interaction points. It should be noted that the cargo-carrying attributes of the AGV are obtained through gravity sensors installed on the vehicle and image recognition equipment. Specifically, the logic for generating the task area layer is as follows: The task map of AGV is divided into m×n cells, each cell is labeled D(m,n), where m is the horizontal index of the cell and n is the vertical index of the cell; Define the attributes of each cell as follows: D(m, n) = {S(m, n), A(m, n)}, where m and n are both integers greater than 0; Where S(m,n) is the occupancy attribute of the cell in the m-th row and n-th column, and A(m,n) is the region attribute vector of the cell in the m-th row and n-th column; S(m,n)∈{0,1}, when S(m,n)=0, the corresponding cell is a passable cell, when S(m,n)=1, the corresponding cell is an obstacle cell; Define the region attribute vector for the corresponding cell based on the interaction point data, A(m, n) = [a1 m,n a2 m,n ]; Among them, a1 m,n For the cargo attributes in rows m and n, a1 m,n ∈{0, 1, 2}, when a1 m,n When a1 = 0, the cells in the m-th and n-th rows cannot hold goods; when a1 m,n When a1 = 1, the corresponding cell can hold small goods; when a1 m,n When the value is 2, the corresponding cell can hold large goods; a2 m,n For the region constraint attributes of the m-th and n-th rows, a2 m,n ∈{0,1},a2 m,n When =0, the cells in row m and row n have no area constraints; a2 m,n When =1, the corresponding cell has a range constraint condition; The output is a task map defined by the attribute definitions and region attribute vectors of each cell, and is output as a task region layer.

[0020] It should be noted that: cargo attributes and area constraint attributes are determined by the operator through classification; the map modeling can be built using ROS and RViz / Gazebo; when the cargo attribute of a cell is 2, it can accommodate cargo with cargo attribute 1, meaning that a cell that can accommodate large cargo can also accommodate small cargo; the area constraint conditions include, but are not limited to, temperature, humidity, and area placement time limits. Specifically, the construction logic of the vector field is as follows: S101: Set the task target cell G in the task area layer, G=(m g n g ), m g n g Let the cell in the m-th row and n-th column be labeled with the index g, where g is an integer greater than 0. Initialize the vector field value V(m) in the target cell G. g n g ), making the vector value V(m) g n g ) is the zero vector, i.e., V(m) g n g )= ; S102: Calculate the backpropagation guiding vector based on the task target cell to obtain the guiding vector for each cell: V(m, n) = =unit(G-(m,n); It should be noted that the calculation method can include, but is not limited to, gradient descent and distance field propagation, where V(m,n) is the unit vector pointing from the point to the target point. S103: Based on the attribute definitions of each cell, modify the guiding vector V(m,n) of the cell to generate an optimized guiding vector Vx(m,n), thus completing the construction of the vector field; Specifically, the logic for correcting the guiding vector V(m, n) of the cell is as follows: Iterate through each cell D(m, n). If the occupancy attribute S(m, n) of a cell D(m, n) is 1, it is determined to be an obstacle cell. The guiding vector V(m, n) is then corrected to a zero vector, and the optimized guiding vector Vx(m, n) is output. ; If the occupancy attribute S(m,n) of cell D(m,n) is 0, obtain the correction coefficient β and correct the guiding vector V(m,n) of the cell. The correction formula for V(m,n) is: Vx(m,n) = β × V(m,n). The logic for obtaining the correction coefficient β is as follows: Define the cargo load attribute of the AGV as L, where L∈{0,1,2}. When L=0, the AGV is unloaded; when L=1, the AGV is loaded with small goods; and when L=2, the AGV is loaded with large goods. Then, compare the AGV cargo load attribute L with the cargo attribute a1 of the cell. m,n If the comparison is performed, and the cargo attribute L of agv is less than the cargo attribute a1, then... m,n The output is β=1; If the cargo attribute L of AGV is greater than or equal to the cargo attribute a1 m,n The output β = 1 - γ × (L - a1) m,n ), where γ is the attenuation factor, γ∈(0,1); It should be noted that the attenuation factor is set by the operator through analysis of historical data, and the main setting range is 0.5 to 0.8.

[0021] By combining the cargo-carrying attributes of AGVs with the cargo-carrying attributes and regional constraint attributes of each cell in the task map, multi-dimensional attribute modeling of the task area layer is achieved. An adjustable vector correction coefficient is introduced to dynamically adapt the path guidance under different load conditions, which significantly improves the safety, rationality and environmental adaptability of the path planning results. At the same time, the vector semantics of target cells and obstacle cells are clearly distinguished, which enhances the robustness of the guidance vector field in the path control process and effectively supports the autonomous operation and task adaptation of AGVs in complex working environments.

[0022] Real-time Example 2 During AGV operation, in addition to fixed obstacles (such as walls and shelves), dynamic obstacles are encountered, such as manually driven forklifts, other AGVs, or personnel crossing the road. These obstacles are non-static and unpredictable, which can cause existing vector field paths to fail or pose a collision risk. To ensure the real-time performance and safety of AGV paths, it is necessary to detect and predict dynamic obstacles in the task area layer and correct existing guide vectors, thereby improving the dynamic adaptability of the path. To solve the above-mentioned problems, this invention proposes an AGV path optimization method based on vector fields and dynamic prediction, which further includes: S200: Acquire dynamic obstacle data and perform trajectory analysis to generate a set of trajectory points and location prediction data; Specifically, the dynamic obstacle data includes target identifier, timestamp, location information, and speed information; It should be noted that the dynamic obstacle data is acquired through collection devices deployed within the AGV mission map, including but not limited to lidar, depth cameras, infrared sensors, etc., combined with real-time target detection and recognition algorithms, and obstacle status data is acquired through periodic scanning. Specifically, the logic of the trajectory analysis is as follows: The identifier of the j-th dynamic obstacle target entering the mission map is denoted as ID(j), where j is an integer greater than 0; Collect the current timestamp Ti of the dynamic obstacle ID(j), where i is a constant greater than or equal to 0, and the location information Pj(Ti) = (m j (Ti), n j (Ti), velocity information Uj(Ti) = (um j (Ti), un j (Ti)); The location information of dynamic obstacle ID(j) is integrated to construct a set of trajectory points: δj={Pj(T0),Pj(T1),……,Pj(Tk)},0<k≤i; Let the current timestamp of the dynamic obstacle ID(j) be T. Predict the position state of each obstacle at time Z in the future using linear extrapolation to generate position prediction data. The prediction formula is: P^(T+f)=Pj(T)+Uj(T)×f,f∈[0,Z]; S300: Constructs a heat map based on trajectory point set and location prediction data, compares the heat value of each cell in the heat map with the preset heat threshold to obtain high heat areas and output heat mask; Specifically, the logic for constructing a heatmap based on a set of trajectory points and location prediction data is as follows: The position prediction data P^(T+f) of each dynamic obstacle ID(j) is plotted into the corresponding cell D(m,n) of the task area layer; The heat value HT(m,n) is generated by statistically analyzing the frequency of occurrence of trajectory points in all trajectory point combinations and the location prediction data in each cell. The formula for calculating the heat value HT(m,n) is as follows: HT(m,n)= ; In the formula, To determine whether the location prediction data P^(T+f) falls into cell D(m,n), Wt is a preset weighting factor; It should be noted that the weighting factor Wt is calculated and set by the operator using the exponential decay function in the prior art, and can be output as different weighting factors based on time changes; Obtain the heat values ​​HT(m,n) of each cell D(m,n) to construct a heatmap; Specifically, the logic of comparing the heat value of each cell in the heat map with a preset heat threshold to obtain high-heat areas and output a heat mask is as follows: Compare the heat value HT(m, n) of each cell with the preset heat threshold η; If HT(m,n)≥η, mark the corresponding cell as a high-heat region and output the thermal mask Lh(m,n)=1; If HT(m,n) < η, mark the corresponding cell as a normal area and output the thermal mask Lh(m,n) = 0.

[0023] S400: Based on heatmaps and high-heat masks, an obstacle avoidance vector is generated for each cell. The obstacle avoidance vector and the optimized guidance vector are integrated to generate the final guidance vector and guide the AGV. Specifically, the logic for generating the final guiding vector is as follows: Extract the optimized guiding vector Vx(m,n) from cell D(m,n) in the task area layer, and simultaneously extract the corresponding heat value HT(m,n) and heat mask Lh(m,n). The cells with thermal mask Lh(m,n)=1 retain the optimized guiding vector as the final guiding vector for output. The cell with thermal mask Lh(m,n)=0 is designated as the cell to be corrected; For the cell to be corrected, extend one cell in the task area layer to form a 3×3 region centered on the current cell D(m,n), and calculate the thermal gradient ∇HT(m,n) within the adjacent region of the current cell: ∇HT(m,n)=( , ); The avoidance vector V is calculated based on the thermal gradient. avoid Avoid vector V avoid The calculation formula is: V avoid (m,n)=- ; Among them, gradient norm =( 2 +( 2 ; If the gradient norm When the gradient is less than or equal to the preset minimum gradient, the output V is... avoid (m, n) = 0, the thermal value is stable, and there is no need for obstacle avoidance; If the gradient norm When the gradient exceeds the preset minimum gradient, the thermal value becomes unstable, requiring obstacle avoidance guidance. The original guidance vector Vx(m,n) and the obstacle avoidance vector V of the cell that needs obstacle avoidance guidance are... avoidThe vectors (m, n) are integrated to generate the final guiding vector Ve(m, n). The formula for calculating the final guiding vector Ve(m, n) is as follows: Ve(m,n)=(1-y×Vx(m,n+y×V avoid (m, n), where y is a preset optimization factor, y∈(0,1).

[0024] It should be noted that the optimization factor was obtained through experiments and can achieve optimal adjustment of the final guiding vector.

[0025] This embodiment extracts high-heat areas by comparing thermal values ​​with thermal thresholds, effectively identifying potential collision risk areas. It then calculates avoidance vectors using thermal gradients to dynamically correct the original guidance path. Finally, by fusing and optimizing the guidance vectors and obstacle avoidance vectors to generate the final guidance path, the AGV can achieve highly safe and continuous path navigation and task execution in complex dynamic environments, improving the intelligence and stability of path planning.

[0026] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0027] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0028] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0029] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only one AGV path optimization method based on vector fields and dynamic prediction. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0030] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0031] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0032] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0033] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AGV path optimization method based on vector fields and dynamic prediction, characterized in that, The method includes: Acquire the interaction point data, transport data, and task map of the AGV, and generate a task area layer based on the interaction point data and task map. Use the task area layer as a spatial carrier to construct the vector field for the AGV to run. Acquire dynamic obstacle data and perform trajectory analysis to generate a set of trajectory points and location prediction data; A heat map is constructed based on the set of trajectory points and location prediction data. The heat value of each cell in the heat map is compared with the preset heat threshold to obtain high heat areas and output heat masks. Based on heatmaps and high-heat masks, obstacle avoidance vectors are generated for each cell. The obstacle avoidance vectors and optimized guidance vectors are integrated to generate the final guidance vector and guide the AGV.

2. The agv path optimization method based on vector field and dynamic prediction according to claim 1, characterized in that, The transport data includes AGV cargo attributes and AGV operating speed. The cargo attributes include no load, small cargo load, and large cargo load. The interaction point data includes the cargo attributes and area constraint attributes of the interaction points.

3. The agv path optimization method based on vector field and dynamic prediction according to claim 2, characterized in that, The logic for generating the task area layer is as follows: The task map of AGV is divided into m×n cells, each cell is labeled D(m,n), where m is the horizontal index of the cell and n is the vertical index of the cell; Define the attributes of each cell as follows: D(m, n) = {S(m, n), A(m, n)}, where m and n are both integers greater than 0; Where S(m,n) is the occupancy attribute of the cell in the m-th row and n-th column, and A(m,n) is the region attribute vector of the cell in the m-th row and n-th column; S(m,n)∈{0,1}, when S(m,n)=0, the corresponding cell is a passable cell, when S(m,n)=1, the corresponding cell is an obstacle cell; Define the region attribute vector for the corresponding cell based on the interaction point data, A(m, n) = [a1 m,n a2 m,n ]; Among them, a1 m,n For the cargo attributes in rows m and n, a1 m,n ∈{0, 1, 2}; When a1 m,n When =0, the cells in the m-th and n-th rows cannot hold goods; When a1 m,n When =1, the corresponding cell can hold small goods; When a1 m,n When the value is 2, the corresponding cell can hold large goods; a2 m,n For the region constraint attributes of the m-th and n-th rows, a2 m,n ∈{0,1}; a2 m,n When =0, the cells in the m-th and n-th rows have no area constraints; a2 m,n When =1, the corresponding cell has a range constraint condition; The output is a task map defined by the attribute definitions and region attribute vectors of each cell, and is output as a task region layer.

4. The agv path optimization method based on vector field and dynamic prediction according to claim 3, characterized in that, The construction logic of the vector field is as follows: In the task area layer, set the task target cell G, G=(m g n g ), m g n g Let the cell in row m and column n be labeled with index g, where g is an integer greater than 0. Initialize the vector field value V(m) in the target cell G. g n g ), making the vector value V(m) g n g ) is the zero vector, i.e., V(m) g n g )= ; Calculate the backpropagation guiding vector based on the task target cell to obtain the guiding vector for each cell: V(m, n) = =unit(G-(m,n)); Based on the attribute definitions of each cell, the guiding vector V(m,n) of the cell is modified to generate an optimized guiding vector Vx(m,n), thus completing the construction of the vector field.

5. The agv path optimization method based on vector field and dynamic prediction according to claim 4, characterized in that, The logic for correcting the guiding vector V(m, n) of the cell is as follows: Iterate through each cell D(m, n). If the occupancy attribute S(m, n) of a cell D(m, n) is 1, it is determined to be an obstacle cell. The guiding vector V(m, n) is then corrected to a zero vector, and the optimized guiding vector Vx(m, n) is output. ; If the occupancy attribute S(m,n) of cell D(m,n) is 0, obtain the correction coefficient β and correct the guiding vector V(m,n) of the cell. The correction formula for V(m,n) is: Vx(m,n) = β × V(m,n). The logic for obtaining the correction coefficient β is as follows: Define the cargo load attribute of the AGV as L, where L∈{0,1,2}. When L=0, the AGV is unloaded; when L=1, the AGV is loaded with small goods; and when L=2, the AGV is loaded with large goods. Then, compare the AGV cargo load attribute L with the cargo attribute a1 of the cell. m,n If the comparison is performed, and the cargo attribute L of agv is less than the cargo attribute a1, then... m,n The output is β=1; If the cargo attribute L of AGV is greater than or equal to the cargo attribute a1 m,n The output β = 1 - γ × (L - a1) m,n ), where γ is the attenuation factor, and γ∈(0,1).

6. The agv path optimization method based on vector field and dynamic prediction according to claim 5, characterized in that, The dynamic obstacle data includes target identifier, timestamp, location information, and speed information.

7. The agv path optimization method based on vector field and dynamic prediction according to claim 6, characterized in that, The logic of trajectory analysis is as follows: The identifier of the j-th dynamic obstacle target entering the mission map is denoted as ID(j), where j is an integer greater than 0; Collect the current timestamp Ti of the dynamic obstacle ID(j), where i is a constant greater than or equal to 0, and the location information Pj(Ti) = (m j (Ti), n j (Ti)), velocity information Uj(Ti) = (um j (Ti), un j (Ti)); The location information of dynamic obstacle ID(j) is integrated to construct a set of trajectory points: δj={Pj(T0),Pj(T1),……,Pj(Tk)},0<k≤i; Let the current timestamp of the dynamic obstacle ID(j) be T. Predict the position state of each obstacle at time Z in the future using linear extrapolation to generate position prediction data. The prediction formula is: P^(T+f)=Pj(T)+Uj(T)×f,f∈[0,Z].

8. The agv path optimization method based on vector field and dynamic prediction according to claim 7, characterized in that, The logic for constructing a heatmap based on trajectory point sets and location prediction data is as follows: The position prediction data P^(T+f) of each dynamic obstacle ID(j) is plotted into grid cells to map to the corresponding cell D(m,n) of the task area layer; The heat value HT(m,n) is generated by statistically analyzing the frequency of occurrence of trajectory points in all trajectory point combinations and the location prediction data in each cell. The formula for calculating the heat value HT(m,n) is as follows: HT(m,n)= ; In the formula, To determine whether the location prediction data P^(T+f) falls into cell D(m,n), Wt is a preset weighting factor; Obtain the heat values ​​HT(m,n) of each cell D(m,n) to construct a heatmap.

9. The agv path optimization method based on vector field and dynamic prediction according to claim 8, characterized in that, The logic for comparing the heat values ​​of each cell in the heatmap with preset heat thresholds to obtain high-heat areas and output heat masks is as follows: Compare the heat value HT(m, n) of each cell with the preset heat threshold η; If HT(m,n)≥η, mark the corresponding cell as a high-heat region and output the thermal mask Lh(m,n)=1; If HT(m,n) < η, mark the corresponding cell as a normal area and output the thermal mask Lh(m,n) = 0.

10. The agv path optimization method based on vector field and dynamic prediction according to claim 9, characterized in that, The final logic for generating the guiding vector is as follows: Extract the optimized guiding vector Vx(m,n) from cell D(m,n) in the task area layer, and simultaneously extract the corresponding heat value HT(m,n) and heat mask Lh(m,n). The cells with thermal mask Lh(m,n)=1 retain the optimized guiding vector as the final guiding vector for output; The cell with thermal mask Lh(m,n)=0 is recorded as the cell to be corrected; For the cell to be corrected, extend one cell in the task area layer to form a 3×3 region centered on the current cell D(m,n), and calculate the thermal gradient ∇HT(m,n) within the adjacent region of the current cell: ∇HT(m,n)=( , ); The avoidance vector V is calculated based on the thermal gradient. avoid Avoid vector V avoid The calculation formula is: V avoid (m,n)=- ; Among them, gradient norm = ( 2 +( 2 ; If the gradient norm When the gradient is less than or equal to the preset minimum gradient, the output V is... avoid (m, n) = 0, the thermal value is stable, and there is no need for obstacle avoidance; If the gradient norm When the gradient exceeds the preset minimum gradient, the thermal value becomes unstable, requiring obstacle avoidance guidance. The original guidance vector Vx(m,n) and the obstacle avoidance vector V of the cell that needs obstacle avoidance guidance are... avoid The vectors (m, n) are integrated to generate the final guiding vector Ve(m, n). The formula for calculating the final guiding vector Ve(m, n) is as follows: Ve(m,n)=(1-y×Vx(m,n+y×V avoid (m, n), where y is a preset optimization factor, y∈(0,1).

Images