AGV real-time path data planning processing method and system based on particle swarm optimization

Abstract

The invention relates to the technical field of data processing, and discloses an AGV real-time path data planning processing method and system based on a particle swarm algorithm. The method comprises the steps of collecting environmental space connectivity, constraint density and dynamic element indexes to calculate topology complexity, adopting two-dimensional search when the complexity is smaller than 2.0, adopting high-dimensional search when the complexity is larger than or equal to 4.0, constructing particle structures of corresponding dimensions, distributing fitness weight coefficients, and updating particle positions and speeds. And when the environmental complexity change exceeds 0.5, switching the dimension and reconstructing a particle swarm, extracting path coordinates from the global optimal particle, verifying a safe distance and space-time conflicts, and outputting path data. The technical problem that a search strategy cannot be dynamically adjusted according to environment complexity due to dimension solidification in an existing AGV path planning technology is solved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for processing AGV real-time path data planning based on a particle swarm algorithm. Background Art

[0002] Existing AGV path planning technologies primarily utilize methods such as the A* algorithm, the Dijkstra algorithm, and basic particle swarm optimization (PSO). These techniques create an environmental map and set optimization objectives to plan the AGV's path from its starting point to its destination. Traditional PSO algorithms employ a fixed search space dimension for AGV path planning, typically using a two-dimensional or three-dimensional coordinate system to represent the AGV's location. Through an iterative optimization process, they search for the optimal path that meets basic requirements such as path length and safety.

[0003] However, existing technologies suffer from significant dimensional adaptability issues. Traditional methods are unable to dynamically adjust the dimensional configuration of the search space based on changes in environmental complexity. This results in wasted computational resources when using high-dimensional searches in simple environments, while low-dimensional searches in complex environments fail to fully account for important factors such as time coordination and device status. Furthermore, existing fitness assessment methods employ a fixed weight distribution strategy, failing to adaptively adjust the importance of various evaluation metrics based on the characteristics of different dimensional searches, thus impacting the accuracy and efficiency of path planning.

[0004] More critically, when the AGV working environment changes dynamically, existing technologies lack an effective dimension switching mechanism and are unable to perceive changes in the environment's topological complexity in real time during algorithm operation and adjust the search strategy accordingly. This limitation of dimensional solidification makes it impossible for the algorithm to maintain continuous optimization performance in complex and changing industrial environments. In particular, in scenarios where multi-dimensional factors such as multi-AGV coordination, equipment status synchronization, and energy consumption optimization need to be considered, traditional methods find it difficult to provide comprehensive and efficient path planning solutions. Summary of the Invention

[0005] The present application provides an AGV real-time path data planning and processing method and system based on a particle swarm algorithm, which is used to solve the technical problem in existing AGV path planning technology that the dimension is fixed and the search strategy cannot be dynamically adjusted according to the complexity of the environment.

[0006] In the first aspect, the present application provides an AGV real-time path data planning and processing method based on a particle swarm algorithm, the AGV real-time path data planning and processing method based on a particle swarm algorithm comprising: collecting the spatial connectivity index, constraint density index and dynamic element index of the AGV working environment, calculating the environment topology complexity value, determining a two-dimensional search when the complexity is less than 2.0, and determining a high-dimensional search when the complexity is greater than or equal to 4.0; constructing a particle structure containing a position vector and a velocity vector based on the determined search dimension, wherein the particles contain coordinate information in a two-dimensional search, and the particles add a time dimension and a device status dimension in a high-dimensional search; allocating a fitness function weight coefficient according to the search dimension, wherein the path length weight is 0.4 in a two-dimensional search, and the weights of the six evaluation indicators are all 1 / 6 in a high-dimensional search, and calculating the fitness value of each particle; updating the particle position and velocity vector, extracting the global optimal position and the individual optimal position, triggering dimension switching when the environment complexity change exceeds 0.5, and reconstructing the particle swarm structure; extracting a path coordinate sequence from the global optimal particle, verifying that the minimum distance between the path and the obstacle is greater than the safety margin, detecting spatiotemporal occupancy conflicts, and outputting path data containing coordinates and timestamps.

[0007] In a second aspect, the present application provides an AGV real-time path data planning and processing system based on a particle swarm algorithm, the AGV real-time path data planning and processing system based on a particle swarm algorithm comprising:

[0008] The acquisition module is used to collect the spatial connectivity index, constraint density index, and dynamic element index of the AGV working environment, calculate the topological complexity value of the environment, and determine a two-dimensional search when the complexity is less than 2.0, and determine a high-dimensional search when the complexity is greater than or equal to 4.0;

[0009] The search module is used to construct a particle structure containing position vectors and velocity vectors based on a determined search dimension. In two-dimensional search, the particles contain coordinate information. In high-dimensional search, the particles add time dimensions and device status dimensions.

[0010] The allocation module is used to allocate fitness function weight coefficients according to the search dimension. The path length weight is 0.4 in two-dimensional search, and the weights of the six evaluation indicators are all 1 / 6 in high-dimensional search. The fitness value of each particle is calculated;

[0011] The switching module is used to update the particle position and velocity vector, extract the global optimal position and the individual optimal position, trigger the dimension switching when the environmental complexity changes by more than 0.5, and reconstruct the particle swarm structure;

[0012] The output module is used to extract the path coordinate sequence from the global optimal particle, verify that the minimum distance between the path and the obstacle is greater than the safety margin, detect space-time occupancy conflicts, and output path data containing coordinates and timestamps.

[0013] In a third aspect, an AGV real-time path data planning and processing device based on a particle swarm algorithm is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the AGV real-time path data planning and processing device based on the particle swarm algorithm executes the above-mentioned AGV real-time path data planning and processing method based on the particle swarm algorithm.

[0014] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, enables the computer to execute the above-mentioned AGV real-time path data planning and processing method based on the particle swarm algorithm.

[0015] The technical solution provided in this application significantly improves the efficiency and accuracy of AGV path planning through a dynamic evaluation of environmental topological complexity and a dimensionality-adaptive adjustment mechanism. Specifically, the environmental topological complexity value is calculated by collecting spatial connectivity index, constraint density index, and dynamic element index. The search dimension is dynamically determined based on a complexity threshold (a two-dimensional search is used for values less than 2.0, and a high-dimensional search is used for values greater than or equal to 4.0). This effectively avoids the computational resource waste of traditional fixed-dimensional methods in simple environments and the insufficient search capability in complex ones. The adaptive particle structure constructed based on the determined search dimension can be flexibly adjusted according to environmental requirements. In two-dimensional searches, particles only contain basic coordinate information to ensure computational efficiency. In high-dimensional searches, particles add time and device status dimensions to address complex coordination requirements. This dimensionality-adaptive mechanism enables the algorithm to maintain optimal performance in environments of varying complexity. The fitness function weight coefficients assigned according to the search dimension further optimize the accuracy of path evaluation. In two-dimensional searches, the path length is weighted 0.4 to prioritize distance. In high-dimensional searches, the six evaluation indicators are weighted 1 / 6 to achieve multi-objective balance, ensuring that the evaluation focus in different search modes matches the environmental characteristics.

[0016] In the specific application of AGV real-time path planning, the particle swarm algorithm of this invention achieves continuous optimization in dynamic environments through a dimension switching and particle swarm structure reconstruction mechanism triggered when the environmental complexity changes by more than 0.5. This effectively solves the efficiency problem of traditional methods that require algorithm restarts when the environment changes. A multi-level verification mechanism that extracts the path coordinate sequence from the global optimal particle and performs safety distance verification and spatiotemporal conflict detection ensures the security and executable nature of the output path data. Especially in complex industrial environments with multiple AGVs working together, the spatiotemporal conflict detection function significantly reduces the risk of equipment collisions and the probability of operation interruptions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 This is a schematic diagram of an embodiment of an AGV real-time path data planning and processing method based on a particle swarm algorithm in an embodiment of the present application;

[0019] Figure 2 This is a schematic diagram of an embodiment of an AGV real-time path data planning and processing system based on a particle swarm algorithm in an embodiment of the present application;

[0020] Figure 3 It is a schematic block diagram of the structure of an AGV real-time path data planning and processing device based on a particle swarm algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The embodiments of the present application provide a method and system for real-time path data planning and processing of AGVs based on a particle swarm algorithm. The terms "first," "second," "third," "fourth," and the like (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the data used in this way can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products, or devices.

[0022] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the AGV real-time path data planning and processing method based on the particle swarm algorithm includes:

[0023] Step S101: Collect the spatial connectivity index, constraint density index, and dynamic element index of the AGV working environment, calculate the environment topology complexity value, and determine a two-dimensional search when the complexity is less than 2.0, and determine a high-dimensional search when the complexity is greater than or equal to 4.0;

[0024] Step S102: construct a particle structure including a position vector and a velocity vector based on the determined search dimension. In a two-dimensional search, the particles include coordinate information. In a high-dimensional search, the particles add a time dimension and a device state dimension.

[0025] Step S103: assign fitness function weight coefficients according to the search dimension. For two-dimensional search, the path length weight is 0.4. For high-dimensional search, the weights of the six evaluation indicators are all 1 / 6. Calculate the fitness value of each particle.

[0026] Step S104: Update the particle position and velocity vector, extract the global optimal position and the individual optimal position, trigger dimension switching when the environmental complexity changes by more than 0.5, and reconstruct the particle swarm structure;

[0027] Step S105: extract the path coordinate sequence from the global optimal particle, verify that the minimum distance between the path and the obstacle is greater than the safety margin, detect space-time occupancy conflicts, and output path data including coordinates and timestamps.

[0028] It is understandable that the execution subject of this application can be an AGV real-time path data planning and processing system based on a particle swarm algorithm, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking a server as the execution subject as an example.

[0029] Specifically, after the lidar scan obtains the environmental spatial data, the algorithm calculates the number of connections between the traversable path nodes and divides it by the total number of nodes to obtain the spatial connectivity index. At the same time, it calculates the constraint density index by counting the ratio of the number of obstacles in the unit space to the total area of the environment, and monitors the changes in the movement frequency of mobile obstacles to obtain the dynamic element index. The three indices are weighted and summed with weights of 0.4, 0.3, and 0.3 to calculate the environmental topological complexity value. When the value is less than 2.0, a two-dimensional search mode is determined, and when the value is greater than or equal to 4.0, a high-dimensional search mode is determined, which solves the problem that traditional algorithms cannot adjust the search dimension according to the complexity of the environment.

[0030] An adaptive particle structure is constructed based on a determined search dimension. To address the limitation of fixed particle dimensions in existing technologies, each particle contains a position vector with horizontal and vertical coordinates and a corresponding velocity vector in two-dimensional search. In high-dimensional search, the height coordinate, timestamp, device coordination status, and energy optimization status dimensions are added to the two-dimensional structure to form a six-dimensional particle structure. The total number of particle swarms is dynamically adjusted according to the number of dimensions. By improving Latin hypercube sampling, the uniform distribution of particles in each dimensional space is ensured. Each particle is assigned a dimension identifier and a topological adaptability weight, which solves the problem that existing technologies cannot adapt to environments of different complexities.

[0031] The problem of inaccurate fitness evaluation in the existing technology is solved through a dimension-aware weight allocation mechanism. In two-dimensional search, the path length weight is set to 0.4, the safety weight is 0.3, and the smoothness weight is 0.3. The time efficiency, energy consumption and equipment coordination weights are all 0. In high-dimensional search, the weights of the six evaluation indicators are all set to 1 / 6. The algorithm substitutes the particle path data into the path length calculation function to calculate the combined value of the Euclidean distance and the corner penalty to obtain the path length evaluation value. The inverse of the minimum distance between the path point and the obstacle is used as the safety evaluation value, and the smoothness evaluation value is calculated by the curvature change rate. The fitness value of each particle is obtained by weighted summing of each evaluation value according to the weight configuration. The global optimal particle and the individual optimal position are obtained by sorting and comparison.

[0032] A cross-dimensional information inheritance mechanism is adopted to solve the problem of optimization experience loss when switching dimensions in the existing technology. The particle velocity vector is updated according to the global optimal particle and the individual optimal position. The current velocity is multiplied by the inertia weight to obtain the inertia velocity component, the difference between the individual optimal position and the current position is multiplied by the learning factor one to obtain the individual learning velocity component, and the difference between the global optimal position and the current position is multiplied by the learning factor two to obtain the global learning velocity component. The three component vectors are summed and then boundary constraints are processed to obtain a new velocity vector, which is added to the current position to obtain the updated particle position. When the change in the environment topology complexity exceeds 0.5, the dimension switch is triggered, the particle historical optimization information is extracted, and the particle swarm structure is reconstructed according to the new dimension requirements, which solves the problem of slow response speed of the algorithm in a dynamic environment in the existing technology.

[0033] A multi-level verification mechanism is used to solve the problem of insufficient path safety verification in the existing technology. The path node coordinates are extracted from the global optimal position and smoothed interpolation is performed to obtain a continuous path coordinate sequence. The distance between each path point and the obstacle is calculated and compared with the safety margin for safety verification. A spatiotemporal occupancy matrix is constructed to represent the spatial occupancy of the AGV at different times. Conflict detection is performed on the spatiotemporal occupied areas with other mobile objects. The verified path coordinates are combined with the timestamp to output the final path data.

[0034] Taking a smart warehouse as an example, the AGV working environment includes dense shelf areas and open aisle areas. After LiDAR scanning, 120 spatial nodes were identified, of which 80 nodes had through-connections. The spatial connectivity index was calculated as 80 divided by 120, which equals 0.67. Obstacle statistics found 45 fixed obstacles in an area of 300 square meters. The constraint density index was 45 divided by 300, which equals 0.15. Dynamic element monitoring found 3 other AGVs and 2 mobile shelves, with a movement frequency of 8 position changes per minute. The dynamic element index was 0.8, and the environmental topological complexity value was 0.4 multiplied by 0. .67 plus 0.3 multiplied by 0.15 plus 0.3 multiplied by 0.8 equals 0.553, which is less than the threshold of 2.0, so a two-dimensional search mode is adopted. 30 two-dimensional particles are initialized. Each particle contains the starting point coordinates, the target point coordinates and the corresponding velocity vector. The fitness weights are assigned to the path length 0.4, the safety 0.3, and the smoothness 0.3. After 50 iterative updates, the global optimal particle generates a path with a length of 15 meters. The distance to the nearest obstacle is 1.2 meters, which is greater than the safety margin requirement of 0.8 meters. The spatiotemporal conflict detection shows that there is no collision risk. The output contains the coordinates of 12 path points and the corresponding timestamps.

[0035] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0036] The working environment is scanned and processed by lidar and visual sensors to obtain environmental spatial data. Based on the environmental spatial data, the topological connectivity of the traversable path is calculated to obtain the spatial connectivity index.

[0037] Perform obstacle statistical analysis on the environmental spatial data to obtain obstacle distribution data, and calculate the constraint density index based on the total number of obstacles in the obstacle distribution data and the total area of the environment;

[0038] Monitor and process the movement frequency of mobile obstacles and other AGVs in the environmental spatial data to obtain dynamic element motion data, and calculate the dynamic element index based on the frequency change of the dynamic element motion data;

[0039] The spatial connectivity index, constraint density index and dynamic element index are weighted and summed to obtain the environmental topological complexity value. When the environmental topological complexity value is less than 2.0, a two-dimensional search mark is output. When the environmental topological complexity value is greater than or equal to 4.0, a high-dimensional search mark is output.

[0040] Specifically, the problem of dimensional fixation in existing technologies is solved through multi-sensor data fusion. LiDAR uses point cloud data collection within the scanning angle range. Each scanning point contains distance value and angle information. The visual sensor extracts environmental feature points through image recognition algorithms. The two sensor data are fused after coordinate transformation and time synchronization processing to generate environmental spatial data. This data contains the coordinates of all identifiable spatial nodes, obstacle boundary information, and passable area identification. The calculation of the spatial connectivity index is based on connectivity analysis in graph theory. The algorithm traverses all spatial nodes and counts the connection relationship between each node and its adjacent nodes. Connectivity is defined as the number of other nodes that a single node can directly reach. The spatial connectivity index is equal to the sum of the connectivity of all nodes divided by the total number of nodes. This index reflects the richness of path selection in the environment.

[0041] Obstacle statistical analysis and processing classifies and counts obstacle information in the environmental spatial data. The algorithm identifies the boundary coordinates of fixed obstacles such as shelves, walls, and equipment, and calculates the occupied area of each obstacle through pixel counting or geometric shape analysis. The total number of obstacles includes all identified independent obstacles. The total area of the environment is calculated from the environmental boundary coordinates to obtain the planar area of the entire working area. The constraint density index is calculated as follows:

[0042]

[0043] in represents the constraint density index, represents the total number of obstacles, represents the total area of the environment, Represents the environment height, which quantifies the density of obstacles in the environment.

[0044] Dynamic element motion data is collected through continuous time series position monitoring. The algorithm tracks the position coordinates of moving obstacles and other AGVs in real time, and records the position changes within each time interval. The movement frequency is obtained by calculating the number of changes in the position coordinates per unit time. The dynamic element index calculation formula is:

[0045]

[0046] in represents the dynamic factor index, represents the motion frequency of the i-th dynamic object, Indicates the total number of dynamic objects, Indicates the total monitoring time. This index reflects the level of dynamic changes in the environment.

[0047] The environmental topological complexity value is processed by weighted summation to combine the three indices into a single evaluation value. The weight of the spatial connectivity index is set to 0.4 to reflect the importance of connectivity in path planning, the weight of the constraint density index is set to 0.3 to reflect the influence of obstacle distribution, and the weight of the dynamic element index is set to 0.3 to consider the complexity of the dynamic environment. The weighted summation is used to calculate the final environmental topological complexity value. When the value is less than 2.0, the environmental structure is judged to be relatively simple and a two-dimensional search mark is output. When the value is greater than or equal to 4.0, the environment is judged to be highly complex and a high-dimensional search mark is output. When it is between the two thresholds, the three-dimensional search mode is selected according to the specific value size.

[0048] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0049] Performing basic construction processing on the particle structure according to the two-dimensional search identifier to obtain a two-dimensional position vector including abscissa and ordinate and a corresponding two-dimensional velocity vector, and combining the two-dimensional position vector and the two-dimensional velocity vector to form a two-dimensional particle structure;

[0050] Perform dimension expansion processing on the two-dimensional particle structure according to the high-dimensional search identifier to obtain height dimension data including height coordinates and time dimension data including timestamps, and add the height dimension data and time dimension data to the two-dimensional position vector to form a high-dimensional position vector;

[0051] Performing device state dimension supplementation processing on the high-dimensional position vector to obtain device state dimension data including device coordination state and energy consumption optimization state, and merging the device state dimension data with the high-dimensional position vector to form an extended high-dimensional position vector;

[0052] Dynamically adjust the particle swarm size according to the search dimension type to obtain the particle number adjustment parameter, and multiply the basic particle number by the particle number adjustment parameter to calculate the total number of particle swarms;

[0053] The two-dimensional particle structure or the extended high-dimensional position vector is spatially distributed by improved Latin hypercube sampling to obtain an initial particle swarm uniformly distributed in each dimension, and each particle is assigned a dimension identifier and a topological adaptability weight.

[0054] Specifically, the basic construction processing of the two-dimensional particle structure is triggered by receiving a two-dimensional search identifier. The algorithm first assigns a unique particle identifier to each particle, and then creates a two-dimensional position vector containing horizontal and vertical coordinates. The horizontal and vertical coordinate values are obtained by randomly generating within the environmental boundary range. At the same time, a corresponding two-dimensional velocity vector is created. The initial values of the horizontal and vertical components of the velocity vector are set to zero or small random values. The two-dimensional position vector and the two-dimensional velocity vector are encapsulated and combined through a data structure to form a two-dimensional particle structure. This structure contains the spatial position information and motion trend information of the particle, laying the foundation for the subsequent path search. Data framework. After receiving the high-dimensional search identifier, the dimensionality expansion processing performs dimensionality enhancement on the existing two-dimensional particle structure. The generation of height dimension data takes into account the three-dimensional shelf structure and multi-layer channels in the AGV working environment. The height coordinate value is discretized according to the vertical spatial distribution in the environment, usually including fixed height values such as the ground layer, the middle layer of the shelf, and the top layer of the shelf. The time dimension data reflects the timestamp information of the AGV arriving at a specific location. The timestamp is estimated and initialized by dividing the path length by the average running speed of the AGV. The height dimension data and the time dimension data are added to the end of the original two-dimensional position vector through vector splicing operations to form a four-dimensional high-dimensional position vector containing horizontal coordinates, vertical coordinates, height coordinates and timestamps. This expansion process expands the plane path planning into three-dimensional space-time path planning.

[0055] The supplementary processing of the equipment status dimension is aimed at the equipment coordination needs in complex industrial environments. The equipment coordination status data records the interaction status of the AGV and peripheral equipment, including discrete state values such as the synchronization status with the conveyor belt, the docking status with the charging pile, and the avoidance status with other AGVs. The energy consumption optimization status data records the current battery power level and power consumption mode selection of the AGV. The power level is expressed as a percentage value, and the power consumption mode includes enumeration values such as energy-saving mode, standard mode and high-speed mode. The equipment coordination status and energy consumption optimization status constitute the equipment status dimension data, which is combined with the four-dimensional high-dimensional position vector through data merging operations to form an extended high-dimensional position vector containing six-dimensional information. This vector comprehensively describes the position, time and status characteristics of the AGV in a complex environment.

[0056] The dynamic adjustment of the particle swarm size is based on the complexity difference of the search dimension type. The algorithm calculates the particle number adjustment parameter according to the number of dimensions. In two-dimensional search, the adjustment parameter is set to 1.0 to keep the basic particle number unchanged. In three-dimensional search, the adjustment parameter is set to 1.5 to increase the number of particles to cover a larger search space. In high-dimensional search, the adjustment parameter is set to 2.0 to further increase the number of particles to cope with the complexity of high-dimensional search. The basic particle number is determined according to the environment scale and computing resources, and is usually set to one-quarter to one-half of the number of environment nodes. The total number of particle swarms is calculated by multiplying the basic particle number by the particle number adjustment parameter. This dynamic adjustment mechanism balances search effect and computing efficiency. The spatial distribution processing of improved Latin hypercube sampling solves the problem of uneven particle distribution caused by traditional random initialization. The sampling algorithm divides the value range of each dimension into intervals equal to the number of particles, ensuring that each interval contains the coordinate component of exactly one particle. The algorithm generates random pairings of intervals of each dimension through permutations and combinations to form a uniform distribution of particles in multidimensional space. The dimension identifier records the current dimensional configuration information of each particle, including the number of dimensions and the data type of each dimension. The topological adaptability weight is calculated based on the distance relationship between the initial position of the particle and the complex area of the environment. Particles closer to the complex area obtain higher weight values. This weight value affects the search behavior of the particles in the subsequent particle update process.

[0057] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0058] The fitness function weight coefficients are two-dimensionally allocated according to the two-dimensional search identifier, resulting in a two-dimensional weight configuration with a path length weight coefficient of 0.4, a safety weight coefficient of 0.3, and a smoothness weight coefficient of 0.3. The time efficiency, energy consumption, and equipment coordination weight coefficients are set to 0.

[0059] According to the high-dimensional search identifier, the weight coefficients of the fitness function are processed in a high-dimensional manner, and a high-dimensional weight configuration is obtained in which the weight coefficients of the six evaluation indicators of path length, safety, smoothness, time efficiency, energy consumption and equipment coordination are all 1 / 6.

[0060] Substitute the path data of each particle in the initial particle swarm into the path length calculation function to perform distance calculation processing to obtain a path length evaluation value, substitute the path length evaluation value and the obstacle distance data into the safety evaluation function to perform safety calculation processing to obtain a safety evaluation value;

[0061] The evaluation values are weighted and summed according to the two-dimensional weight configuration or the high-dimensional weight configuration to obtain the fitness value of each particle. The fitness values are sorted and compared to obtain the global optimal particle and the individual optimal position of each particle.

[0062] Specifically, the allocation processing of the two-dimensional weight configuration triggers the reconfiguration of the weight coefficient after receiving the two-dimensional search identifier. The algorithm sets the path length weight coefficient to 0.4 to reflect the dominance of path distance in simple environments, the safety weight coefficient to 0.3 to reflect the basic requirements of AGV obstacle avoidance, and the smoothness weight coefficient to 0.3 to ensure the continuity of the path and the comfort of AGV operation. At the same time, the time efficiency, energy consumption and equipment coordination weight coefficients are all set to 0, because the influence of these complex factors is relatively small in the two-dimensional search environment. The two-dimensional weight configuration forms a weight vector containing six elements, of which the first three elements are non-zero values and the last three elements are zero values.

[0063] After receiving the high-dimensional search identifier, the allocation processing of the high-dimensional weight configuration sets the weight coefficients of the six evaluation indicators to one-sixth. This uniform distribution strategy reflects the balanced importance of various factors in complex environments. Path length, safety, smoothness, time efficiency, energy consumption, and equipment coordination each occupy the same weight ratio. The high-dimensional weight configuration also forms a six-element weight vector, but all element values are 0.1667. This uniform distribution avoids the problem of over-emphasis on a single indicator and neglecting other important factors in a complex environment.

[0064] The path length calculation function takes the path data of each particle in the initial particle swarm as input. The path data contains a series of continuous coordinate points from the starting point to the end point. The function calculates the straight-line distance between adjacent path points using the Euclidean distance formula. The cumulative sum of the distances of all adjacent points forms the total path length. At the same time, the corner change in the path is calculated, and the corner angle is multiplied by the corner penalty coefficient and added to the path length evaluation value. The path length evaluation function is expressed as:

[0065]

[0066] in represents the total path length, m represents the total number of path points, and represents the coordinates of the kth path point, represents the corner penalty coefficient, represents the direction angle of the kth path segment.

[0067] The safety assessment function takes the path length assessment value and the obstacle distance data extracted from the environmental space data as input. The algorithm traverses each point on the path, calculates the shortest distance between the point and the nearest obstacle boundary, and finds the value with the minimum distance to the obstacle among all points on the path as the path safety distance. The safety assessment value is calculated as the inverse of the path safety distance. The smaller the distance, the larger the safety assessment value, indicating a higher degree of danger. The safety assessment function is expressed as:

[0068]

[0069] in represents the safety assessment value, It represents the distance between the kth path point and the nearest obstacle. When the minimum distance is less than the AGV radius, the function value is set to infinity, indicating that the path is infeasible.

[0070] The weighted summation process numerically synthesizes each evaluation value according to the current weight configuration. The algorithm multiplies the path length evaluation value by the corresponding weight coefficient, the safety evaluation value by the safety weight coefficient, the smoothness evaluation value calculated by the path curvature change rate and then multiplied by the smoothness weight coefficient, the time efficiency evaluation value calculated by the total path time and then multiplied by the time efficiency weight coefficient, the energy consumption evaluation value calculated according to the AGV power consumption model and then multiplied by the energy consumption weight coefficient, and the device coordination evaluation value calculated by the number of device conflicts and then multiplied by the device coordination weight coefficient. The fitness function is expressed as:

[0071]

[0072] in represents the particle fitness value, arrive Represents the weight coefficients of the six evaluation indicators, represents the smoothness evaluation value, represents the time efficiency evaluation value, Indicates the energy consumption assessment value, Represents the device coordination evaluation value. The smaller the fitness value, the better the path quality.

[0073] The sorting and comparison process arranges the fitness values of all particles in ascending order according to their numerical values. The particle with the smallest fitness value is identified as the current global optimal particle. The algorithm also maintains the historical optimal record of each particle and compares the current fitness value with the particle's historical optimal fitness value. If the current value is smaller, the individual optimal position of the particle is updated to the current position. Otherwise, the original individual optimal position remains unchanged. The determination of the global optimal particle and the individual optimal position of each particle lays the data foundation for subsequent particle position updates.

[0074] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0075] According to the global optimal particle and the individual optimal position, the velocity vector of each particle in the initial particle group is updated and calculated to obtain a new velocity vector. Based on the new velocity vector, the position vector of each particle is updated to obtain the updated particle position.

[0076] Substitute the updated particle position into the fitness function to recalculate the fitness to obtain the updated fitness value. Perform a global comparison on the updated fitness value to obtain the global optimal position and individual optimal position of the current iteration.

[0077] Continuously monitor the environment topology complexity value to obtain the complexity change. When the complexity change exceeds 0.5, the dimension switching judgment process is triggered to obtain the dimension switching signal.

[0078] According to the dimension switching signal, the current particle swarm is subjected to information extraction processing to obtain the particle history optimization information. Based on the particle history optimization information and the new dimension requirements, the particle swarm structure is reconstructed to obtain the particle swarm structure after dimension adjustment.

[0079] Specifically, the velocity vector update calculation process is based on the standard update rule of the particle swarm optimization algorithm. The algorithm extracts the position coordinates of the current global optimal particle and the individual optimal position coordinates of each particle as guidance information, and processes each particle in the initial particle swarm one by one. First, the inertia velocity component is calculated by multiplying the current velocity vector of the particle by the inertia weight. The inertia weight value is dynamically adjusted according to the ratio of the current number of iterations to the total number of iterations. In the early iterations, the inertia weight is larger to maintain the global search ability of the particle. In the later iterations, the inertia weight is smaller to enhance the local fine search. Then, the individual learning velocity component is calculated by multiplying the difference between the individual optimal position and the current position by the learning factor one and the random number. The learning factor one controls the intensity of the particle moving toward its own historical optimal position. Then, the global learning velocity component is calculated by multiplying the difference between the global optimal position and the current position by the learning factor two and the random number. The learning factor two controls the intensity of the particle moving toward the optimal position of the group. The three velocity components are synthesized into a new velocity vector through vector addition operation. The boundary limit processing of each dimensional component of the new velocity vector is required to ensure that it does not exceed the preset maximum velocity range. The position update process performs vector addition on the new velocity vector and the current position vector of the particle to obtain the candidate new position of the particle. The algorithm performs boundary constraint check on the candidate new position to ensure that the position coordinates do not exceed the environmental space range. At the same time, kinematic constraint verification is performed to check whether the new position meets the maximum steering angle and minimum turning radius restrictions of the AGV. The candidate position that passes the constraint verification becomes the updated particle position. The position that fails the verification needs to be corrected or regenerated. The updated particle position replaces the original position coordinates to complete the position update of a single particle. After all particles complete the position update, a new generation of particle swarm is formed.

[0080] The fitness recalculation process substitutes the updated particle position data into the multi-dimensional fitness function established previously. The algorithm calculates the path length evaluation value, safety evaluation value, smoothness evaluation value and evaluation values of other dimensions corresponding to each particle's new position one by one, and performs weighted summation of each evaluation value according to the current weight configuration to obtain the updated fitness value. The updated fitness value reflects the path quality level of the particle at the new position. The global comparison process numerically compares the updated fitness values of all particles to find the particle corresponding to the minimum value as the global optimal position of the current iteration. At the same time, the algorithm compares the updated fitness value of each particle with its historical individual optimal fitness value. When the updated value is better, the current position is set as the new individual optimal position and the corresponding fitness record is updated. Otherwise, the original individual optimal position remains unchanged. The global optimal position of the current iteration and the individual optimal position of each particle lay the data foundation for the speed update of the next iteration.

[0081] The continuous monitoring and processing of environmental topological complexity calculates the latest spatial connectivity index, constraint density index and dynamic element index by collecting environmental sensor data in real time. The algorithm calculates the difference between the latest calculated environmental topological complexity value and the complexity value at the previous moment to obtain the complexity change. The absolute value of the complexity change reflects the severity of the environmental change. When the absolute value of the change exceeds the set threshold of 0.5, it indicates that the environment has changed significantly and the search strategy needs to be adjusted. The algorithm triggers the dimension switching judgment processing to evaluate whether the current dimension configuration is still suitable for the new environmental complexity level. The judgment result generates a dimension switching signal to indicate whether dimension adjustment is required. The dimension switching signal contains information such as the target dimension type and the urgency of the switch.

[0082] The particle swarm structure reconstruction process starts the information extraction program after receiving the dimension switching signal. The algorithm traverses the current particle swarm to extract key information such as the position coordinates, velocity vector, fitness value and search history of each particle to form a particle history optimization information data set. This data set records the search experience and optimization results of the particle swarm in the current dimension. Then, the particle structure is redesigned according to the new dimension requirements. When the dimension increases, the initial value is assigned to the new dimension through interpolation and statistical analysis. When the dimension decreases, the main features of the high-dimensional information are extracted through principal component analysis and projected into the low-dimensional space. The reconstruction process integrates the effective components in the historical optimization information into the particle structure of the new dimension, forming a dimension-adjusted particle swarm structure that retains historical experience and adapts to the new environment. The reconstructed particle swarm continues to perform iterative optimization until the optimal path that meets the requirements is found.

[0083] In a specific embodiment, the process of performing the step of updating the velocity vector of each particle in the initial particle swarm according to the global optimal particle and the individual optimal position may specifically include the following steps:

[0084] The inertia weight adjustment process is performed on the historical velocity vector of the current particle to obtain the inertia velocity component. The learning factor is calculated based on the difference between the individual optimal position and the current particle position to obtain the individual learning velocity component.

[0085] The learning factor 2 is calculated based on the difference between the global optimal particle position and the current particle position to obtain the global learning speed component. The inertia weight and the learning factor are dimensionally adjusted according to the dimension identifier to obtain the dimension adjustment parameter.

[0086] Perform vector summation on the inertial velocity component, the individual learning velocity component, and the global learning velocity component to obtain a velocity update vector, and perform boundary constraint processing on the velocity update vector based on the dimension adjustment parameter to obtain a constrained velocity vector;

[0087] The constrained velocity vector and the current particle position vector are updated to obtain a new particle position. The new particle position is subjected to kinematic constraint verification to obtain an updated particle position that conforms to the AGV motion characteristics.

[0088] Specifically, the inertia weight adjustment process performs a numerical scaling operation on the historical velocity vector of the current particle. The inertia weight value is dynamically calculated based on the ratio of the current number of iterations to the total number of iterations. At the beginning of the iteration, the inertia weight is set to a larger value to maintain the global exploration ability of the particle. In the later stages of the iteration, the inertia weight is gradually reduced to enhance the local search accuracy. The algorithm multiplies each dimensional component of the particle's historical velocity vector by the inertia weight value to obtain the inertial velocity component, which retains the particle's original motion trend and search direction. The size of the inertial velocity component directly affects the particle's movement distance and exploration range in the search space. The calculation of the individual learning speed component is based on the particle's own historical optimal experience. The algorithm calculates the coordinate difference between the individual optimal position and the current particle position in each dimension. The difference vector reflects the direction and distance of the particle moving towards its own historical optimal position. The learning factor one controls the intensity of the particle's learning towards the individual optimal position. It is usually set to a value of around 1.5 to balance autonomous learning and group collaboration. The algorithm multiplies the coordinate difference vector with the learning factor one, and introduces random numbers to ensure the randomness and diversity of the search. The random numbers are evenly distributed between 0 and 1 to avoid over-determinism in particle behavior. The calculation result of the individual learning speed component is the product of the difference vector, the learning factor one and the random number. This component drives the particle to approach its own historical optimal position.

[0089] The calculation and processing of the global learning rate component uses the wisdom of the group to guide the movement of particles. The algorithm extracts the position coordinates of the current global optimal particle as a representative of the group experience, and calculates the coordinate difference vector between the global optimal position and the current particle position. The difference vector points to the best area discovered by the group. The learning factor two controls the initiative of the particle to move toward the global optimal position. The value is usually set to 2.0 to reflect the importance of group collaboration. The algorithm multiplies the global coordinate difference vector with the learning factor two, and also introduces an independent random number to maintain the uncertainty of the search. The global learning rate component is the product of the global difference vector, the learning factor two and the random number. This component guides the particles to concentrate in the optimal area of the group.

[0090] The dimensionality adjustment process modifies the inertia weight and learning factor according to the particle's dimension identifier. The dimension identifier records the particle's current dimensional configuration information, including the number of dimensions and the data type of each dimension. The algorithm calculates the dimension adjustment parameter based on the number of dimensions. The adjustment parameter of two-dimensional particles is set to 1.0 to maintain the standard parameter value. The adjustment parameter of three-dimensional particles is set to 0.9 to slightly reduce the search intensity. The adjustment parameter of high-dimensional particles is set to 0.8 to further reduce the search step size to prevent search divergence. The dimension adjustment parameter is multiplied with the inertia weight, learning factor one, and learning factor two respectively to obtain the correction parameter adapted to the current dimension. The correction parameter ensures that the search behavior of particles in different dimensions maintains reasonable convergence characteristics.

[0091] The vector summation process combines the three velocity components into a unified velocity update vector through vector addition. The algorithm adds the inertial velocity component, individual learning velocity component, and global learning velocity component dimension by dimension. The addition operation follows the vector operation rules to ensure that the velocity information of each dimension is correctly merged. The velocity update vector contains the comprehensive movement trend of the particle in each dimension. This vector comprehensively considers the three factors of the particle's historical inertia, individual experience, and group wisdom. The boundary constraint processing numerically limits the various components of the velocity update vector. The algorithm checks whether the velocity component of each dimension exceeds the preset maximum velocity threshold. The component exceeding the threshold is truncated to the maximum allowable value. The constrained velocity vector ensures that the particle movement is not too violent, causing the search to get out of control.

[0092] The position update process performs vector addition on the constrained velocity vector and the current particle position vector. The algorithm adds each dimensional component of the velocity vector to the corresponding component of the position vector to obtain the new particle position. The new position coordinates reflect the next position of the particle in the search space. The kinematic constraint verification process checks whether the new particle position meets the physical motion restrictions of the AGV. The verification content includes the maximum steering angle restriction, the minimum turning radius restriction and the maximum acceleration restriction. The algorithm calculates the steering angle between adjacent position points to ensure that it does not exceed the mechanical steering capability of the AGV, calculates the path curvature to ensure that the turning radius meets the kinematic characteristics of the AGV, and calculates the speed change rate to ensure that the acceleration is within a reasonable range. The new position that passes the verification becomes the updated particle position, and the position that fails the verification is corrected or recalculated until the constraint conditions are met.

[0093] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0094] Extracting the path node coordinate sequence from the global optimal position of the current iteration to obtain the original path coordinate data, and performing smoothing interpolation processing on the original path coordinate data to obtain a continuous path coordinate sequence;

[0095] Calculate the distance between each path point and the obstacle in the obstacle distribution data based on the continuous path coordinate sequence to obtain path safety distance data. Compare and verify the path safety distance data with the preset safety margin to obtain a safety verification result.

[0096] According to the continuous path coordinate sequence, a spatiotemporal occupancy matrix is constructed to obtain the spatiotemporal occupancy area of the AGV. The spatiotemporal occupancy area of the AGV is subjected to conflict detection processing with the spatiotemporal occupancy area of other mobile objects to obtain the spatiotemporal conflict detection result.

[0097] The continuous path coordinate sequence is combined with the corresponding timestamp data to obtain the spatiotemporal path data. The spatiotemporal path data is finally verified based on the security verification results and spatiotemporal conflict detection results to obtain the path data containing coordinates and timestamps.

[0098] Specifically, the path node coordinate sequence extraction process reads the particle position information from the global optimal position data structure of the current iteration. The global optimal position contains the set of optimized coordinate points of the particle in the search space. The algorithm extracts the coordinate data of each path node according to the logical order of path planning. The original path coordinate data contains the starting point coordinates, several intermediate node coordinates and the end point coordinates. These discrete coordinate points constitute the basic path framework for AGV movement. The smoothing interpolation process uses the cubic spline interpolation algorithm to perform continuous processing on the original path coordinate data. The algorithm inserts intermediate coordinate points between adjacent path nodes. The coordinates of the interpolation points are calculated by cubic polynomial functions. The interpolation process ensures that the continuity and smoothness of the path meet the motion characteristics requirements of the AGV. The continuous path coordinate sequence contains densely distributed path point coordinates. The distance between points is set according to the control accuracy requirements of the AGV, usually one tenth to one fifth of the AGV length. The calculation and processing of path safety distance data is based on the spatial relationship analysis of the continuous path coordinate sequence and the environmental obstacle distribution data. The algorithm traverses each coordinate point on the continuous path and calculates the Euclidean distance between the point and the boundaries of all obstacles in the environment. The obstacle distribution data comes from the obstacle position and shape information obtained from the previous environmental scan. The distance calculation considers the geometric dimensions of the AGV and simplifies the AGV into a rectangle or circle for collision detection. The path safety distance data records the shortest distance between each path point and the nearest obstacle. The safety verification result is obtained by numerically comparing the path safety distance data with the preset safety margin. The preset safety margin is determined according to the size and running speed of the AGV, usually set to the maximum size of the AGV plus the speed-related braking distance. The comparison verification process checks whether the safety distance of all path points is greater than the safety margin. When the safety distance of any path point is less than the margin, the verification result is failed. When all path points meet the safety requirements, the verification result is passed.

[0099] The spatiotemporal occupancy matrix construction process extends the continuous path coordinate sequence to the time dimension to form a four-dimensional spatiotemporal data structure. The algorithm calculates the arrival time of each path point based on the AGV's running speed. The timestamps are accumulated from the starting point of the path, considering the acceleration and deceleration process of the AGV. The spatiotemporal occupancy matrix records the spatial area occupied by the AGV at a specific moment. Each matrix element contains time coordinates, spatial coordinates and occupancy status information. The AGV spatiotemporal occupancy area is defined as the spatiotemporal volume occupied by the AGV during the entire operation process. The conflict detection process analyzes the overlap between the AGV spatiotemporal occupancy area and the spatiotemporal occupancy areas of other moving objects. Other moving objects include other AGVs, forklifts, personnel and other entities with moving trajectories. The algorithm checks whether there is an overlapping part of the same time and space in the two spatiotemporal areas. Overlap detection is achieved through the intersection operation of spatiotemporal coordinates. The spatiotemporal conflict detection results record all detected conflict times, conflict locations and conflict object information.

[0100] The spatiotemporal path data combination processing pairs and associates the continuous path coordinate sequence with the corresponding timestamp data. The algorithm assigns a corresponding time tag to each path coordinate point to form a spatiotemporal path data structure containing spatial coordinates and time information. The spatiotemporal path data fully describes the motion trajectory and time arrangement of the AGV. The final verification processing comprehensively considers the safety verification results and the spatiotemporal conflict detection results to conduct a comprehensive evaluation of the spatiotemporal path data. When the safety verification result is passed and the spatiotemporal conflict detection result shows no conflict, the spatiotemporal path data passes the final verification and becomes a valid path planning result. When a problem occurs in any verification link, the algorithm triggers the path correction program or re-planning program. The final output path data contains a fully verified coordinate sequence and corresponding timestamp. The path data format contains four data fields: the horizontal coordinate, the vertical coordinate, the height coordinate and the timestamp of each path point.

[0101] The above describes the AGV real-time path data planning and processing method based on the particle swarm algorithm in the embodiment of the present application. The following describes the AGV real-time path data planning and processing system based on the particle swarm algorithm in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the AGV real-time path data planning and processing system based on the particle swarm algorithm includes:

[0102] The acquisition module is used to collect the spatial connectivity index, constraint density index, and dynamic element index of the AGV working environment, calculate the topological complexity value of the environment, and determine a two-dimensional search when the complexity is less than 2.0, and determine a high-dimensional search when the complexity is greater than or equal to 4.0;

[0103] The search module is used to construct a particle structure containing position vectors and velocity vectors based on a determined search dimension. In two-dimensional search, the particles contain coordinate information. In high-dimensional search, the particles add time dimensions and device status dimensions.

[0104] The allocation module is used to allocate fitness function weight coefficients according to the search dimension. The path length weight is 0.4 in two-dimensional search, and the weights of the six evaluation indicators are all 1 / 6 in high-dimensional search. The fitness value of each particle is calculated;

[0105] The switching module is used to update the particle position and velocity vector, extract the global optimal position and the individual optimal position, trigger the dimension switching when the environmental complexity changes by more than 0.5, and reconstruct the particle swarm structure;

[0106] The output module is used to extract the path coordinate sequence from the global optimal particle, verify that the minimum distance between the path and the obstacle is greater than the safety margin, detect space-time occupancy conflicts, and output path data containing coordinates and timestamps.

[0107] above Figure 2The AGV real-time path data planning and processing system based on the particle swarm algorithm in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The AGV real-time path data planning and processing device based on the particle swarm algorithm in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0108] Reference Figure 3 In the embodiment of the present invention, a device for planning and processing AGV real-time path data based on particle swarm algorithm is also provided. The device for planning and processing AGV real-time path data based on particle swarm algorithm can be a server, and its internal structure can be as follows: Figure 3 As shown. The AGV real-time path data planning and processing device based on the particle swarm algorithm includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the computer-designed processor is used to provide computing and control capabilities. The memory of the AGV real-time path data planning and processing device based on the particle swarm algorithm includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the AGV real-time path data planning and processing device based on the particle swarm algorithm is used to store the corresponding data in this embodiment. The network interface of the AGV real-time path data planning and processing device based on the particle swarm algorithm is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0109] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the AGV real-time path data planning and processing device based on the particle swarm algorithm to which the solution of the present invention is applied.

[0110] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions. When the instructions are executed on a computer, the computer executes the steps of the AGV real-time path data planning and processing method based on the particle swarm algorithm.

[0111] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0112] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling an AGV real-time path data planning and processing device based on a particle swarm algorithm (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0113] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A real-time path data planning and processing method for AGV based on particle swarm optimization, characterized in that: The method comprises: The spatial connectivity index, constraint density index, and dynamic element index of the AGV working environment are collected, and the topological complexity value of the environment is calculated. When the complexity is less than 2.0, a two-dimensional search is determined, and when the complexity is greater than or equal to 4.0, a high-dimensional search is determined; Based on the determined search dimension, a particle structure containing position vectors and velocity vectors is constructed. In two-dimensional search, the particles contain coordinate information, while in high-dimensional search, the particles add time dimensions and device status dimensions. Assign fitness function weight coefficients according to the search dimension. The path length weight is 0.4 for two-dimensional search and 1 / 6 for high-dimensional search. Calculate the fitness value of each particle. Update the particle position and velocity vector, extract the global optimal position and individual optimal position, trigger dimension switching when the environmental complexity changes by more than 0.5, and reconstruct the particle swarm structure; Extract the path coordinate sequence from the global optimal particle, verify that the minimum distance between the path and the obstacle is greater than the safety margin, detect space-time occupancy conflicts, and output path data containing coordinates and timestamps.

2. The AGV real-time path data planning and processing method based on particle swarm optimization according to claim 1 is characterized in that: The method collects the spatial connectivity index, constraint density index, and dynamic element index of the AGV working environment, calculates the topological complexity value of the environment, determines a two-dimensional search when the complexity is less than 2.0, and determines a high-dimensional search when the complexity is greater than or equal to 4.0, including: Scanning and processing the working environment through laser radar and visual sensors to obtain environmental spatial data, and calculating the topological connectivity of the traversable paths based on the environmental spatial data to obtain a spatial connectivity index; Performing obstacle statistical analysis on the environmental spatial data to obtain obstacle distribution data, and calculating a constraint density index based on the total number of obstacles in the obstacle distribution data and the total area of the environment; Performing motion frequency monitoring processing on mobile obstacles and other AGVs in the environmental spatial data to obtain dynamic element motion data, and calculating a dynamic element index based on frequency changes of the dynamic element motion data; The spatial connectivity index, the constraint density index and the dynamic element index are weightedly summed to obtain an environmental topology complexity value. When the environmental topology complexity value is less than 2.0, a two-dimensional search identifier is output; when the environmental topology complexity value is greater than or equal to 4.0, a high-dimensional search identifier is output.

3. The AGV real-time path data planning and processing method based on particle swarm optimization according to claim 1 is characterized in that: The particle structure including position vectors and velocity vectors is constructed based on the determined search dimension. In a two-dimensional search, the particles include coordinate information. In a high-dimensional search, the particles add time dimension and device status dimension, including: Performing basic construction processing on the particle structure according to the two-dimensional search identifier to obtain a two-dimensional position vector including a horizontal coordinate and a vertical coordinate and a corresponding two-dimensional velocity vector, and combining the two-dimensional position vector and the two-dimensional velocity vector to form a two-dimensional particle structure; Performing dimension expansion processing on the two-dimensional particle structure according to the high-dimensional search identifier to obtain height dimension data including height coordinates and time dimension data including timestamps, and adding the height dimension data and the time dimension data to the two-dimensional position vector to form a high-dimensional position vector; Performing device state dimension supplementation processing on the high-dimensional position vector to obtain device state dimension data including device coordination state and energy consumption optimization state, and merging the device state dimension data with the high-dimensional position vector to form an extended high-dimensional position vector; Dynamically adjust the particle swarm size according to the search dimension type to obtain a particle number adjustment parameter, and multiply the basic particle number by the particle number adjustment parameter to calculate the total number of the particle swarm; The two-dimensional particle structure or the extended high-dimensional position vector is spatially distributed by improved Latin hypercube sampling to obtain an initial particle group uniformly distributed in each dimension, and a dimension identifier and a topological adaptability weight are assigned to each particle.

4. The AGV real-time path data planning and processing method based on particle swarm optimization according to claim 1 is characterized in that: The fitness function weight coefficient is assigned according to the search dimension. The path length weight is 0.4 in two-dimensional search and the weights of the six evaluation indicators are 1 / 6 in high-dimensional search. The fitness value of each particle is calculated, including: The fitness function weight coefficients are two-dimensionally allocated according to the two-dimensional search identifier, resulting in a two-dimensional weight configuration with a path length weight coefficient of 0.4, a safety weight coefficient of 0.3, and a smoothness weight coefficient of 0.

3. The time efficiency, energy consumption, and equipment coordination weight coefficients are set to 0. Performing high-dimensional distribution processing on the fitness function weight coefficient according to the high-dimensional search identifier, obtaining a high-dimensional weight configuration in which the weight coefficients of the six evaluation indicators of path length, safety, smoothness, time efficiency, energy consumption and equipment coordination are all 1 / 6; Substituting the path data of each particle in the initial particle swarm into a path length calculation function to perform distance calculation processing to obtain a path length evaluation value, and substituting the path length evaluation value and obstacle distance data into a safety evaluation function to perform safety calculation processing to obtain a safety evaluation value; A weighted summation process is performed on each evaluation value according to the two-dimensional weight configuration or the high-dimensional weight configuration to obtain a fitness value of each particle, and the fitness values are sorted and compared to obtain a global optimal particle and an individual optimal position of each particle.

5. The AGV real-time path data planning and processing method based on particle swarm algorithm according to claim 4 is characterized in that: The updating of particle positions and velocity vectors, extracting the global optimal position and individual optimal position, triggering dimension switching when the environmental complexity changes by more than 0.5, and reconstructing the particle swarm structure include: Performing an update calculation process on the velocity vector of each particle in the initial particle swarm according to the global optimal particle and the individual optimal position to obtain a new velocity vector, and performing a position update process on the position vector of each particle based on the new velocity vector to obtain an updated particle position; Substituting the updated particle position into the fitness function to perform fitness recalculation processing to obtain an updated fitness value, performing a global comparison processing on the updated fitness value to obtain a global optimal position and an individual optimal position of the current iteration; Continuously monitoring the environment topology complexity value to obtain a complexity change, and triggering a dimension switching judgment process when the complexity change exceeds 0.5 to obtain a dimension switching signal; The current particle swarm is subjected to information extraction processing according to the dimension switching signal to obtain particle history optimization information, and the particle swarm structure is reconstructed based on the particle history optimization information and the new dimension requirement to obtain a particle swarm structure after dimension adjustment.

6. The AGV real-time path data planning and processing method based on particle swarm algorithm according to claim 5 is characterized in that: The updating and calculating of the velocity vector of each particle in the initial particle swarm according to the global optimal particle and the individual optimal position to obtain a new velocity vector includes: Performing inertia weight adjustment processing on the historical velocity vector of the current particle to obtain an inertia velocity component, and performing learning factor one calculation processing based on the difference between the individual optimal position and the current particle position to obtain an individual learning velocity component; Calculating a learning factor 2 based on the difference between the global optimal particle position and the current particle position to obtain a global learning speed component, and performing dimension adjustment on the inertia weight and the learning factor according to the dimension identifier to obtain a dimension adjustment parameter; Performing vector summation processing on the inertial velocity component, the individual learning velocity component, and the global learning velocity component to obtain a velocity update vector, and performing boundary constraint processing on the velocity update vector based on the dimension adjustment parameter to obtain a constrained velocity vector; The constrained velocity vector and the current particle position vector are subjected to position update processing to obtain a new particle position, and the new particle position is subjected to kinematic constraint verification processing to obtain an updated particle position that conforms to the AGV motion characteristics.

7. The AGV real-time path data planning and processing method based on particle swarm optimization according to claim 1 is characterized in that: The method extracts the path coordinate sequence from the global optimal particle, verifies that the minimum distance between the path and the obstacle is greater than the safety margin, detects space-time occupancy conflicts, and outputs path data containing coordinates and timestamps, including: Extracting a path node coordinate sequence from the global optimal position of the current iteration to obtain original path coordinate data, and performing smoothing interpolation processing on the original path coordinate data to obtain a continuous path coordinate sequence; Calculating the distance between each path point and the obstacle in the obstacle distribution data based on the continuous path coordinate sequence to obtain path safety distance data, and performing a comparison and verification process on the path safety distance data with a preset safety margin to obtain a safety verification result; Constructing a spatiotemporal occupancy matrix based on the continuous path coordinate sequence to obtain the spatiotemporal occupancy area of the AGV, performing conflict detection processing on the spatiotemporal occupancy area of the AGV and the spatiotemporal occupancy areas of other mobile objects, and obtaining a spatiotemporal conflict detection result; The continuous path coordinate sequence is combined with the corresponding timestamp data to obtain spatiotemporal path data, and the spatiotemporal path data is finally verified based on the security verification result and the spatiotemporal conflict detection result to obtain path data including coordinates and timestamps.

8. An AGV real-time path data planning and processing system based on particle swarm algorithm, characterized in that: The method for planning and processing AGV real-time path data based on the particle swarm algorithm according to any one of claims 1 to 7 is used to implement the method, wherein the system for planning and processing AGV real-time path data based on the particle swarm algorithm comprises: The acquisition module is used to collect the spatial connectivity index, constraint density index, and dynamic element index of the AGV working environment, calculate the topological complexity value of the environment, and determine a two-dimensional search when the complexity is less than 2.0, and determine a high-dimensional search when the complexity is greater than or equal to 4.0; The search module is used to construct a particle structure containing position vectors and velocity vectors based on a determined search dimension. In two-dimensional search, the particles contain coordinate information. In high-dimensional search, the particles add time dimensions and device status dimensions. The allocation module is used to allocate fitness function weight coefficients according to the search dimension. The path length weight is 0.4 in two-dimensional search, and the weights of the six evaluation indicators are all 1 / 6 in high-dimensional search. The fitness value of each particle is calculated; The switching module is used to update the particle position and velocity vector, extract the global optimal position and the individual optimal position, trigger the dimension switching when the environmental complexity changes by more than 0.5, and reconstruct the particle swarm structure; The output module is used to extract the path coordinate sequence from the global optimal particle, verify that the minimum distance between the path and the obstacle is greater than the safety margin, detect space-time occupancy conflicts, and output path data containing coordinates and timestamps.

9. An AGV real-time path data planning and processing device based on particle swarm algorithm, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the AGV real-time path data planning and processing method based on the particle swarm algorithm according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor executes the AGV real-time path data planning and processing method based on the particle swarm algorithm according to any one of claims 1 to 7.

Images