A box storage robot path planning method based on an improved particle swarm algorithm

Abstract

The application discloses a box storage robot path planning method based on an improved particle swarm algorithm and belongs to the technical field of path optimization. The method comprises the following steps: S1, performing mathematical modeling on a movement mode of the box storage robot; S2, constructing an optimization function with the minimum picking time as a target; S3, configuring a simulation experiment environment; and S4, performing picking operation path optimization on the box storage robot by using the improved particle swarm algorithm in the simulation experiment environment. The application can accurately describe the movement state of the box storage robot in the picking operation and the picking operation time, can find a global optimal solution with a higher probability compared with a standard particle swarm algorithm under a case that goods boxes are delivered to a human picking system, and the convergence iteration number is superior to that of the standard particle swarm algorithm.

Description

Technical Field

[0001] This invention belongs to the field of path optimization technology, specifically relating to a path planning method for a box-type warehouse robot based on an improved particle swarm optimization algorithm. Background Technology

[0002] With rapid economic development, the increasing demand from customers for diverse, small-batch, and personalized products is placing higher demands on industries such as footwear and apparel, 3PL, e-commerce, 3C manufacturing, and retail. The contradiction between the massive number of SKUs and the shallow depth of BOM (Bill of Materials) is becoming increasingly serious, leading to heavy warehouse storage pressure and difficulties in managing the picking process. Meanwhile, rising labor costs are also posing significant challenges to these industries. Autonomous Case-Handling Robotics (ACRs), as picking robots in a case-to-person picking system, offer advantages such as high picking efficiency, multi-case handling, high positioning accuracy, and stable operation. However, ACRs require addressing path planning and trajectory planning issues during the picking process.

[0003] In existing technologies, the main methods for solving path planning problems include Dijkstra's algorithm, A* algorithm, dynamic programming algorithm, genetic algorithm, and particle swarm optimization algorithm. Among them, particle swarm optimization algorithm has the advantages of strong versatility, ease of implementation, and easier finding of the global optimum, but it also has the disadvantages of being prone to getting trapped in local optima and having low optimization accuracy. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, the purpose of this invention is to provide a path planning method for box-type warehouse robots based on an improved particle swarm optimization algorithm, which can optimize the path in a box-to-person picking system.

[0005] This invention provides the following technical solution:

[0006] A path planning method for a box-type storage robot based on an improved particle swarm optimization algorithm includes the following steps:

[0007] S1. Mathematically model the motion patterns of the box-type warehousing robot, including horizontal movement, fork lifting and lowering movement, fork storage and retrieval operations, and turning.

[0008] S2. Based on the motion analysis of the box-type storage robot in step 1, construct an optimization function with the goal of minimizing picking time;

[0009] S3. Configure the simulation experimental environment, including warehouse environment parameters, box-type storage robot parameters, and picking task box positions;

[0010] S4. Using an improved particle swarm optimization algorithm, the picking operation path of the box-type storage robot is optimized in the simulation experimental environment configured in step 3.

[0011] Furthermore, in step S4, the specific process of improving the particle swarm algorithm is as follows:

[0012] 1) Initialize the position and velocity of the particle swarm;

[0013] 2) Calculate the fitness value of each particle;

[0014] 3) Store individual particles in Ppb; compare the fitness values ​​of each particle and store the globally optimal particle in Pgb;

[0015] 4) Update the individual swapper learning probability r1 and the global swapper learning probability r2 according to the iteration time;

[0016] 5) Calculate the swap sequence ss, and update the particle position and particle velocity;

[0017] 6) Update particle fitness values;

[0018] 7) Compare the best fitness values ​​of the particles during the iteration process. If the fitness value is better, update the best individual particle Ppb.

[0019] 8) Compare all current particles Ppb and Pgb, and update the globally optimal particle Pgb;

[0020] 9) If the termination condition is met, the iterative search process ends and the optimal particle is output as the final optimization result; otherwise, return to 4) and continue the iterative search.

[0021] Furthermore, in step S1,

[0022] The formula for calculating the horizontal movement distance is:

[0023]

[0024]

[0025] Where D ij_ x represents the distance traveled along the x-axis from cargo box position i to cargo box position j. ij_y Let s be the y-axis distance that cargo box i moves from position j to position j. ij s represents the number of transport channels that the ACR needs to cross during horizontal movement. ij =ceil(y j ,g)-ceil(y i ,g), w is the width of the storage compartment, d is the depth of the storage compartment, g is the number of storage locations per row or column of shelving, l is the width of the picking aisle, and L is the width of the transport aisle;

[0026] The formula for calculating the forklift travel distance is:

[0027] Di_z =(z i +k i -z i +k i )×h

[0028] Where D i_z k represents the distance traveled along the z-axis from cargo box position i to cargo box position j. i Let h be the coordinates of the basket, and h be the height of the cargo compartment.

[0029] The time for forklift loading and unloading operations is:

[0030] T λ =num

[0031] T λ This represents the time for a single forklift loading / unloading operation, where num is a user-defined fixed value.

[0032] The formula for calculating the number of turns is:

[0033]

[0034] Where, μ ij This represents the number of turns.

[0035] Furthermore, in step S1,

[0036] The formula for calculating horizontal movement time is:

[0037]

[0038] Where T1 is the total horizontal movement time, T ij Let A be the time for horizontal movement from container i to container j. ij For the picking operation constraint, n is the number of task sequences per vehicle trip;

[0039] The formula for calculating the fork lifting and moving time is:

[0040]

[0041] Where T2 is the total lifting and moving time of the forks, and V z The lifting and moving speed of the forks;

[0042] The formula for calculating the time for storage and retrieval operations is:

[0043]

[0044] T3 represents the total storage and retrieval operation time;

[0045] The formula for calculating turning time is:

[0046]

[0047] Where T4 is the total turning time, T μ This refers to the time for a single turn.

[0048] Furthermore, the optimization function in step S2, which aims to minimize picking time, is the same as the calculation formula for initializing the particle population fitness in step S4, both being:

[0049]

[0050] Furthermore, in step S4, the improved particle swarm velocity update formula is as follows:

[0051]

[0052] Where Vi(t+1) represents the particle velocity at time t+1, Vi(t) represents the particle velocity at time t, r1 represents the individual exchanger learning probability, r2 represents the global exchanger learning probability, Ppb represents the individual optimal particle, Pgb represents the global optimal particle, Xi(t) represents the particle position at time t, (Ppb-Xi(t)) represents the individual exchanger sequence, and (Pgb-Xi(t)) represents the global exchanger sequence.

[0053] The improved particle swarm particle position update formula is as follows:

[0054]

[0055] Where Xi(t+1) represents the particle position at time t+1;

[0056] The formula for calculating individual exchange subsequences is:

[0057] SS pb =P pb -X i (t)

[0058] Where SSPb represents the individual exchange subsequence;

[0059] The formula for calculating the global swapped subsequence is:

[0060] SS gb =P gb -X i (t)

[0061] Where SSgb represents the global exchange subsequence;

[0062] The formula for calculating the learning probability of a nonlinear asynchronous commutator is:

[0063]

[0064]

[0065] Where r1max is the maximum individual swapper learning probability, r1min is the minimum individual swapper learning probability, r2max is the maximum global swapper learning probability, r2max is the minimum global swapper learning probability, t is the number of iterations, and tmax is the maximum number of iterations.

[0066] By employing the above-described technology, the beneficial effects of the present invention compared to the prior art are as follows:

[0067] 1) This invention can accurately describe the motion state and picking time of the box-type warehousing robot in the picking operation; it can find the global optimal solution with a higher probability than the standard particle swarm algorithm in the box-to-person picking system, and the number of convergence iterations is better than that of the standard particle swarm algorithm.

[0068] 2) This invention incorporates a swap sequence and a nonlinear asynchronous swapper learning probability into the standard particle swarm optimization algorithm. Through the swap sequence and the nonlinear asynchronous swapper learning probability, the individual swapper learning probability takes a larger value in the early stage of the algorithm and a smaller value in the later stage, while the global swapper learning probability takes a smaller value in the early stage of the algorithm and a larger value in the later stage. This enables the algorithm to converge quickly in the early stage of the iteration and to continue searching for the global optimal solution after convergence, and its optimization ability is stronger than that of the standard particle swarm optimization algorithm. Attached Figure Description

[0069] Figure 1 This is a flowchart of the steps of the method of the present invention;

[0070] Figure 2 This is a basic flowchart of the improved particle swarm algorithm of this invention;

[0071] Figure 3 This is a graph showing the change in the learning probability of the nonlinear asynchronous commutator of this invention with the number of iterations.

[0072] Figure 4 This diagram illustrates the process of the optimal fitness value decreasing with the number of iterations in the picking operation model of this invention using both the standard particle swarm optimization algorithm and the improved particle swarm optimization algorithm proposed in this invention, over 500 iterations. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0074] Conversely, this invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of the invention as defined in the claims. Furthermore, to provide a better understanding of the invention, certain specific details are described in detail below. However, those skilled in the art will fully understand the invention even without these detailed descriptions.

[0075] Example:

[0076] Please see Figure 1-4 The simulation experiment was conducted in a warehouse with a regular layout of 8 rows, 32 columns, and 16 layers. There was one ACR (Automatic Picker) with 8 baskets. The I / O point coordinates were (1,0,0), depth d = 0.5m, width w = 0.45m, height h = 0.35m, picking aisle length l = 1.1m, transport aisle length L = 1.4m, ACR basket count k = 8, ACR horizontal movement speed Vx = 1.8m / s, ACR fork lifting speed Vz = 0.5m / s, ACR horizontal acceleration / deceleration ax = 0.6m / s, ACR single-cycle storage / retrieval time Tλ = 7.5s, and ACR single-cycle turning time Tμ = 3s. At a certain moment, the picking task contained 40 items, therefore requiring 5 AMR (Automatic Picker) operations to complete the picking operation.

[0077] like Figure 1 The diagram shown is a flowchart of the method in this embodiment, including the following steps:

[0078] Step 1: Perform mathematical modeling of the motion patterns of the box-type storage robot, including horizontal movement, fork lifting and lowering movement, fork storage and retrieval operations, and turning.

[0079] Step 2: Based on the motion analysis of the box-type storage robot in Step 1, construct an optimization function with the goal of minimizing picking time;

[0080] Step 3: Configure the simulation experimental environment, including warehouse environment parameters, box-type storage robot parameters, and picking task box positions;

[0081] Step 4: Optimize the picking operation path of the box-type storage robot using the improved particle swarm optimization algorithm in the simulation experimental environment configured in Step 3.

[0082] The specific process of improving the particle swarm optimization algorithm is as follows:

[0083] 1) Initialize the position and velocity of the particle swarm;

[0084] 2) Calculate the fitness value of each particle;

[0085] 3) Store individual particles in Ppb; compare the fitness values ​​of each particle and store the globally optimal particle in Pgb;

[0086] 4) Update the individual swapper learning probability r1 and the global swapper learning probability r2 according to the iteration time;

[0087] 5) Calculate the swap sequence ss, and update the particle position and particle velocity;

[0088] 6) Update particle fitness values;

[0089] 7) Compare the best fitness values ​​of the particles during the iteration process. If the fitness value is better, update the best individual particle Ppb.

[0090] 8) Compare all current particles Ppb and Pgb, and update the globally optimal particle Pgb;

[0091] 9) If the termination condition is met, the iterative search process ends and the optimal particle is output as the final optimization result; otherwise, return to 4) and continue the iterative search.

[0092] Table 1 shows the location of the picking task boxes in step 3 of this embodiment.

[0093] Picking operation number Cargo container position (x, y, z) Picking operation number Cargo container position (x, y, z) 1 (3，11，10) 21 (4，1，9) 2 (8，12，5) 22 (6，18，11) 3 (8，9，15) 23 (6，30，8) 4 (3，26，8) 24 (7，18，13) 5 (7，10，4) 25 (4，12，16) 6 (6，20，2) 26 (7，15，，6) 7 (7，30，16) 27 (8，13，2) 8 (2，25，5) 28 (5，14，4) 9 (2，31，2) 29 (7，1，1) 10 (2，10，1) 30 (3，19，3) 11 (8，10，8) 31 (2，32，9) 12 (5，7，14) 32 (3，29，8) 13 (1，25，9) 33 (1，3，15) 14 (4，9，12) 34 (7，11，9) 15 (6，12，12) 35 (3，12，10) 16 (4，22，10) 36 (3，20，4) 17 (5，31，7) 37 (4，3，11) 18 (4，9，7) 38 (3，22，9) 19 (3，28，6) 39 (1，14，7) 20 (3，9，2) 40 (4，26，8))

[0094] Table 1

[0095] The formula for calculating the picking time includes:

[0096] The formula for calculating the horizontal movement distance is:

[0097]

[0098]

[0099] Where D ij _x represents the distance moved along the x-axis from cargo box position i to cargo box position j, and D ij_y Let s be the y-axis distance that cargo box i moves from position j to position j. ij The number of transport aisles that the ACR needs to cross during horizontal movement is given by w, where w is the shelf width, d is the shelf depth, g is the number of storage locations per row or column of shelving, l is the picking aisle width, and L is the transport aisle width.

[0100] In step 1, the formula for calculating the number of transport channels that the ACR needs to cross during horizontal movement is:

[0101] s ij =|ceil(y j ,g)-ceil(y i ,g)|

[0102] The formula for calculating the forklift travel distance is:

[0103] D i_z =(z i +|k i -z i |+k i )×h

[0104] Where D i_z k represents the distance traveled along the z-axis from cargo box position i to cargo box position j. i Let h be the coordinates of the basket, and h be the height of the compartment.

[0105] The time for forklift loading and unloading operations is:

[0106] T λ =num

[0107] T λ This represents the time required for a single forklift loading / unloading operation, where num is a user-defined fixed value.

[0108] The formula for calculating the number of turns is:

[0109]

[0110] Where, μ ij This represents the number of turns.

[0111] The formula for calculating horizontal movement time is:

[0112]

[0113] Where T1 is the total horizontal movement time, T ij Let A be the time for horizontal movement from container i to container j. ij Constraints for picking operations.

[0114] The formula for calculating the fork lifting and moving time is:

[0115]

[0116] Where T2 is the total lifting and moving time of the forks, and V z This refers to the lifting and moving speed of the forks.

[0117] The formula for calculating the time for storage and retrieval operations is:

[0118]

[0119] T3 represents the total time for storage and retrieval operations.

[0120] The formula for calculating turning time is:

[0121]

[0122] Where T4 is the total turning time, T μ This refers to the time for a single turn.

[0123] The formula for calculating the total picking operation time is:

[0124]

[0125] In this embodiment, the basic parameters of the improved particle swarm optimization algorithm are set as follows: population size = 500, maximum number of iterations tmax = 500, r1max = 0.75, r1max = 0.65, r1max = 0.85, r1max = 0.75. The learning probability of the nonlinear asynchronous squaring sub-particle changes with the number of iterations. When the number of iterations reaches 500, the iteration process ends and the optimal particle is output.

[0126] like Figure 3 The diagram shown is an iterative graph of the nonlinear asynchronous commutator learning probability in an embodiment of the present invention, and its calculation formula is as follows:

[0127]

[0128]

[0129] like Figure 4 The figure shown is a process diagram of how the optimal fitness value of the picking operation model in this embodiment decreases with the number of iterations in the standard particle swarm algorithm and the improved particle swarm algorithm proposed in this invention during 500 iterations.

[0130]

[0131] Table 2

[0132] Table 2 shows the results of the picking operation model in this embodiment after optimization using the standard particle swarm optimization (PSO) algorithm and the improved PSO algorithm proposed in this invention. The comparison shows that the improved PSO algorithm converges quickly in the early stages of iteration and continues to optimize after convergence. Its optimal fitness function decreases rapidly and is superior to that of the standard PSO algorithm. It can be seen that the improved PSO algorithm outperforms the standard PSO algorithm in both convergence iteration speed and global optimization capability.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A path planning method for a box-type storage robot based on an improved particle swarm optimization algorithm, characterized in that: Includes the following steps: S1. Mathematically model the motion patterns of the box-type warehousing robot, including horizontal movement, fork lifting and lowering movement, fork storage and retrieval operations, and turning. S2. Based on the motion analysis of the box-type storage robot in step 1, construct an optimization function with the goal of minimizing picking time; S3. Configure the simulation experimental environment, including warehouse environment parameters, box-type storage robot parameters, and picking task box positions; S4. Using an improved particle swarm optimization algorithm, the picking operation path of the box-type storage robot is optimized in the simulation experimental environment configured in step 3. In step S1 The formula for calculating the horizontal movement distance is: Where D ij_ x represents the distance traveled along the x-axis from cargo box position i to cargo box position j. ij_y Let s be the y-axis distance that cargo box i moves from position j to position j. ij s represents the number of transport channels that the ACR needs to cross during horizontal movement. ij =|ceil(y j ,g)-ceil(y i ,g)|,w is the width of the storage compartment,d is the depth of the storage compartment,g is the number of storage locations per row and column of shelves,l is the width of the picking aisle, andL is the width of the transport aisle; The formula for calculating the forklift travel distance is: D i_z =(z i +|k i -z i |+k i )×h Where D i_z k represents the distance traveled along the z-axis from cargo box position i to cargo box position j. i Let h be the coordinates of the basket, and h be the height of the cargo compartment. The time for forklift loading and unloading operations is: T λ =number T λ This represents the time for a single forklift loading / unloading operation, where num is a user-defined fixed value. The formula for calculating the number of turns is: Where, μ ij Number of turns; The formula for calculating horizontal movement time is: Where T1 is the total horizontal movement time, T ij Let A be the time for horizontal movement from container i to container j. ij For the picking operation constraint, n is the number of task sequences per vehicle trip; The formula for calculating the fork lifting and moving time is: Where T2 is the total lifting and moving time of the forks, and V z The lifting and moving speed of the forks; The formula for calculating the time for storage and retrieval operations is: T3 represents the total storage and retrieval operation time; The formula for calculating turning time is: Where T4 is the total turning time, T μ This refers to the time for a single turn; The optimization function aimed at minimizing picking time in step S2 is the same as the calculation formula for initializing particle population fitness in step S4, both being:

2. The path planning method for a box-type storage robot based on an improved particle swarm optimization algorithm according to claim 1, characterized in that... In step S4, the specific process of improving the particle swarm algorithm is as follows: 1) Initialize the position and velocity of the particle swarm; 2) Calculate the fitness value of each particle; 3) Store individual particles in Ppb; compare the fitness values ​​of each particle and store the globally optimal particle in Pgb; 4) Update the individual swapper learning probability r1 and the global swapper learning probability r2 according to the iteration time; 5) Calculate the swap sequence ss, and update the particle position and particle velocity; 6) Update particle fitness values; 7) Compare the best fitness values ​​of the particles during the iteration process. If the fitness value is better, update the best individual particle Ppb. 8) Compare all current particles Ppb and Pgb, and update the globally optimal particle Pgb; 9) If the termination condition is met, the iterative search process ends and the optimal particle is output as the final optimization result; otherwise, return to 4) and continue the iterative search.

3. The path planning method for a box-type storage robot based on an improved particle swarm optimization algorithm according to claim 2, characterized in that... In step S4, the improved particle swarm velocity update formula is as follows: Where Vi(t+1) represents the particle velocity at time t+1, Vi(t) represents the particle velocity at time t, r1 represents the individual exchanger learning probability, r2 represents the global exchanger learning probability, Ppb represents the individual optimal particle, Pgb represents the global optimal particle, Xi(t) represents the particle position at time t, (Ppb-Xi(t)) represents the individual exchanger sequence, and (Pgb-Xi(t)) represents the global exchanger sequence. The improved particle swarm particle position update formula is as follows: Where Xi(t+1) represents the particle position at time t+1; The formula for calculating individual exchange subsequences is: SS pb =P pb -X i (t) Where SSPb represents the individual exchange subsequence; The formula for calculating the global swapped subsequence is: SS gb =P gb -X i (t) Where SSgb represents the global exchange subsequence; The formula for calculating the learning probability of a nonlinear asynchronous commutator is: Where r1max is the maximum individual swapper learning probability, r1min is the minimum individual swapper learning probability, r2max is the maximum global swapper learning probability, r2min is the minimum global swapper learning probability, t is the number of iterations, and tmax is the maximum number of iterations.

Images