Multi-AGV path planning and obstacle avoidance method based on improved artificial bee colony algorithm

Abstract

The present invention proposes a multi-AGV path planning and obstacle avoidance method based on an improved artificial bee colony algorithm, aiming to solve the key technical challenges of multi-AGV system path planning and obstacle avoidance in modern industrial automation. The method constructs a dynamically updateable grid environment model to accurately represent the starting point, end point, static obstacles and dynamic obstacles in the working environment, and uses an improved artificial bee colony algorithm to globally optimize the path planning of multiple AGVs, while satisfying the constraints of obstacle avoidance and multi-AGV path conflicts, ensuring the efficient operation of the system and the collision-free nature of the path. In addition, the present invention also includes path tracking control and conflict detection and obstacle avoidance control to ensure that the AGV travels accurately along the planned path, and detects conflict points in the multi-AGV path in real time, and adopts different response strategies to achieve path coordination and scheduling. Finally, the present invention can effectively improve the collaborative efficiency and safety of the multi-AGV system.

Description

Technical Field

[0001] The present invention relates to the field of path planning, and in particular to a multi-AGV path planning and obstacle avoidance method based on an improved artificial bee colony algorithm. Background Art

[0002] With the rapid development of robotics technology, multi-AGV systems (Multi-Robot Systems) have been widely used in modern industrial automation. These systems have demonstrated strong problem-solving capabilities, high robustness and reliability through teamwork and interaction, and are widely used in scenarios such as smart warehousing, oil detection and search and rescue. However, multi-robot path planning and obstacle avoidance have always been key technical challenges faced by such systems. For example, in path planning, not only the path planning of each robot should be considered, but also the optimization of multiple goals, such as the shortest path length, the lowest energy consumption, the shortest task completion time, etc., while also meeting multiple constraints such as collision avoidance. In terms of obstacle avoidance, collaborative obstacle avoidance between robots is difficult, and not only does it require real-time and accurate acquisition of other robots' position, speed and other information to avoid collisions; but also when multiple robots are avoiding obstacles at the same time, mutual interference, obstacle avoidance deadlock and other problems are prone to occur. Current path planning methods are roughly divided into classical methods, graph-based algorithms and heuristic methods:

[0003] Classical methods include artificial potential field method, landmark method and unit decomposition method. These methods have certain effects in solving basic path planning problems, but they may perform poorly in complex environments due to high computational complexity. In addition, such methods are prone to fall into local optimality and cannot find the global optimal solution.

[0004] Graph-based search algorithms, such as Dijkstra's algorithm and Algorithms are suitable for modeling meshed environments. Although these algorithms can generate feasible paths in static and simple environments, the computational overhead increases significantly in dynamic or complex environments, resulting in low efficiency.

[0005] In recent years, meta-heuristic algorithms such as particle swarm optimization, genetic algorithm, ant colony optimization and artificial bee colony algorithm have received extensive attention. These methods have the ability to escape local optimality by simulating evolution or foraging behavior in nature, and they perform well in multi-objective optimization. However, there is still a trade-off between the exploration and development capabilities of these algorithms, which may lead to slow convergence or insufficient solution accuracy.

[0006] The artificial bee colony algorithm is inspired by the foraging behavior of bees. It is widely used in single-robot and multi-robot path planning because of its strong exploration ability, few parameters and easy implementation. However, the original artificial bee colony algorithm has deficiencies in development ability, which is particularly evident in multi-robot path planning that requires real-time performance and high precision. In addition, the solution update strategy of the traditional artificial bee colony algorithm is single, which may lead to low path planning efficiency in complex environments. To this end, recent studies have improved the artificial bee colony algorithm by integrating learning strategies or optimization operations of other algorithms. For example, the global optimal individual information is introduced to enhance the search directionality, or the learning strategy of the teaching-learning optimization algorithm is adopted to improve the development ability. These improvements have achieved remarkable results in balancing the exploration and development capabilities of the algorithm and improving the efficiency and safety of path planning. However, the existing methods still have deficiencies in practicality and robustness. Summary of the invention

[0007] In view of the above content, the present invention provides a multi-AGV (Automated Guided Vehicle) path planning and obstacle avoidance method based on an improved artificial bee colony algorithm.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A multi-AGV path planning and obstacle avoidance algorithm based on an improved artificial bee colony algorithm includes the following steps:

[0010] Step 1, environmental modeling, model the warehouse environment in two-dimensional space, construct a path planning space containing obstacles and free space, and calibrate the starting point and target point of each AGV. The modeling process can be in the form of gridding or directed graphs to define path nodes and path edges, while considering the geometric features of obstacles (polygons, circular boundaries). When selecting path nodes and path edges, it is necessary to detect the feasibility of the path to ensure that the path does not overlap with obstacles, thereby providing clear environmental information for subsequent path planning.

[0011] Step 2, path planning, uses the improved artificial bee colony algorithm to generate the optimal path for multiple AGVs, thereby achieving efficient navigation from the starting point to the target point. The objective function of path planning comprehensively considers the minimization of path length, the safety distance constraints between multiple AGVs, and the safety distance constraints with obstacles to ensure the optimality and feasibility of the path. In the path planning process, the path is dynamically adjusted to adapt to complex environments and operational requirements through the collaborative optimization of employed bees, observer bees, and scout bees.

[0012] Step 3, motion control, to ensure that the AGV moves accurately along the planned path, define its kinematic model, including the changing relationship between position, speed and orientation, and adjust the linear speed and angular velocity through the proportional-integral-differential (PID) controller. The motion control module will make real-time feedback adjustments to the error between the actual position and the target position of the AGV, while satisfying the constraints of speed and angular velocity, thereby achieving smooth tracking and high-precision control of the path.

[0013] Step 4, conflict resolution. For the path conflicts that may occur during the operation of multiple AGVs, conflict detection, classification and resolution strategies (speed adjustment, path replanning) are used to ensure the safety and efficiency of the system. In the case of minor conflicts, speed adjustment, path fine-tuning or local path adjustment are used to reduce the risk of collision; in the case of severe conflicts or path congestion, complex conflict problems are resolved through strategies such as path replanning, regional partitioning or time-sharing scheduling. After all conflicts are resolved, the results are output in the form of the final path set for each AGV. The final path will be used as the task allocation plan for the AGV, indicating the driving route and scheduling timing of each AGV.

[0014] In step 1, step 1 environment modeling, the specific steps are as follows:

[0015] 1-1. Construct a warehouse environment model containing obstacles and free space in two-dimensional space, calibrate the starting point and target point of each AGV, and model the environment in the form of gridding or directed graph. Define the directed graph represents the path planning space, where Represents a set of path nodes. Indicates path nodes, represents the set of path edges, Represents a slave node To Node The edge,

[0016] Assume that the set of obstacles in the storage environment is ,in Indicates obstacles. The obstacles are represented by boundary polygons or circular boundaries. If a circular boundary is used, the obstacles The boundary is defined by the center point and radius OK, the mathematical expression is:

[0017] ,

[0018] 1-2. In the process of selecting path nodes and path edges, it is necessary to check the feasibility of the path to ensure that the path does not overlap with obstacles and that the path edges are Slave Node To Node Determine whether it is in contact with an obstacle by the following conditions Collision occurs:

[0019] ,

[0020] in, is the parameterized variable of the path edge, , if this inequality holds, then the path No obstacles A collision occurred.

[0021] In step 2, the path planning of step 2 is as follows:

[0022] 2-1. In path planning, based on the improved artificial bee colony algorithm, the optimal path from the starting point to the target point is generated. The optimization goal of the objective function is defined as:

[0023] ,

[0024] represents the minimization objective of the path length, namely:

[0025] ,

[0026] in, For the The current location of the AGVs, For its target location,

[0027] Represents the safety distance constraint between multiple AGVs:

[0028] ,

[0029] in, Indicates The minimum distance between an AGV and other AGVs, is the penalty factor,

[0030] Represents the safety distance constraint from obstacles:

[0031] ,

[0032] in, Indicates The minimum distance between an AGV and the nearest obstacle, is the safety distance threshold, is the weight factor.

[0033] Step 2-1 includes the following steps:

[0034] 2-11. Hired bee stage. In the initial stage of the algorithm, the generated The population of solutions , each solution Represents the path planning scheme of AGV. The path consists of the coordinates of multiple path nodes. Indicates that, assuming the total number of steps of the path is , then The solution is expressed as:

[0035] ,

[0036] These solutions are randomly generated in the search space and meet the feasibility constraints of the path. The employed bees Perform local search to generate new candidate solutions , the search formula is as follows:

[0037] ,

[0038] in, The current solution The weight of the dimension, The global optimal solution The weight of the dimension, , are two different individual indices randomly selected from the population, A random individual No. The weight of the dimension, A random individual No. The weight of the dimension, is the random disturbance coefficient, ,

[0039] Fitness function of candidate solution The fitness function can be expressed as follows through the path planning objective function calculation:

[0040] ,

[0041] in, They represent the minimization objective of path length, the safety distance constraint between multiple AGVs, and the safety distance constraint with obstacles, respectively. is the weight coefficient, satisfying ,

[0042] 2-12. Observation bee stage: The second stage of the improved artificial bee colony algorithm is the observation bee stage. The observation bee selects the solution that needs to be further searched according to the fitness value of the solution, using a probabilistic selection mechanism.

[0043] ,

[0044] in, Indicates The fitness of a solution, is the sum of the fitness of all solutions. Solutions with high fitness are more likely to be selected.

[0045] Observe the bee's solution to the choice Further local search is performed, using a teaching-learning based search strategy, and new candidate solutions are Given by:

[0046] ,

[0047] in, is the index of an individual randomly selected from the population, It is Solution No. The weight of the dimension, is randomly selected Solution No. The weight of the dimension, The value range is The random number, when When , the solution tends to be a better solution. When , the search range of the solution is near the original position.

[0048] The observation bee evaluates the fitness of the new solution and decides whether to accept the solution through a greedy selection strategy. The selection principle is:

[0049] ,

[0050] If the fitness of the new solution Better than the original solution , then replace the original solution with the new solution,

[0051] 2-13. Scouting bee stage. The third stage of the improved artificial bee colony algorithm is the scout bee stage. exist If the solution is not updated in the iteration, it is considered that the solution has fallen into the local optimal solution. At this time, the scout bee mechanism is activated to generate a new solution for the solution that has fallen into the local optimal solution. It is randomly generated within the boundary of the search space, and its generation formula is as follows:

[0052] ,

[0053] in, and The search space is The minimum and maximum values ​​of the dimension, yes Random numbers in the range ensure that the generated solutions are randomly distributed within the boundary.

[0054] If the following conditions are met, it is determined to be a local optimal solution:

[0055] ,

[0056] in, Indicates The solution is The fitness of the generation, Indicates The solution is The fitness of the generation, is the convergence threshold of detection, if and If the absolute value of the difference is less than the convergence threshold, it means that the fitness change between the previous and next generations is ignored and it is determined to be a local optimal solution. After all conflicts are resolved, the results are output. The path output is in the form of the final path set of each AGV. The final path will be used as the task allocation plan for the AGV, indicating the driving route and scheduling timing of each AGV.

[0057] Step 3 motion control is as follows:

[0058] 3-1. The motion state of AGV can be composed of position, direction and speed, which can be expressed as , at each time step In the AGV, the motion is controlled by the following equation:

[0059]

[0060] in, is the linear speed, For the direction, is the angular velocity, is the time step.

[0061] The linear velocity and angular velocity of the AGV need to be constrained to ensure the feasibility of control. The constraints are as follows:

[0062] ,

[0063] in, , Respectively represent the minimum and maximum values ​​of the linear velocity, , Represent the minimum and maximum angular velocity respectively.

[0064] Assume that the target trajectory of AGV is , the current location is , then its tracking error is:

[0065] ,

[0066] 3-2. Based on the idea of ​​error feedback control, a proportional-integral-differential PID controller is used to adjust the control input of the AGV:

[0067] ,

[0068] in, , , To control the gain.

[0069] After receiving the control input, the motion control module receives the path point data from the path planning module and converts these path points into control signals to indicate the speed and direction of the AGV. The generation of control signals follows the optimal control principle to ensure that the AGV reaches the target point in the shortest time and avoids unnecessary pauses or path deviations.

[0070] In step 4, the details are as follows:

[0071] 4-1 Mild conflict handling: When the number of conflicting paths on the same node is less than 1, the following strategy is adopted:

[0072] When there are only 3 or less points of conflict, the AGV speed is adjusted to avoid collision with other AGVs. The speed of an AGV is , and its target speed is , when the minimum distance to other AGVs is detected Less than the set safety distance threshold When , the following speed adjustment strategy is adopted:

[0073] ,

[0074] in, is the reaction time, is the minimum distance between the current AGV and the nearest neighbor AGV, The upper limit of adjustment is set to the target speed , ensuring that the speed does not exceed the target speed even if the safety distance allows, the current speed Dynamically adjusted based on safety distance, the speed of increase depends on and The difference.

[0075] When there are more than 3 conflicting nodes, the AGV’s path is adjusted at the conflicting nodes to avoid collision with other AGVs. The goal of path adjustment is to adjust the AGV’s forward direction. Adjust to , the adjustment formula is as follows:

[0076] ,

[0077] in, To adjust the step size, is the angle between the current path direction and the conflict-free path, is a sign function indicating the direction of adjustment.

[0078] If the collision risk is small, that is, there is only one collision node, a collision may still occur. Then, the path of the AGV is slightly adjusted to keep it away from the conflict area. The fine-tuning of the path adopts the following strategy. The fine-tuning angle is , and its calculation formula is as follows:

[0079] ,

[0080] in, is the adjustment factor, is the minimum distance between the current AGV and the nearest neighbor AGV, To set the safety distance threshold, the AGV's travel direction is adjusted to:

[0081] ,

[0082] 4-2 Severe conflict handling strategies,

[0083] When the number of conflicting paths to the same node is greater than or equal to 2, the following strategy is adopted:

[0084] When it is found that the number of conflicting paths is greater than or equal to 2 and cannot be resolved by simple speed or path adjustment, the AGV adopts a path replanning strategy. The goal of replanning is to minimize the path length and avoid conflicts with the currently running AGV. The objective function of path replanning is as follows:

[0085] ,

[0086] in, and is the weight coefficient, For the The minimum distance between an AGV and other AGVs, is the coordinate of the target point after path replanning, is the target location,

[0087] The AGV operation area is divided into several sub-areas, so that the AGV can operate in its own sub-area, thereby reducing the probability of path conflicts. The goal of partitioning is to divide the area into Sub-areas , the path conflict rate of each sub-area The following conditions are met:

[0088] ,

[0089] in, is the overall conflict rate, is the conflict rate threshold of the sub-region, and the optimization goal of the partition is:

[0090] ,

[0091] During the path planning process, the number of AGVs in each time window is controlled to reduce the possibility of path conflicts. The goal of time-sharing scheduling is to minimize the number of AGVs in each time window. The following scheduling strategies are adopted:

[0092] ,

[0093] in, is the indicator function, indicating the Arrival time of AGVs Is it in the time window Inside, is the number of AGVs in the time window, and the scheduling goal is to minimize the maximum number of AGVs in all time windows:

[0094] ,

[0095] Indicates the entire task planning time range Find all possible time windows The maximum number of AGVs in the time window is the maximum value, that is, the number of AGVs in each time window is examined to find the window with the most AGVs running simultaneously. The function is to adjust the start time of AGV and reduce the maximum peak value as much as possible.

[0096] If the number of AGVs in a certain time window exceeds the preset upper limit , then use the following method to adjust the start time of AGV:

[0097] ,

[0098] in, Delay time, used to push the start time of AGV back

[0099] For each AGV path, after the obstacle avoidance adjustment, the obstacle avoidance cost of the path needs to be recalculated. It can be calculated by the following formula:

[0100] ,

[0101] in, is the number of nodes on the path, is the number of obstacles, Path Node To obstacles The distance It is a distance penalty function. When the distance between the path node and the obstacle is small, the penalty function value increases, thereby increasing the obstacle avoidance cost.

[0102] After the obstacle avoidance adjustment is completed, the optimal path is selected by comparing the fitness before and after the path optimization. If the path after the obstacle avoidance adjustment has a lower fitness value, the current path is updated to the new path.

[0103] After completing the path coordination and scheduling, the path of each AGV is finally output. The path output is in the form of a final path set for each AGV. The final path will serve as the task allocation plan for the AGV, indicating the driving route and scheduling timing of each AGV.

[0104] An AGV control system is used to implement the multi-AGV path planning and obstacle avoidance method based on the improved artificial bee colony algorithm. The control system includes a path planning module, a dynamic obstacle avoidance module and a multi-AGV collaborative scheduling module. The path planning module generates an optimal path that meets environmental constraints; the dynamic obstacle avoidance module is used to detect obstacles and conflicts between multiple AGVs during operation, and adjust the speed or path of the AGV according to the hierarchical obstacle avoidance strategy; the multi-AGV collaborative scheduling module optimizes AGV task allocation through a time-sharing scheduling mechanism, and coordinates path conflict resolution to achieve efficient collaborative operation of AGVs.

[0105] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the multi-AGV path planning and obstacle avoidance method based on the improved artificial bee colony algorithm is implemented.

[0106] A computer-readable storage medium stores computer instructions, which, when executed by a processor, implement the multi-AGV path planning and obstacle avoidance method based on an improved artificial bee colony algorithm.

[0107] Compared with the prior art, the advantages of the present invention are as follows:

[0108] The improved artificial bee colony algorithm proposed in the present invention significantly improves the efficiency and quality of multi-AGV path planning. Traditional path planning algorithms can usually only operate under a single goal or a small number of constraints, and cannot take into account the complex requirements of path length optimization and multi-AGV collaborative operation. The present invention incorporates the minimization of path length, the safety distance constraints between multiple AGVs, and the safety distance constraints with obstacles into a unified objective function framework, and realizes global search and local optimization of path planning through the multi-stage optimization mechanism of the artificial bee colony algorithm (employed bees, observation bees, and scout bees). Compared with the prior art, the path planning strategy of the present invention is more robust and flexible, and can quickly generate the optimal path that meets the constraints even in a dynamic and complex environment;

[0109] The present invention proposes a hierarchical processing method for various conflict types in terms of obstacle avoidance strategy, thereby achieving high efficiency and accuracy in dynamic obstacle avoidance. In mild conflict scenarios, the present invention can quickly respond to temporary conflict needs between multiple AGVs through speed adjustment, path adjustment, and path fine-tuning strategies. For example, when the minimum distance between AGVs is lower than the set threshold, the system will minimize the risk of conflict by adjusting the speed or path direction in real time. In severe conflict scenarios, such as severe congestion at the intersection of paths or overlap of multiple paths, the present invention effectively disperses the operation trajectory of the AGV through path replanning strategies and regional zoning operation methods, thereby avoiding overall stagnation of system operation. The introduction of hierarchical obstacle avoidance strategies enables the system of the present invention to maintain stable operating efficiency in high-density warehousing environments;

[0110] The multi-AGV collaborative design of the present invention further improves the overall performance of the system. In the conflict resolution between multiple AGVs, the present invention realizes real-time perception of environmental information and AGV status through the monitoring module, and optimizes the task allocation of AGVs and resource utilization of time windows through the time-sharing scheduling mechanism. In actual operation, the time-sharing scheduling strategy can balance the operating load of each AGV in the time domain, thereby effectively reducing the risk of path conflicts during peak time periods. At the same time, the conflict resolution module realizes accurate identification and rapid response of path conflicts by comprehensively considering the dynamic changes of the environment and the prediction of AGV behavior, which significantly improves the intelligence level of the system in complex environments;

[0111] Therefore, the multi-AGV path planning and obstacle avoidance method of the present invention is based on the improved artificial bee colony algorithm, combined with the hierarchical obstacle avoidance strategy and multi-AGV cooperation mechanism, which not only improves the optimization ability of path planning, but also greatly enhances the efficiency of dynamic obstacle avoidance and conflict resolution. Compared with the prior art, the present invention can better meet the efficient operation requirements of multiple AGVs in a dynamic environment, is suitable for a variety of practical scenarios in the fields of warehousing logistics, industrial manufacturing, etc., and has broad application prospects and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Figure 1 It is a multi-AGV path planning and obstacle avoidance algorithm process.

[0113] Figure 2 It is to improve the detailed process of artificial bee colony algorithm.

[0114] Figure 3 This is a simulation experiment diagram of multi-AGV path coordination and obstacle avoidance scheduling.

[0115] Figure 4 is a scatter plot of resolved conflicts,

[0116] Figure 5 A line chart comparison showing the overall system runtime. DETAILED DESCRIPTION

[0117] The technical solutions in the implementation of the present invention are described clearly and completely below in conjunction with the accompanying drawings in the examples of the present invention.

[0118] Combination Figure 1 ,The present invention provides a multi-AGV path planning and obstacle avoidance algorithm.

[0119] Step 1: Environment modeling. The specific steps are as follows:

[0120] 1-1. Construct a warehouse environment model containing obstacles and free space in two-dimensional space, calibrate the starting point and target point of each AGV, and model the environment in the form of gridding or directed graph. Define the directed graph represents the path planning space, where Represents a set of path nodes. Indicates path nodes, represents the set of path edges, Represents a slave node To Node The edge,

[0121] Assume that the set of obstacles in the storage environment is ,in Indicates obstacles. The obstacles are represented by boundary polygons or circular boundaries. If a circular boundary is used, the obstacles The boundary is defined by the center point and radius OK, the mathematical expression is:

[0122] ,

[0123] 1-2. In the process of selecting path nodes and path edges, it is necessary to check the feasibility of the path to ensure that the path does not overlap with obstacles and that the path edges are Slave Node To Node Determine whether it is in contact with an obstacle by the following conditions Collision occurs:

[0124] ,

[0125] in, is the parameterized variable of the path edge, , if this inequality holds, then the path No obstacles A collision occurred.

[0126] Step 2: Path planning: Generate the optimal path from the starting point to the target point based on the improved artificial bee colony algorithm, taking into account the path length, the safety distance between multiple AGVs, and the safety distance from obstacles.

[0127] 2-1. In path planning, based on the improved artificial bee colony algorithm, the optimal path from the starting point to the target point is generated. The optimization goal of the objective function is defined as:

[0128] ,

[0129] represents the minimization objective of the path length, namely:

[0130] ,

[0131] in, For the The current location of the AGVs, For its target location,

[0132] Represents the safety distance constraint between multiple AGVs:

[0133] ,

[0134] in, Indicates The minimum distance between an AGV and other AGVs, is the penalty factor,

[0135] Represents the safety distance constraint from obstacles:

[0136] ,

[0137] in, Indicates The minimum distance between an AGV and the nearest obstacle, is the safety distance threshold, is the weight factor.

[0138] Step 3: Motion control: Define the kinematic model of the AGV and use a PID controller to control the linear and angular velocity of the AGV to ensure that the AGV moves accurately along the planned path.

[0139] 3-1. The motion state of AGV can be composed of position, direction and speed, which can be expressed as , at each time step In the AGV, the motion is controlled by the following equation:

[0140]

[0141] in, is the linear speed, For the direction, is the angular velocity, is the time step.

[0142] The linear velocity and angular velocity of the AGV need to be constrained to ensure the feasibility of control. The constraints are as follows:

[0143] ,

[0144] in, , Respectively represent the minimum and maximum values ​​of the linear velocity, , Represent the minimum and maximum angular velocity respectively.

[0145] Assume that the target trajectory of AGV is , the current location is , then its tracking error is:

[0146] ,

[0147] 3-2. Based on the idea of ​​error feedback control, a proportional-integral-differential PID controller is used to adjust the control input of the AGV:

[0148] ,

[0149] in, , , To control the gain.

[0150] After receiving the control input, the motion control module receives the path point data from the path planning module and converts these path points into control signals to indicate the speed and direction of the AGV. The generation of control signals follows the optimal control principle to ensure that the AGV reaches the target point in the shortest time and avoids unnecessary pauses or path deviations.

[0151] Step 4: Conflict resolution: Path conflicts between multiple AGVs are resolved through conflict detection, classification and resolution strategies (such as speed adjustment, path replanning, etc.) to ensure safe and efficient operation of the system.

[0152] 4-1 Mild conflict handling: When the number of conflicting paths on the same node is less than 1, the following strategy is adopted:

[0153] When there are only 3 or less points of conflict, the AGV speed is adjusted to avoid collision with other AGVs. The speed of an AGV is , and its target speed is , when the minimum distance to other AGVs is detected Less than the set safety distance threshold When , the following speed adjustment strategy is adopted:

[0154] ,

[0155] in, is the reaction time, is the minimum distance between the current AGV and the nearest neighbor AGV, The upper limit of adjustment is set to the target speed , ensuring that the speed does not exceed the target speed even if the safety distance allows, the current speed Dynamically adjusted based on safety distance, the speed of increase depends on and The difference.

[0156] When there are more than 3 conflicting nodes, the AGV’s path is adjusted at the conflicting nodes to avoid collision with other AGVs. The goal of path adjustment is to adjust the AGV’s forward direction. Adjust to , the adjustment formula is as follows:

[0157] ,

[0158] in, To adjust the step size, is the angle between the current path direction and the conflict-free path, is a sign function indicating the direction of adjustment.

[0159] If the collision risk is small, that is, there is only one collision node, a collision may still occur. Then, the path of the AGV is slightly adjusted to keep it away from the conflict area. The fine-tuning of the path adopts the following strategy. The fine-tuning angle is , and its calculation formula is as follows:

[0160] ,

[0161] in, is the adjustment factor, is the minimum distance between the current AGV and the nearest neighbor AGV, To set the safety distance threshold, the AGV's travel direction is adjusted to:

[0162] ,

[0163] 4-2 Severe conflict handling strategies,

[0164] When the number of conflicting paths to the same node is greater than or equal to 2, the following strategy is adopted:

[0165] When it is found that the number of conflicting paths is greater than or equal to 2 and cannot be resolved by simple speed or path adjustment, the AGV adopts a path replanning strategy. The goal of replanning is to minimize the path length and avoid conflicts with the currently running AGV. The objective function of path replanning is as follows:

[0166] ,

[0167] in, and is the weight coefficient, For the The minimum distance between an AGV and other AGVs, is the coordinate of the target point after path replanning, is the target location,

[0168] The AGV operation area is divided into several sub-areas, so that the AGV can operate in its own sub-area, thereby reducing the probability of path conflicts. The goal of partitioning is to divide the area into Sub-areas , the path conflict rate of each sub-area The following conditions are met:

[0169] ,

[0170] in, is the overall conflict rate, is the conflict rate threshold of the sub-region, and the optimization goal of the partition is:

[0171] ,

[0172] During the path planning process, the number of AGVs in each time window is controlled to reduce the possibility of path conflicts. The goal of time-sharing scheduling is to minimize the number of AGVs in each time window. The following scheduling strategies are adopted:

[0173] ,

[0174] in, is the indicator function, indicating the Arrival time of AGVs Is it in the time window Inside, is the number of AGVs in the time window, and the scheduling goal is to minimize the maximum number of AGVs in all time windows:

[0175] ,

[0176] Indicates the entire task planning time range Find all possible time windows The maximum number of AGVs in the time window is the maximum value, that is, the number of AGVs in each time window is examined to find the window with the most AGVs running simultaneously. The function is to adjust the start time of AGV and reduce the maximum peak value as much as possible.

[0177] If the number of AGVs in a certain time window exceeds the preset upper limit , then use the following method to adjust the start time of AGV:

[0178] ,

[0179] in, It is the delay time, which is used to push back the start time of AGV.

[0180] For each AGV path, after the obstacle avoidance adjustment, the obstacle avoidance cost of the path needs to be recalculated. It can be calculated by the following formula:

[0181] ,

[0182] in, is the number of nodes on the path, is the number of obstacles, Path Node To obstacles The distance It is a distance penalty function. When the distance between the path node and the obstacle is small, the penalty function value increases, thereby increasing the obstacle avoidance cost.

[0183] After the obstacle avoidance adjustment is completed, the optimal path is selected by comparing the fitness before and after the path optimization. If the path after the obstacle avoidance adjustment has a lower fitness value, the current path is updated to the new path.

[0184] After completing the path coordination and scheduling, the path of each AGV is finally output. The path output is in the form of a final path set for each AGV. The final path will serve as the task allocation plan for the AGV, indicating the driving route and scheduling timing of each AGV.

[0185] Figure 2 The specific implementation method of the improved artificial bee colony algorithm in the present invention is described in detail. The algorithm is used in the path optimization stage of the multi-AGV path planning and obstacle avoidance method.

[0186] Step 2-1 includes the following steps:

[0187] 2-11. Hired bee stage. The first stage of the improved artificial bee colony algorithm is the hired bee stage. In the initial stage of the algorithm, a The population of solutions , each solution Represents the path planning scheme of AGV. The path consists of the coordinates of multiple path nodes. Indicates that, assuming the total number of steps of the path is , then The solution is expressed as:

[0188] ,

[0189] These solutions are randomly generated in the search space and meet the feasibility constraints of the path. The employed bees Perform local search to generate new candidate solutions , the search formula is as follows:

[0190] ,

[0191] in, The current solution The weight of the dimension, The global optimal solution The weight of the dimension, , are two different individual indices randomly selected from the population, A random individual No. The weight of the dimension, A random individual No. The weight of the dimension, is the random disturbance coefficient, ,

[0192] Fitness function of candidate solution The fitness function can be expressed as follows through the path planning objective function calculation:

[0193] ,

[0194] in, They represent the minimization objective of path length, the safety distance constraint between multiple AGVs, and the safety distance constraint with obstacles, respectively. is the weight coefficient, satisfying ,

[0195] 2-12. Observation bee stage: The second stage of the improved artificial bee colony algorithm is the observation bee stage. The observation bee selects the solution that needs to be further searched according to the fitness value of the solution, using a probabilistic selection mechanism.

[0196] ,

[0197] in, Indicates The fitness of a solution, is the sum of the fitness of all solutions. Solutions with high fitness are more likely to be selected.

[0198] Observe the bee's solution to the choice Further local search is performed, using a teaching-learning based search strategy, and new candidate solutions are Given by:

[0199] ,

[0200] in, is the index of an individual randomly selected from the population, It is Solution No. The weight of the dimension, is randomly selected Solution No. The weight of the dimension, The value range is The random number, when When , the solution tends to be a better solution. When , the search range of the solution is near the original position.

[0201] The observation bee evaluates the fitness of the new solution and decides whether to accept the solution through a greedy selection strategy. The selection principle is:

[0202] ,

[0203] If the fitness of the new solution Better than the original solution , then replace the original solution with the new solution,

[0204] 2-13. Scouting bee stage. The third stage of the improved artificial bee colony algorithm is the scout bee stage. exist If the solution is not updated in the iteration, it is considered that the solution has fallen into the local optimal solution. At this time, the scout bee mechanism is activated to generate a new solution for the solution that has fallen into the local optimal solution. It is randomly generated within the boundary of the search space, and its generation formula is as follows:

[0205] ,

[0206] in, and The search space is The minimum and maximum values ​​of the dimension, yes Random numbers in the range ensure that the generated solutions are randomly distributed within the boundary.

[0207] If the following conditions are met, it is determined to be a local optimal solution:

[0208] ,

[0209] in, Indicates The solution is The fitness of the generation, Indicates The solution is The fitness of the generation, The algorithm continuously iterates and updates until the set stop condition is met, the maximum number of iterations is reached, or the fitness value of the path optimization converges. If the stop condition is met, the current optimal path is output.

[0210] Combination Figure 3, the present invention provides a multi-AGV path coordination and obstacle avoidance scheduling simulation experiment. The background in the figure represents an environment with obstacles, the obstacle area is marked as a black grid, and the path is represented by different colors. Each path represents the motion trajectory of an AGV from the starting point to the end point, and each grid passed by the path is connected into a line. The starting point (green) and end point (red) of each AGV, as well as the direction of each path are marked in the figure. The goal of path planning is to ensure that the AGV travels smoothly in a complex environment while avoiding collisions with obstacles. In the path optimization process, an artificial bee colony algorithm is used, which continuously adjusts the path by simulating the foraging behavior of bees to find the optimal route, minimize conflicts between paths and the length of the path, and adopt appropriate strategies for retreat. The path in the figure clearly shows how a multi-AGV system can operate efficiently in a shared environment.

[0211] Combination Figure 4 , the present invention provides conflict detection results in a multi-AGV system. The scatter plot in the figure indicates the conflict points that have been resolved in the path. The conflict points are where multiple AGVs overlap at the same position or adjacent positions. Each conflict point is marked with a dot, and the color or shape of the dot may correspond to different types of conflicts. For example, when two paths conflict at a certain position, the figure will show the numbers of the two AGVs and the specific location where the conflict occurred. If multiple AGVs conflict, it will also be shown in the figure as "multi-path conflict", indicating that congestion needs to be avoided by partitioning or replanning the path. The record of conflicts helps to optimize the path planning algorithm, further improve the coordination between AGVs, and ensure that multiple AGVs can operate efficiently and safely in the same environment.

[0212] Combination Figure 5 The present invention provides a line chart of the overall system running time. Compared with the traditional method of directly using the artificial bee colony algorithm to solve the problem in a larger environment with more path conflicts, the method of the present invention still maintains better time performance. The main reason is that there is no need to re-plan the path, a series of strategies are used to fine-tune the path, and it avoids falling into the local optimal solution.

[0213] It should be noted that the above embodiments are not intended to limit the protection scope of the present invention, and equivalent changes or substitutions made on the basis of the above technical solutions all fall within the protection scope of the claims of the present invention.

Claims

1. A multi-AGV path planning and obstacle avoidance method based on an improved artificial bee colony algorithm, characterized in that: The method comprises the following steps: Step 1: Environmental modeling: construct a warehouse environment model containing obstacles and free space in two-dimensional space, calibrate the starting point and target point of the AGV, and use a grid-based modeling method to perform modeling. Step 2: Path planning: Generate the optimal path from the starting point to the target point based on the improved artificial bee colony algorithm, taking into account the path length, the safety distance between multiple AGVs, and the safety distance from obstacles. Step 3: Motion control: define the kinematic model of the AGV and use a PID controller to control the linear and angular velocity of the AGV to ensure that the AGV moves accurately along the planned path. Step 4: Conflict resolution: resolve the path conflicts between multiple AGVs through conflict detection, classification and resolution strategies to ensure the safe and efficient operation of the system. After all conflicts are resolved, the results are output. The path output is in the form of the final path set for each AGV. The final path will be used as the task allocation plan for the AGV, indicating the driving route and scheduling timing of each AGV. Among them, step 1 is environment modeling, and the specific steps are as follows: 1-1. Construct a warehouse environment model containing obstacles and free space in two-dimensional space, calibrate the starting point and target point of each AGV, and model the environment in the form of gridding or directed graph. Define a directed graph G(D, E) to represent the path planning space, where D = {d1, d2, ..., d n } represents the path node set, d i represents the i-th path node, E=(e ij |d i ,d j ∈D} represents the path edge set, e ij Represents the slave node d i To node d j The edge, Assume that the set of obstacles in the storage environment is O = {o1, o2, ..., o m }, where 0 i Represents the ith obstacle. The obstacle is represented by a boundary polygon or a circular boundary. If a circular boundary is used, the obstacle 0 i The boundary is defined by the center point and radius OK, the mathematical expression is: 1-2. In the process of selecting path nodes and path edges, it is necessary to check the feasibility of the path to ensure that the path does not overlap with obstacles and the path edges are ij From node d i (x i ,y i ) to node d j (x j ,y j ) Determine whether it is in contact with the obstacle O by the following conditions: k Collision occurs: Among them, τ is the parameterized variable of the path edge, τ∈[0,1]. If this inequality holds, it means that the path e ij No obstacles k A collision occurs; Step 2: Path planning is as follows: 2-1. In path planning, based on the improved artificial bee colony algorithm, the optimal path from the starting point to the target point is generated. The optimization goal of the objective function is defined as: F=f1+f2+f3 f1 represents the minimization objective of the path length, namely: Among them, (x i，new ,y i，new ) is the current position of the i-th AGV, (x i，T ,y i，T ) is its target location, f2 represents the safety distance constraint between multiple AGVs: Among them, D min，i represents the minimum distance between the i-th AGV and other AGVs, η is the penalty factor, f3 represents the safety distance constraint from obstacles: Among them, ObsDist min,i represents the minimum distance between the i-th AGV and the nearest obstacle, Q is the safety distance threshold, and ζ is the weight factor; Step 3 motion control is as follows: 3-1. The motion state of AGV can be composed of position, direction and speed, denoted as [x i ,y i ,θ i , v i ], in each time step Δt, the motion of the AGV is controlled by the following equation: Among them, v i is the linear velocity, θ i is the direction, ω is the angular velocity, Δt is the time step, The linear velocity and angular velocity of the AGV need to be constrained to ensure the feasibility of control. The constraints are as follows: v min ≤v i ≤v max Oh, oh min ≤ω i ≤ω max Among them, v min , v max Respectively represent the minimum and maximum values ​​of the linear velocity, ω min ,ω max Represent the minimum and maximum angular velocity, respectively. Assume that the target trajectory of AGV is (x ref ,y ref ), the current position is (x i ,y i ), then its tracking error is: e x =x ref -x i ,he y =y ref -y i 3-2. Based on the idea of ​​error feedback control, a proportional-integral-differential PID controller is used to adjust the control input of the AGV: Among them, K p , K i , K d To control the gain; Step 4 conflict resolution is as follows: 4-1 Mild conflict handling: When the number of conflicting paths on the same node is less than 1, the following strategy is adopted: When there are only 3 or less points of conflict, the speed of the AGV is adjusted to avoid collision with other AGVs. Assume that the speed of the ath AGV is v a , whose target speed is v a，target , when the minimum distance Dis to other AGVs is detected min Less than the set safety distance threshold Dis safe When , the following speed adjustment strategy is adopted: Among them, t react is the reaction time, Dis min is the minimum distance between the current AGV and the nearest neighbor AGV, The upper limit of adjustment is set to the target speed v a，target , ensuring that the speed will not exceed the target speed even if the safety distance allows, the current speed v a Dynamic adjustment based on safety distance, the speed of increase depends on Dis min and Dis safe The difference, When there are more than 3 conflicting nodes, the AGV’s path is adjusted at the conflicting nodes to avoid collision with other AGVs. The goal of path adjustment is to adjust the AGV’s forward direction θ a Adjust to The adjustment formula is as follows: Among them, α is the adjustment step size, Δθ is the angle between the current path direction and the conflict-free path, sign(·) is the sign function, indicating the direction of adjustment, If the collision risk is small, that is, there is only one collision node, a collision may still occur. Then, the path of the AGV is slightly adjusted to keep it away from the conflict area. The fine-tuning of the path adopts the following strategy. The fine-tuning angle is Δφ, and its calculation formula is as follows: Among them, γ is the adjustment coefficient, Dis min is the minimum distance between the current AGV and the nearest neighbor AGV, Dis safe To set the safety distance threshold, the AGV's travel direction is adjusted to: 4-2 Severe conflict handling strategies, When the number of conflicting paths to the same node is greater than or equal to 2, the following strategy is adopted: When it is found that the number of conflicting paths is greater than or equal to 2 and cannot be resolved by simple speed or path adjustment, the AGV adopts a path replanning strategy. The goal of replanning is to minimize the path length and avoid conflicts with the currently running AGV. The objective function of path replanning is as follows: Among them, λ1 and λ2 are weight coefficients, Dis min，i is the minimum distance between the i-th AGV and other AGVs, (x i，new ,y i，new ) is the coordinate of the target point after path replanning, (x i，T ,y i，T ) is the target position, The operation area of ​​AGV is divided into several sub-areas, so that AGV can operate in its own sub-area, thereby reducing the probability of path conflict. The goal of partitioning is to divide the area into < sub-areas R1, R2, ..., R M , the path conflict rate C of each sub-area k The following conditions are met: Among them, C total is the overall conflict rate, C threshold is the conflict rate threshold of the sub-region, and the optimization goal of the partition is: During the path planning process, the number of AGVs in each time window is controlled to reduce the possibility of path conflicts. The goal of time-sharing scheduling is to minimize the number of AGVs in each time window. The following scheduling strategies are adopted: Among them, 1(·) is the indicator function, which represents the arrival time t of the i-th AGV i Whether it is within the time window [t, t+Δt], N t is the number of AGVs in the time window, and the scheduling goal is to minimize the maximum number of AGVs in all time windows: It means to find the maximum number of AGVs in all possible time windows [t, t+Δt] within the entire task planning time range T, that is, to examine the number of AGVs in each time window and find the window with the most AGVs running simultaneously. The role of min is to adjust the startup time of AGV to minimize the maximum peak value. If the number of AGVs in a certain time window exceeds the preset upper limit N max , then use the following method to adjust the start time of AGV: Among them, Δt shift It is the delay time, which is used to push the start time of AGV backwards; Step 2-1 includes the following steps: 2-11. Hired bee stage: At the initial stage of the algorithm, a population S = {s1, s2, ..., s N }, each solution s i Represents the path planning scheme of AGV. The path consists of the coordinates of multiple path nodes (x i，j ,y i，j ) means that, assuming the total number of steps of the path is L, the i-th solution is expressed as: S i ={(x i，1 ,y i，1 ), (x i，2 ,y i，2 ),…,(x i，L ,y i，L )} These solutions are randomly generated in the search space, satisfying the feasibility constraints of the path, and the employed bees are responsible for the current solution s i Perform local search to generate new candidate solutions s′ i，j , the search formula is as follows: s′ i，j =s i，j +ρ·(g j，best -s i，j )+μ·(s k1，j -s k2，j ) Among them, s i，j is the j-th dimension component in the current solution, g j,best is the component of the jth dimension in the global optimal solution, k1 and k2 are two different individual indices randomly selected from the population, is the j-th dimension component of random individual k1, is the j-th dimension component of random individual k2, ρ, μ are random perturbation coefficients, ρ, μ∈[0,1], The fitness function f(s) of the candidate solution is calculated through the path planning objective function, and the fitness function can be expressed as: f(s)=λ1·f1+λ2·f2+λ3·f3 Among them, f1, f2, f3 represent the minimization goal of path length, the safety distance constraint between multiple AGVs, and the safety distance constraint with obstacles, respectively. λ1, λ2, λ3 are weight coefficients, satisfying λ1+λ2+λ3=1. 2-12. Observation bee stage: The observation bee selects the solution that needs to be further searched according to the fitness value of the solution, using a probability selection mechanism, and the probability calculation formula is: Among them, f(s i ) represents the fitness of the i-th solution, is the sum of the fitness of all solutions. Solutions with high fitness are more likely to be selected. Observe the solution of the bee to the choice i Further local search is performed, using a teaching-learning based search strategy, and the new candidate solution s′ i,j Given by: Among them, k is the index of an individual randomly selected from the population, and s i，j is the i-th solution s i The j-th dimension of k，j is the kth solution s chosen randomly k The j-th dimension component of , rand is a random number in the range [0, 1], when f(s k )<f(s t ), the solution tends to be a better solution. When f(s k )≥f(s i ), the search range of the solution is near the original position, The observation bee evaluates the fitness of the new solution and decides whether to accept the solution through a greedy selection strategy. The selection principle is: If the fitness of the new solution f(s′ i ) is better than the original solution f(s i ), then replace the original solution with the new one, 2-13, Scouting Bee Phase, if a solution i In T lilmit If the solution is not updated in the iteration, it is considered that the solution has fallen into the local optimal solution. At this time, the scout bee mechanism is activated to generate a new solution for the solution that has fallen into the local optimal solution. It is randomly generated within the boundary of the search space, and its generation formula is as follows: Among them, s min，j and max，j are the minimum and maximum values ​​of the search space in the jth dimension, respectively. φ is a random number in the range of [0, 1] to ensure that the generated solutions are randomly distributed within the boundary. If the following conditions are met, it is determined to be a local optimal solution: in, represents the fitness of the ith solution in the tth generation, represents the fitness of the ith solution in the t-1th generation, ò is the convergence threshold for detection, if and If the absolute value of the difference is less than the convergence threshold, it means that the fitness change between the previous and next generations is ignored and it is determined to be a local optimal solution. After all conflicts are resolved, the results are output. The path output is in the form of the final path set of each AGV. The final path will be used as the task allocation plan for the AGV, indicating the driving route and scheduling timing of each AGV.

2. An AGV control system, characterized in that: Used to implement the multi-AGV path planning and obstacle avoidance method based on the improved artificial bee colony algorithm as described in claim 1, the control system includes a path planning module, a dynamic obstacle avoidance module and a multi-AGV collaborative scheduling module, The path planning module generates the optimal path that meets environmental constraints; the dynamic obstacle avoidance module is used to detect obstacles and conflicts between multiple AGVs during operation, and adjust the speed or path of the AGV according to the hierarchical obstacle avoidance strategy; the multi-AGV collaborative scheduling module optimizes AGV task allocation through a time-sharing scheduling mechanism, and coordinates path conflict resolution to achieve efficient collaborative operation of AGVs.

3. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the multi-AGV path planning and obstacle avoidance method based on the improved artificial bee colony algorithm as described in claim 1 above is implemented.

4. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instruction is executed by the processor, the multi-AGV path planning and obstacle avoidance method based on the improved artificial bee colony algorithm as described in claim 1 is implemented.

Images