Robot path planning method for warehouse logistics cluster operation

Abstract

The invention discloses a robot path planning method for warehouse logistics cluster operation, and belongs to the technical field of path planning. The method comprises the following steps: constructing a two-dimensional model diagram by using a grid method; the ant colony algorithm is improved; an improved ant colony algorithm is adopted to plan an operation path for each logistics robot; a local obstacle avoidance strategy is carried out through a dynamic window algorithm, and whether obstacles exist around is detected; and the collision type is judged, and a corresponding collision avoidance strategy is adopted until all the robots safely and correctly arrive at the preset target point. According to the method, the pheromone concentration is initialized by introducing the Sigmoid function, blind search in the early stage is avoided, the self-adaptive slope parameter is introduced, and the steep degree of the Sigmoid function is dynamically adjusted according to environmental characteristics; meanwhile, the ant colony algorithm and the dynamic window algorithm are fused, parameters are dynamically adjusted according to the number of iterations, the obstacle rejection weight is introduced, the algorithm is prevented from falling into local optimum, and global path planning and local dynamic obstacle avoidance are achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of path planning, and particularly relates to a robot path planning method for warehouse logistics cluster operation. BACKGROUND

[0002] In modern logistics and warehouse systems, multi-logistics robots are increasingly widely used, and these robots need to efficiently and safely complete tasks such as cargo handling and sorting in complex dynamic environments. Path planning, as a core problem in a multi-logistics robot system, aims to find an optimal or suboptimal path for each robot from a starting point to an ending point while avoiding collisions. Swarm intelligence algorithms, especially ant colony algorithms, are widely used in path planning problems due to their self-organization, positive feedback, and distributed computing characteristics. The ant colony algorithm uses the positive feedback mechanism of pheromone to find the optimal path by simulating the foraging behavior of ants, and has good global search capability. However, the traditional ant colony algorithm still faces some challenges in multi-logistics robot path planning.

[0003] Therefore, how to solve the slow convergence speed and the easy falling into local optimal solution of the ant colony algorithm in robot path planning, and improve the path planning efficiency and accuracy of multi-logistics robots in complex environments is a technical problem to be solved by the application. SUMMARY

[0004] The application aims to provide a robot path planning method for warehouse logistics cluster operation to solve the problems in the background.

[0005] The application is achieved in the following way: a robot path planning method for warehouse logistics cluster operation, characterized in that the method comprises the following steps:

[0006] Step S1: constructing a two-dimensional model graph by using a grid method;

[0007] Step S2: improving the ant colony algorithm;

[0008] Step S3: planning operation paths for each logistics robot by using the improved ant colony algorithm;

[0009] Step S4: performing a local obstacle avoidance strategy by using a dynamic window algorithm to detect whether there are obstacles around; if there are, performing step S5 operation;

[0010] Step S5: judging the collision type and taking corresponding collision avoidance strategies until all robots safely and correctly reach the predetermined target point.

[0011] Preferably, in step S1, the two-dimensional model graph is constructed by using a grid method, and the specific operation is as follows:

[0012] The two-dimensional model map is composed of 900 grids, and the grids are numbered from 1 to 900 from left to right and from top to bottom starting from the first row; the two-dimensional model map is marked, and the white area represents the space that the intelligent vehicle can pass through, and the black area represents the static obstacle that cannot pass through;

[0013] In order to determine the position of the intelligent vehicle in the map, a unique number is assigned to each grid, and the corresponding coordinates (i x , i y ) are found through the number;

[0014]

[0015] Wherein, mod is the remainder operation, and ceil is the positive integer operation;

[0016] When the intelligent vehicle moves on the map, it only needs to record the number of the grid it passes through to know its walking path.

[0017] Preferably, the ant colony algorithm in step S2 is improved, and the specific operation is as follows:

[0018] Step S2-1: setting the parameters of the improved ant colony algorithm, including the initial value of pheromone, the pheromone evaporation coefficient, initializing the parameters of the dynamic window algorithm;

[0019] Step S2-2: introducing Sigmoid function to set the initial concentration of pheromone to avoid blind search in the early stage;

[0020] Step S2-3: placing M ants;

[0021] Step S2-4: selecting the path according to the improved state transition probability, and introducing the obstacle repulsion weight in the state transition probability formula;

[0022] Step S2-5: judging whether the kth ant completes the search, if the kth ant completes the search, the next ant continues to search until M ants complete the search; if the kth ant does not complete the search, jump to step S2-3 for execution;

[0023] Step S2-6: judging whether M ants complete the search, if yes, then jump to step S2-3 for execution;

[0024] Step S2-7: global updating according to the improved pheromone updating method, and outputting the optimal solution.

[0025] Preferably, in step S2-4, the path is selected according to the improved state transition probability, and the specific operation is as follows:

[0026] The improved state transition probability is:

[0027]

[0028] wherein, α is pheromone enhancement coefficient, representing the importance of pheromone on the path (i, j), the greater the value, the greater the influence of pheromone concentration; β is heuristic function factor, representing the importance of the next grid point j to the current grid point i, the greater the value, the greater the importance; σ is the weight parameter of obstacle repulsion weight; allowed k is the set of grid points allowed for the next step of the ant k, τ ij is the pheromone concentration on the path (i, j), η ij is the transition expectation degree of the robot from the grid point i to the grid point j; ω ij is the obstacle repulsion weight from the grid point i to the grid point j;

[0029] [ω is (t)] σ represents the reciprocal of the weight of the ant k moving from the free grid i to the obstacle grid s, and d(i, s) represents the repulsion distance from the free grid i to the obstacle grid s, then:

[0030]

[0031] wherein, d ij (t) represents the obstacle density on the path (i, j) at time t; N ij (t): the number of obstacles on the path (i, j) at time t; L ij is the Euclidean distance between the node i and the node j;

[0032]

[0033] wherein, represents the number of times that the ant k traverses the grid j, and G represents the target grid; d(i, G) is the Euclidean distance from the free grid i to the target grid G.

[0034] Preferably, the Sigmoid function is introduced in the step S2-2 to set the initial concentration of pheromone, avoiding blind search in the early stage, and specifically:

[0035] The Sigmoid function converts the initial pheromone concentration from a constant value to a variable related to the position of the grid node, guiding the ant to have more diversity and randomness when selecting the path at the starting point;

[0036] An adaptive slope parameter h(t) is introduced to dynamically adjust the steepness of the Sigmoid function according to the environmental characteristics, and the expression of the initial pheromone τ(0) is:

[0037]

[0038] wherein q0 is the initial pheromone concentration; x is the Euclidean distance from the current node i to the target node j; d avg is the average distance from the start point (x s ,y s ) to the target point (x e ,y e ); h(t) is the adaptive slope parameter, h base is the base slope parameter of the Sigmoid function, used to control the steepness of the initial pheromone distribution, N obs is the number of obstacles around the current node, N total is the total number of obstacles in the environment.

[0039] Preferably, an obstacle repulsion weight is introduced in the state transition probability formula in step S2-4, specifically:

[0040] The pheromone enhancement coefficient a and the heuristic factor b are improved, and an adaptive parameter adjustment strategy is introduced to dynamically adjust the parameters according to the number of iterations:

[0041]

[0042] wherein t represents at time t, T max represents the maximum number of iterations.

[0043] Preferably, the global update is performed according to the improved pheromone update method in step S2-7, and the improved pheromone update method is:

[0044]

[0045] wherein represents the pheromone increment of the best path, represents the pheromone decrement of the worst path, ζ, λ1 respectively represent the number of ants on the best and worst paths in the current search, ω represents the pheromone adjustment factor, L best represents the optimal path length output by the current search, L worst represents the worst path length output by the current search.

[0046] Preferably, the collision type is determined in step S5, and the improved ant colony algorithm is combined with the DWA algorithm to avoid collision, specifically:

[0047] Rolling window movement: the robot moves along the global optimal path with a step size ε, and the sensor scans the rolling window area with a radius of R = 3 in real time;

[0048] Dynamic obstacle detection: the environmental information in the window is updated every 2 moving steps, and whether there is a potential collision risk is detected

[0049] Collision prediction model: the robot model is regarded as a circular mass point with a diameter of D, and the maximum moving speed is V_max;

[0050] Collision condition: whether the predicted dynamic obstacle and the future trajectory of the robot coincide in a certain grid;

[0051] Obstacle avoidance response time: if a collision risk is detected, the robot pauses for a time Δt = D / (2V_R), where V_R is the current speed;

[0052] Feedback and obstacle avoidance: the collision prediction results are transmitted to the local obstacle avoidance module in real time, triggering the deceleration, detour or pause strategy to ensure safe passage.

[0053] Compared with the prior art, the present application has the following improvements and advantages:

[0054] 1. By introducing the Sigmoid function to initialize the pheromone concentration, blind search in the early stage is avoided, an adaptive slope parameter h(t) is introduced, the steepness of the Sigmoid function is dynamically adjusted according to the environmental characteristics, the state transition probability formula is improved, the ant colony algorithm is combined with the dynamic window algorithm, the parameters are dynamically adjusted according to the iteration number, the obstacle repulsion weight is introduced in the formula, the algorithm is prevented from falling into local optimum, the global path planning can be realized, and local dynamic obstacle avoidance can also be realized.

[0055] 2. By increasing the number of pheromones of the best path in the pheromone update rule and reducing the number of pheromones of the worst path, more ants can select the best path with more pheromones, so as to improve the convergence speed of the algorithm; the ant colony algorithm process is improved, each ant independently finds a path and shares global pheromones, so as to speed up the entire search process. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The improved ant colony algorithm process diagram of the present application.

[0057] Figure 2 The process diagram of the fusion algorithm of the present application.

[0058] Figure 3 The collision detection area diagram of the DWA algorithm of the present application.

[0059] Figure 4 The path planning diagram of the improved ant colony algorithm of the present application.

[0060] Figure 5 The convergence curve trend diagram of the improved ant colony algorithm of the present application.

[0061] Figure 6 The figure is a collision-free route diagram planned by the fusion algorithm of the application for each robot.

[0062] Figure 7 The figure is a convergence curve change trend diagram of the fusion algorithm of the application.

[0063] Figure 8 The figure is a collision avoidance strategy diagram between robots of the application. DETAILED DESCRIPTION

[0064] The application is further described below in combination with the drawings.

[0065] As shown in the figure, a robot path planning method for warehouse logistics cluster operation has the following steps: Figure 1

[0066] Step S1: Construct a two-dimensional model diagram by using a grid method;

[0067] The two-dimensional model diagram is composed of 900 grids, and the grids are numbered from 1 to 900 from left to right and from top to bottom starting from the first row; mark the two-dimensional model diagram, and the white area represents the space through which the intelligent vehicle can pass, and the black area represents the static obstacle that cannot be passed;

[0068] In order to determine the position of the intelligent vehicle in the map, a unique number is assigned to each grid, and the corresponding coordinates (i x , i y ) are found through the number;

[0069]

[0070] Wherein, mod is the remainder operation, and ceil is the positive integer operation;

[0071] When the intelligent vehicle moves on the map, it only needs to record the number of the grid it passes through to know its walking path;

[0072] In step S2, the ant colony algorithm is improved, and the specific operation is as follows:

[0073] Step S2-1: Set the parameters of the improved ant colony algorithm, including the initial value of pheromone, the pheromone evaporation coefficient, and initialize the parameters of the dynamic window algorithm;

[0074] Step S2-2: Introduce the Sigmoid function to set the initial concentration of pheromone to avoid blind search in the early stage;

[0075] ​The sigmoid function converts the initial pheromone concentration from a constant value to a variable related to the position of the grid node. When the guide ants select the path at the starting point, it has more diversity and randomness, which is beneficial to explore the search space; when the path is selected near the target point, the guiding effect of pheromone is fully played, thereby improving the global optimal solution search ability of the algorithm;

[0076] An adaptive slope parameter h(t) is introduced to dynamically adjust the steepness of the sigmoid function according to the environmental characteristics. The expression of the initial pheromone τ(0) is:

[0077]

[0078] Where q0 is the initial pheromone concentration; x is the Euclidean distance from the current node i to the target node j; d avg is the average distance from the starting point (x s , y s ) to the target point (x e , y e ); h(t) is the adaptive slope parameter, h base is the reference slope parameter of the sigmoid function, which is used to control the steepness of the initial pheromone distribution; N obs is the number of obstacles around the current node, and N total is the total number of obstacles in the environment.

[0079] Step S2-3: Place M ants;

[0080] Step S2-4: Select the path according to the improved state transition probability, and introduce the obstacle repulsion weight in the state transition probability formula;

[0081] The improved state transition probability is:

[0082]

[0083] Where α is the pheromone enhancement coefficient, representing the importance of the pheromone on the (i,j) path. The greater the value, the greater the influence of the pheromone concentration; β is the heuristic function factor, representing the importance of the next grid point j to the current grid point i. The greater the value, the greater the importance; σ is the weight parameter of the obstacle repulsion weight; allowed k is the set of grid points that the ant k can access next, τ ij is the pheromone concentration on the (i,j) path, η ij is the transition expectation degree of the robot from the grid point i to the grid point j; ω ij is the obstacle repulsion weight from the grid point i to the grid point j;

[0084] [ω is (t)]σ represents the weight reciprocal of the ant k moving from the free grid i to the obstacle grid s, and d(i,s) represents the repulsion distance from the free grid i to the obstacle grid s, then:

[0085]

[0086] where d ij (t) represents the obstacle density of the path (i,j) at time t; N ij (t): the number of obstacles on the path (i,j) at time t; L ij is the Euclidean distance between node i and node j;

[0087]

[0088] where, represents the number of times that the ant k traverses the grid j, and G represents the target grid; d(i,G) is the Euclidean distance from the free grid i to the target grid G;

[0089] The obstacle repulsion weight is introduced into the state transition probability formula, which is specifically:

[0090] The pheromone enhancement coefficient α and the heuristic function factor β are improved, and an adaptive parameter adjustment strategy is introduced, which dynamically adjusts the parameters according to the number of iterations:

[0091]

[0092] where t is the time at t, T max represents the maximum number of iterations.

[0093] Step S2-5: Determine whether the kth ant has completed the search. If the kth ant has completed the search, the next ant continues to search until all M ants have completed the search. If the kth ant has not completed the search, jump to step S2-3 for continuous execution;

[0094] Step S2-6: Determine whether all M ants have completed the search. If the search is completed, proceed to the next step, then jump to step S2-3 for continuous execution;

[0095] Step S2-7: Perform global updating according to the improved pheromone updating method, and output the optimal solution;

[0096] The improved pheromone updating method is:

[0097]

[0098] where, represents the pheromone increment of the best path, The pheromone reduction of the worst path is represented by ζ, λ1 represents the number of ants on the best and worst paths in the current search, ω represents the pheromone adjustment factor, and L best The optimal path length output by the current search is represented by L worst The worst path length output by the current search is represented by L

[0099] Step S3: The improved ant colony algorithm is used to plan the work path for each logistics robot.

[0100] Step S4: The dynamic window algorithm is used for local obstacle avoidance strategy, and it is detected whether there is an obstacle around; if there is, step S5 operation is performed.

[0101] Step S5: The collision type is judged, and the corresponding collision avoidance strategy is adopted until all the robots safely and accurately reach the predetermined target point; the improved ant colony algorithm and the DWA algorithm are combined for collision avoidance, specifically as follows:

[0102] Rolling window movement: the robot moves along the global optimal path with a step size ε, and the sensor scans the rolling window area with a radius of R=3 in front of the robot in real time;

[0103] Dynamic obstacle detection: the environmental information in the window is updated every 2 movement steps, and it is detected whether there is a potential collision risk

[0104] Collision prediction model: the robot model is regarded as a circular particle with a diameter of D, and the maximum moving speed is V_max;

[0105] Collision condition: it is predicted whether the future trajectory of the robot and the dynamic obstacle coincide in a certain grid;

[0106] Obstacle avoidance response time: if a collision risk is detected, the robot pauses for a time Δt=D / (2V_R), where V_R is the current speed;

[0107] Feedback and obstacle avoidance: the collision prediction result is transmitted to the local obstacle avoidance module in real time, triggering the deceleration, detour or pause strategy to ensure safe passage.

[0108] During the movement of each robot, collision detection is performed through the DWA algorithm, and all the robots can finally safely and accurately reach the target point. The robot uses the dynamic window method to detect dynamic obstacles and other robots in the nearby environment, and the detection range radius is 3 grid lengths, as shown in Figure 3 When the existence of dynamic obstacles and other robots is detected, it is determined whether a collision will occur through collision prediction analysis, if no collision occurs, the robot is allowed to move along the original global planning path, if there is a collision, the local collision avoidance strategy is adopted, and after avoiding, the dynamic window is used again to collect the current environmental information until the robot reaches the target point.

[0109] As Figure 2 shown, the flow chart of the ant colony algorithm fusion DWA algorithm in the face of obstacles and robot collision problem, in the conflict with dynamic obstacles, will judge its conflict type, using local path planning method to solve the conflict. And the conflict between robots is solved by priority strategy.

[0110] As Figure 4 shown, Figure 4 The middle red represents the improved ant colony algorithm planning path, and the blue is the path after removing the redundant turning points. The path will become shorter.

[0111] Figure 5 The convergence curve changes are shown. The convergence tends to be stable around the 10th generation.

[0112] According to Figure 6 It can be seen that the improved ant colony algorithm fusion DWA algorithm plans a collision-free route from the starting point to the target point for each robot.

[0113] According to Figure 7 It can be seen that the convergence curve changes when the three logistics robots plan the path. The collision avoidance strategy between the robots of the application is shown in Figure 8 By comparing the Euclidean distance from the point that may produce collision to the target point, the priority of the distance is greater than that of the small distance. From the distance comparison in Figure 8 AGV1 has a higher priority than AGV2, so AGV2 will stop and wait for AGV1 to pass first.

[0114] The above only describes the embodiments of the application and is not used to limit the application. The application can have various changes and variations for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the application shall be included in the scope of the claims of the application.

Claims

1. A robot path planning method for warehouse logistics cluster operations, characterized in that: The method includes the following steps: Step S1: Construct a two-dimensional model diagram using the raster method; Step S2: Improve the ant colony algorithm; Step S3: Use the improved ant colony algorithm to plan the operation path for each logistics robot; Step S4: Implement a local obstacle avoidance strategy using a dynamic window algorithm to detect whether there are obstacles in the surrounding area; if so, proceed to step S5. Step S5: Determine the collision type and adopt the corresponding collision avoidance strategy until all robots safely reach the predetermined target point.

2. The robot path planning method for warehousing and logistics cluster operations according to claim 1, characterized in that: In step S1, a two-dimensional model diagram is constructed using the raster method. The specific operation is as follows: The two-dimensional model consists of 900 grids, which are numbered from 1 to 900 from left to right and top to bottom, starting from the first row. The two-dimensional model is marked with white areas representing spaces that the intelligent vehicle can pass through, and black areas representing static obstacles that cannot pass through. To pinpoint the location of the intelligent vehicle on the map, a unique number was assigned to each grid cell, and the corresponding coordinates were located using this number. x i y ); Here, mod is the modulo operation, and ceil is the positive integer operation; When a smart car moves on a map, it only needs to record the grid numbers it passes through to know its travel path.

3. The robot path planning method for warehousing and logistics cluster operations according to claim 1, characterized in that: The ant colony algorithm in step S2 is improved, specifically as follows: Step S2-1: Set the parameters of the improved ant colony algorithm, including the initial pheromone value, the pheromone evaporation coefficient, and the parameters of the dynamic window algorithm; Step S2-2: Introduce the Sigmoid function to set the initial concentration of pheromones, avoiding blind searching in the early stages; Step S2-3: Place M ants; Step S2-4: Select a path according to the improved state transition probability, and introduce obstacle rejection weights into the state transition probability formula; Step S2-5: Determine whether the k-th ant has completed the search. If the k-th ant has completed the search, the next ant continues the search until all M ants have completed the search. If the k-th ant has not completed the search, jump to step S2-3 to continue execution. Step S2-6: Determine whether all M ants have completed the search. If they have, proceed to the next step or jump to step S2-3 to continue execution. Step S2-7: Perform a global update based on the improved pheromone update method and output the optimal solution.

4. The robot path planning method for warehousing and logistics cluster operations according to claim 3, characterized in that: In step S2-2, the Sigmoid function is introduced to set the initial concentration of pheromones, avoiding blind searching in the early stages. Specifically: The Sigmoid function transforms the initial pheromone concentration from a fixed value into a variable related to the position of the grid nodes, guiding ants to choose paths at the starting point with greater diversity and randomness; An adaptive slope parameter h(t) is introduced to dynamically adjust the steepness of the Sigmoid function based on environmental characteristics. The expression for the initial pheromone τ(0) is: Where q0 is the initial pheromone concentration; x is the Euclidean distance from the current node i to the target node j; d avg Starting point (x) s ,y s ) to target point (x e ,y e The average distance of ) ; j(t) is the adaptive slope parameter, j base N is the baseline slope parameter of the sigmoid function, used to control the steepness of the initial pheromone distribution. obs N represents the number of obstacles surrounding the current node. total This represents the total number of obstacles in the environment.

5. The robot path planning method for warehousing and logistics cluster operations according to claim 3, characterized in that: In step S2-4, the path selection based on the improved state transition probability is specifically as follows: The improved state transition probability is: Where α is the pheromone enhancement coefficient, representing the importance of pheromones on the (i,j) path; the larger the value, the greater the influence of pheromone concentration. β is the heuristic function factor, representing the importance of the next grid point j to the current grid point i; the larger the value, the greater the importance. σ is the weighting parameter for obstacle repulsion weights. k Let τ be the set of grid points that ant k is allowed to visit next. ij Let η be the pheromone concentration along the path (i,j). ij ω represents the expected transfer speed of the robot from grid point i to grid point j. ij Obstacle repulsion weight from grid point i to grid point j; [ω is (t)] σ Let d(i,s) represent the reciprocal of the weight of the position ant k moves from free grid i to obstacle grid s, and let d(i,s) represent the repulsion distance from free grid i to obstacle grid s. Where, d ij (t) represents the obstacle density at point (i,j) on the path at time t; N ij (t): The number of obstacles on path (i,j) at time t; L ij Let be the Euclidean distance between node i and node j; in, Let d(i,G) represent the number of times ant k traverses grid j, and G represent the target grid; d(i,G) is the Euclidean distance from free grid i to target grid G.

6. The robot path planning method for warehousing and logistics cluster operations according to claim 3, characterized in that: In step S2-4, an obstacle repulsion weight is introduced into the state transition probability formula, specifically as follows: Improvements were made to the pheromone enhancement coefficient α and the heuristic function factor β, and an adaptive parameter adjustment strategy was introduced to dynamically adjust the parameters based on the number of iterations. Where t represents the time T at time t. max This represents the maximum number of iterations.

7. The robot path planning method for warehousing and logistics cluster operations according to claim 3, characterized in that: In steps S2-7, a global update is performed based on an improved pheromone update method. The improved pheromone update method is as follows: in, This represents the pheromone increment for the optimal path. This represents the pheromone reduction on the worst path, ζ and λ1 represent the number of ants on the best and worst paths in the current search, respectively, ω represents the pheromone adjustment factor, and L... best L represents the optimal path length output in this round of search. worst This indicates the worst path length output in this round of search.

8. The robot path planning method for warehousing and logistics cluster operations according to claim 1, characterized in that: In step S5, the collision type is determined, and collision avoidance is achieved by fusing the improved ant colony algorithm with the DWA algorithm. Specifically: Scrolling window movement: The robot moves along the globally optimal path with a step size ε, and the sensor scans the scrolling window area with a radius of R=3 in front of it in real time; Dynamic obstacle detection: The environmental information within the window is updated every two movement steps to detect potential collision risks. Collision prediction model: The robot model is regarded as a circular point mass with a diameter of D and a maximum moving speed of V_max; Collision condition: Predict whether the dynamic obstacle and the robot's future trajectory will coincide in a certain grid cell; Obstacle avoidance response time: If a collision risk is detected, the robot pauses for a time Δt = D / (2V_R), where V_R is the current speed; Feedback and obstacle avoidance: The collision prediction results are transmitted to the local obstacle avoidance module in real time to trigger deceleration, detour or pause strategies to ensure safe passage.

Images