CN113516429A - Multi-AGV global planning method based on network congestion model - Google Patents

Multi-AGV global planning method based on network congestion model Download PDF

Info

Publication number
CN113516429A
CN113516429A CN202110379885.0A CN202110379885A CN113516429A CN 113516429 A CN113516429 A CN 113516429A CN 202110379885 A CN202110379885 A CN 202110379885A CN 113516429 A CN113516429 A CN 113516429A
Authority
CN
China
Prior art keywords
area
agv
region
time
cost
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110379885.0A
Other languages
Chinese (zh)
Other versions
CN113516429B (en
Inventor
谢巍
周雅静
杨启帆
廉胤东
钱文轩
林丹淇
张宜旭
林健峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN202110379885.0A priority Critical patent/CN113516429B/en
Publication of CN113516429A publication Critical patent/CN113516429A/en
Application granted granted Critical
Publication of CN113516429B publication Critical patent/CN113516429B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0287—Control of position or course in two dimensions specially adapted to land vehicles involving a plurality of land vehicles, e.g. fleet or convoy travelling
    • G05D1/0289—Control of position or course in two dimensions specially adapted to land vehicles involving a plurality of land vehicles, e.g. fleet or convoy travelling with means for avoiding collisions between vehicles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/083—Shipping
    • G06Q10/0835—Relationships between shipper or supplier and carriers
    • G06Q10/08355—Routing methods

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Marketing (AREA)
  • Theoretical Computer Science (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Educational Administration (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a multi-AGV global planning method based on a network congestion model, which comprises the steps of firstly establishing a regionalized map, distributing tasks to AGVs according to transportation requirements, and determining an initial area and a target area of an AGV path; then, on the basis of the distance cost, introducing time cost containing turning cost and a regional density estimation value based on a corrected network congestion diffusion model to update the estimated cost value of the A-algorithm; and finally, performing inter-area global path planning based on an improved A-algorithm to obtain a path area set connecting the current area and the target area. The invention improves the traditional A-star algorithm according to the index of the regional congestion condition. According to the method, the network congestion model is used for predicting the congestion condition of each area, the AGVs are reasonably distributed in each area as much as possible during scheduling, the transportation efficiency of a multi-AGV system is improved, and the complexity of a scheduling algorithm is reduced.

Description

Multi-AGV global planning method based on network congestion model
Technical Field
The invention belongs to the field of intelligent vehicle path planning, and relates to a multi-AGV global planning method based on a network congestion model.
Background
The rise of electronic commerce in China promotes the development of the logistics storage industry, express business volume keeps a high-speed growth situation, sorting and transporting work of a logistics storage is increasingly heavy, a large number of electronic commerce have advanced the field of military storage automation, and the mechanical sorting and transporting is tried to replace manual work, so that the Kiva robot in Amazon is used firstly, various storage AGVs in Jingdong, vegetable and bird and the like are used later, and the storage automation and intelligentization have become trends.
However, at present, domestic intelligent warehousing is still in a development stage, and the prior art still has a space for improvement. In the aspect of a single AGV path planning algorithm, an A algorithm and a D algorithm are adopted in the existing mainstream method, and although the two classical path planning methods can achieve good effects when a single machine path is planned, the traditional A algorithm and the D algorithm have certain limitations on the multi-AGV collaborative path planning in a complex environment. The default map environment is static when the traditional A-x algorithm plans the path, so that if the environment changes, if a new obstacle appears, the original planned path is discarded and needs to be re-planned, and the path planning method is not suitable for complex warehousing environments with real-time change; the improved D-algorithm considers the real-time change of the environment and reduces the calculation amount of the re-planning to a certain extent. When the D algorithm detects the change of the surrounding environment, the AGV replans and finds the optimal path. However, when the number of AGVs increases, the number of times of re-planning increases, which brings challenges to the real-time scheduling task of large-scale AGVs.
Therefore, the invention utilizes the characteristic that the A-algorithm path is fixed, introduces time variables to improve the A-algorithm, greatly predicts the congestion condition of each region by improving the existing Network congestion diffusion model (G.Wen, N.Huang, C.Wang, Network connection differentiation model configuration distribution information, IEEE Access 7(2019) 102064. once-in 072.) (A.Tareen, N.S.Wigreen, R.Mukholdheya, Modeling evaluation of congestion in notification networks, Phys.Rev.E97(2) (2018)20402), and reasonably distributes AGV in each region as much as possible during scheduling, so that the scheduling effect of the A-algorithm is similar to the scheduling effect of the A-algorithm planning control D.9. Auga.9. after scheduling algorithm (AGV.9. update) and the scheduling times of the A-algorithm is reduced by the routing algorithm (C.9. update). Through tests, the method and the device can be suitable for large-scale AGV scheduling, and a scientific balance is carried out between algorithm complexity and scheduling efficiency.
Disclosure of Invention
The invention aims to improve the existing path planning algorithm by correcting a network congestion diffusion model so as to improve the transportation efficiency of a multi-AGV system and the space utilization rate of an intelligent warehouse, and therefore, the multi-AGV global planning method based on the network congestion model is provided.
The invention is realized by at least one of the following technical schemes.
A multi-AGV global planning method based on a network congestion model comprises the following steps:
1) establishing a regional storage environment map, and distributing transportation tasks for each idle AGV;
2) respectively acquiring distance cost and time cost of each region;
3) predicting the AGV density of each area, and establishing an improved network congestion diffusion model according to the scheduling of the storage system;
4) acquiring a density estimated value of each area according to the AGV density of each area;
5) and acquiring the cost value of each region and outputting an optimal path region set.
Preferably, step 1) comprises: and processing the global image of the storage environment to acquire the serial number r and the coordinate of each AGV, and distributing a transportation task for each idle AGV, so as to determine the starting area O and the target area E of the path in the current task of each idle AGV.
Preferably, before the distance cost and the time cost of each region are obtained, a set starting region O is established as a father node, the father node is placed into a close set, an adjacent region of the father node is led into an open set, if the target region is in the open set, the step 5) is performed, and if the target region is not in the open set, the step 3) is performed.
Preferably, the distance cost H for the region i(r，i)For Manhattan distance, i.e. current region i and AGVrAbsolute value of distance difference between task target areas E; let the coordinates of region i be (x)i，yi)，AGVrThe coordinates of the task target area E are (x)(r，E)，y(r，E)) Then the distance cost H of the region i(r，i)The formula of (1) is as follows:
H(r，i)＝|x(r，E)-xi|+|y(r，E)-yi|
time cost F of region iiTravel time t for complete passage through zone iiExtra turn cost TN with zone iiSum, in particular, tiTime taken to completely pass region i if AGVrIn the case of no steering behavior during driving in zone i, TNiIs marked as 0;
for region i, time cost FiApparently a time-invariant constant, and the target area E of an AGV to which a task has been assigned is determined, so H is in a single task(r，i)Also a time invariant constant.
Preferably, the improved network congestion diffusion model is a universal network congestion diffusion model which is based on a Langevin diffusion model and considers dynamic influence, the improved network congestion diffusion model regards the whole storage environment as a network, one area of the storage environment is a node, the driving-in and driving-out of the AGV are regarded as resource exchange among the nodes, and a Langevin diffusion equation of the improved network congestion diffusion model for a specific area i is defined as follows:
Figure BDA0003012536930000031
wherein
Figure BDA0003012536930000032
AGV Density for region i, Ai(t) and Bi(t) static and dynamic parts of the Langevin diffusion equation, Ai(t) is the difference between the number of AGVs entering the area i and the number of AGVs exiting the area i at the predetermined time t, and is defined as follows:
Figure BDA0003012536930000041
Bi(t) is the AGV exchange number of the region set Ne to which the region i is connected, and the random Gaussian white noise matrix xi with the average value of 0i(t) the dynamic error, B, representing the number of AGVs in region i after multiplicationi(t) is specifically defined as follows:
Figure BDA0003012536930000042
wherein, Ci(t) density of free space resources in region i, Cj(t) is the density of idle space resources in region j;
Figure BDA0003012536930000043
the AGV density is the AGV density which is common to the area i at the time t and the area j to be driven into at the time (t + 1); nei is a set of regions to which region i is connected, i.e., the reachable regions of all AGVs at the next time, have
Figure BDA0003012536930000044
K(i，j)(t) is calculated from the following formula:
Figure BDA0003012536930000045
wherein k isiIndicating the degree of importance of the region i, the region set Ne being connected by the region iiThe number of middle zones is related to the geographic location; k(j，i)(t) represents the proportion that the AGV can smoothly move to the area j in the area i at the time t, and is used for evaluating the congestion degree of one area, and if the AGV moves to the area j completely, the value is 1; if the area is congested, the value is less than 1.
Preferably, for a path that is a single channel, Ai(t) and BiThe formula of (t) is:
Figure BDA0003012536930000046
Figure BDA0003012536930000047
wherein the content of the first and second substances,
Figure BDA0003012536930000051
the AGV can move to the area of the area i;
Figure BDA0003012536930000052
representing the target area of area i.
Preferably, the estimate P of the density of the region iiThe formula of (t) is:
Figure BDA0003012536930000053
wherein T isiAverage waiting time for AGV in time neighborhood to pass through zone i:
Figure BDA0003012536930000054
wherein V is the number of the latest data used, if the number beta of AGV passing through the history recording area i is less than m, namely the number of data is insufficient, then V takes the value of beta, otherwise V is set as m,
Figure BDA0003012536930000055
α -th element of the V-order row vector:
Figure BDA0003012536930000056
wherein ET(i，α)For the moment alpha AGV enters zone i, IT(i，α)NF being the time when the alpha AGV enters the area i(i，α)The time it takes for the alpha AGV to completely pass through zone i.
Wherein with respect to the time unit t: the time consumed for passing through an area under the condition of no congestion is taken as a time unit, namely, t is increased by 1 every time an area is driven.
Preferably, the cost value G of the region iiThe formula of (t) is:
Gi(t)＝Pi(t)+Fi+H(r，i)
the regional value GiAnd (t) substituting into the A-algorithm to obtain an optimal region set moving from the current region to the target region.
Preferably, in step 5), the optimal path region set is output by using the a-algorithm, and the cost value g (t) of each region is used as a new estimated cost value of the a-algorithm.
Preferably, in step 5), the optimal path is planned by using a D-x algorithm, and an area path set which reaches the target area and is optimal at the current time is found.
Compared with the prior art, the invention has the beneficial effects that:
compared with the prior art, the improved global path planning method based on the network congestion diffusion model has the following advantages:
according to the improved algorithm, when a single AGV path is planned according to the characteristics of an application scene, congestion information pre-measured according to a known AGV path is introduced as a part of the path cost, so that the congestion waiting time is optimized, the AGV transportation efficiency is improved, and meanwhile, the storage space is fully utilized; the default map environment of the improved algorithm is dynamically changed, the algorithm predicts the dynamic influence of the environment by using all known AGV paths in consideration of the fact that the environment is changed in real time due to the fact that a plurality of AGVs move simultaneously, and the improved algorithm is suitable for efficient control of complex storage environments; the improved algorithm of the invention realizes prediction by using a network congestion diffusion model instead of monitoring the dynamic change of the environment in real time, and considers the dynamic influence before path planning, thereby greatly reducing the re-planning times, improving the scheduling efficiency and having remarkable advantages in time performance.
Drawings
FIG. 1 is a schematic flow chart of a multi-AGV global planning method based on a network congestion model according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a storage environment global map regionalization according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a congestion diffusion model of a warehousing network with two channels according to an embodiment of the present invention;
FIG. 4 is a modified A algorithm actual global routing graph according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a congestion diffusion model of a storage network for a single channel according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a congestion diffusion model of a multi-channel warehousing network according to an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following examples and the accompanying drawings.
Example 1
The invention provides a network congestion model-based multi-AGV global planning method, which specifically comprises the following steps of:
1) establishing a storage environment map and regionalizing, as shown in fig. 2, the storage is divided into 16 areas, and 4 areas (Q) on the right side are set4，Q4，Q12，Q16) Is a starting area and the other areas are target areas. After receiving the task request, the controller acquires the serial number r and the coordinate of each AGV in the storage environment and distributes a transportation task for each idle AGV, so that an initial area O and a target area E of a path in the current task of each idle AGV are determined, and the initial area O is set as a father node;
2) putting the father node into a close set, leading the adjacent area of the father node into an open set, and entering the step 8) if the target area is in the open set, or entering the step 3);
3) respectively obtaining distance cost H of each region in open setrWith time cost F, for a particular region i:
distance cost H(r，i)For Manhattan distance, i.e. current region i and AGVrAbsolute value of distance difference between the task target areas E. Let the coordinates of region i be (x)i，yi)，AGVrThe coordinates of the task target area E are (x)(r，E)，y(r，E)) Then H is(r，i)The formula of (1) is as follows:
H(r，i)＝|x(r，E)-xi|+|y(r，E)-yi|
time cost FiTravel time t for complete passage through zone iiExtra turn cost TN with zone iiSum, in particular, tiI.e. the number of grids passing through the area i completely, and if the AGV has no steering action during the driving process in the area i, then TNiAnd is noted as 0.
For a particular region i, the time cost FiConstant for time, the target area E of an AGV to which a task has been assigned is determined, so in a single task H(r，i)Is also a time invariant constant;
4) predicting AGV density C of each area in open set*(t) of (d). The warehousing network congestion diffusion model suitable for the method is obtained by modifying a universal network congestion diffusion model considering dynamic influence based on a Langevin diffusion model.
The whole storage environment is regarded as a network, one area of the storage environment is a node, and the driving-in and driving-out of the AGV are regarded as resource exchange among the nodes. As shown in FIG. 3, the regions are connected by two channels. The Langevin diffusion equation for the modified model for a particular region i is defined as follows:
Figure BDA0003012536930000081
wherein
Figure BDA0003012536930000082
I.e. AGV Density for region i, Ai(t) and Bi(t) are the static and dynamic parts of the Langevin diffusion equation, respectively. Specifically, Ai(t) is the difference between the number of AGVs entering the area i and the number of AGVs exiting the area i at the predetermined time t, and is defined as follows:
Figure BDA0003012536930000083
Bi(t) is the AGV exchange number of the region set Ne to which the region i is connected, and the random Gaussian white noise matrix xi with the average value of 0i(t) represents the dynamic error of the number of AGVs in zone i after multiplication. B isi(t) is specifically defined as follows:
Figure BDA0003012536930000084
wherein, Ci(t) density of the idle grid in region i;
Figure BDA0003012536930000085
the AGV density is the AGV density which is common to the area i at the time t and the area j to be driven into at the time (t + 1); ne (line of contact)iThe set of regions connected for region i, i.e. the reachable regions for all AGVs at the next time, has
Figure BDA0003012536930000086
K(i，j)(t) is calculated from the following formula:
Figure BDA0003012536930000087
wherein k isiIndicating the degree of importance of the region i, the region set Ne being connected by the region iiThe number of medium areas is related to the geographical location.
5) ObtainTaking the density estimation value P (t) of each area in the open set, and the area density estimation value P of the specific area iiThe formula of (t) is:
Figure BDA0003012536930000091
wherein
Figure BDA0003012536930000092
Is defined as in step 3), TiThe average waiting time for an AGV to pass through area i in the time neighborhood is given by:
Figure BDA0003012536930000093
wherein V is the number of the latest data used in the calculation, if the number beta of the AGV passing through the history recording area i is less than 20 and the number of the data is insufficient, the value of V is beta, otherwise, the value of V is 20,
Figure BDA0003012536930000094
α -th element of the V-order row vector:
Figure BDA0003012536930000095
wherein ET(i，α)For the moment alpha AGV enters zone i, IT(i，α)NF being the time when the alpha AGV enters the area i(i，α)Is the number of grids that the alpha AGV experiences in fully passing through zone i.
The description of the time unit t is as follows: taking the time consumed by passing through one area under the condition of no congestion as a time unit, namely increasing t by 1 every time the vehicle passes through one area;
in the global planning process, the time required for driving to the area is determined by the area direction set, as shown in fig. 4, the arrow in each area indicates the direction of the area, and the number of areas required to reach the starting area from the area along the arrow direction is t.
6) And acquiring the cost value G (t) of each region in the open set as a new estimated cost value of the A-algorithm, and realizing the multi-AGV global path planning based on congestion prediction in the storage environment by utilizing the improved A-algorithm.
Cost value G for specific region iiThe formula of (t) is:
Gi(t)＝Pi(t)+Fi+H(r，i)
7) setting the region with the minimum cost in the current open set as a new father node, moving the new father node out of the open set, and repeating the step 2);
8) and outputting the optimal path region set obtained by planning.
In this embodiment, Q4As a starting region, Q13For the target region, the global path obtained by the improved a-algorithm planning proposed by the present invention is shown in fig. 4. Simulation experiments show that the global path planning algorithm based on the network congestion diffusion model and the D-algorithm based on model predictive control have similar scheduling efficiency, but the operation time is only one tenth of the latter. The improved algorithm provided by the invention scientifically balances the scheduling effect and the operation time, and can be suitable for real-time scheduling of large-scale AGV.
Example 2
In some warehousing systems, there are obstacles that are moved frequently except AGVs, the environment is complicated and variable, and since the path of the obstacles that are moved frequently except AGVs is unpredictable, multiple times of re-planning are still needed based on the improved a-x algorithm in embodiment 1. For this case, the a-algorithm in example 1 is changed to D-algorithm, but still improved by using the network congestion diffusion model. In step 1) in embodiment 1), instead of setting the start area O as a parent node, the target area E is set as a parent node; step 2) instead, a father node is put into a close set, an adjacent area of the father node is led into an open set, if an initial area is in the open set, the step 8) is carried out, and if not, the step 3) is carried out; when obstacles except the AGV enter the storage system, the improved D-algorithm can continuously utilize the previously calculated region path set from the target region to the obstacle appearing region, and a new path for avoiding the obstacle can be obtained only by replanning the region path set from the current region to the obstacle avoiding region of the AGV, so that the planning time is saved.
Cost value G of region iiThe calculation procedure of (t) was the same as in example 1. When all the AGVs are in the warehousing system, when the AGV paths are planned, if the cost value of a certain area is detected to be too high, the D-algorithm is used for re-planning, and the optimal area path set reaching the target area at the current time is found. Although the complexity of the algorithm is higher than that of the algorithm in embodiment 1, the improved D-x algorithm can ensure that all the time of the area path set of the AGV is optimal, and has a better scheduling effect. Through the improvement, the multi-AGV global planning method based on the network congestion model can be applied to the situation that the unpredictable obstacles in the warehouse move frequently, and even if the unpredictable obstacles in the warehouse move frequently, the method can still efficiently complete the scheduling task of large-scale AGV.
Example 3
In some warehousing systems, the AGV may only be able to pass on a single pass if certain areas are limited by the surrounding environment, as shown in fig. 5. The AGVs can only move from area 6 to area 5, but the AGVs within area 5 cannot travel to area 6. For this case, when the network congestion propagation model is applied, the area set Ne connected to the area 5 is used5Albeit region 1, region 6 and region 9. But in the calculation of the static and dynamic part A of the Langevin diffusion equationi(t) and Bi(t), since some paths are single-channel, the formula is changed to the following form:
Figure BDA0003012536930000111
Figure BDA0003012536930000112
wherein the content of the first and second substances,
Figure BDA0003012536930000113
the AGV of the inner zone may move to zone i. Taking region 9 as an example, it
Figure BDA0003012536930000114
Region 5 and region 13.
Figure BDA0003012536930000115
The inner regions may all be target regions of region i, exemplified by region 5
Figure BDA0003012536930000116
Region 1 and region 9.
Cost value G of subsequent region iiThe calculation procedure of (t) was the same as in example 1. Through the improvement, the multi-AGV global planning method based on the network congestion model can be applied to the condition that the warehouse passage is narrow, and even if the AGV can only move in a single passage in some areas, the method can still efficiently finish the scheduling task of the large-scale AGV.
Example 4
In some warehousing systems, to maximize the utilization of AGV agility to improve the operating efficiency of the warehousing system, AGVs are free to travel from the current area to any area adjacent to it. As shown in FIG. 6, the AGV can move from the current area to any area nearby. For this case, when the network congestion diffusion model is applied, the area 6 is taken as an example, and the area set Ne thereof6Extended to Q1、Q2、Q3、Q5、Q7、Q13、Q14And Q15. In obtaining the static and dynamic part A of the Langevin diffusion equationi(t) and Bi(t) the same procedure as in example 1 was repeated, except that the number of regions in the region set was increased.
Cost value G of subsequent region iiThe calculation procedure of (t) was still the same as in example 1. Through the improvement, the network congestion model-based multi-AGV global planning method provided by the invention can be applied to the condition that the AGV moves flexibly, even if the AGV can move to any position at presentAnd the method can still efficiently finish the scheduling task of the large-scale AGV by arranging adjacent areas.
The above embodiments are only for explaining the details to help understanding the technical solution of the present invention, and it is obvious to those skilled in the art that any modifications and substitutions made without departing from the principle of the present invention belong to the protection scope of the present invention.

Claims (10)

1. A multi-AGV global planning method based on a network congestion model is characterized by comprising the following steps:
1) establishing a regional storage environment map, and distributing transportation tasks for each idle AGV;
2) respectively acquiring distance cost and time cost of each region;
3) predicting the AGV density of each area, and establishing an improved network congestion diffusion model according to the scheduling of the storage system;
4) acquiring a density estimated value of each area according to the AGV density of each area;
5) and acquiring the cost value of each region and outputting an optimal path region set.
2. The method for global planning of multiple AGVs based on network congestion model according to claim 1, wherein step 1) comprises: and processing the global image of the storage environment to acquire the serial number r and the coordinate of each AGV, and distributing a transportation task for each idle AGV, so as to determine the starting area O and the target area E of the path in the current task of each idle AGV.
3. The method as claimed in claim 2, wherein before the distance cost and the time cost of each region are obtained, an initial region O is established and set as a parent node, the parent node is placed in a close set, an adjacent region of the parent node is led into an open set, if the target region is in the open set, the step 5) is performed, and if the target region is not in the open set, the step 3) is performed.
4. A network congestion based on claim 1 or 2The multi-AGV global planning method of the model is characterized in that the distance cost H of the region i(r，i)For Manhattan distance, i.e. current region i and AGVrAbsolute value of distance difference between task target areas E; let the coordinates of region i be (x)i，yi)，AGVrThe coordinates of the task target area E are (x)(r，E)，y(r，E)) Then the distance cost H of the region i(r，i)The formula of (1) is as follows:
H(r，i)＝|x(r，E)-xi|+|y(r，E)-yi|
time cost F of region iiTravel time t for complete passage through zone iiExtra turn cost TN with zone iiSum, in particular, tiTime taken to completely pass region i if AGVrNo steering behavior during driving in zone i, TMiIs marked as 0;
for region i, time cost FiApparently a time-invariant constant, and the target area E of an AGV to which a task has been assigned is determined, so H is in a single task(r，i)Also a time invariant constant.
5. The method as claimed in claim 4, wherein the improved network congestion diffusion model is a generalized network congestion diffusion model based on a Langevin diffusion model and considering dynamic influence, the improved network congestion diffusion model regards the entire warehousing environment as a network, one area of the warehousing environment is a node, the inbound and outbound of AGVs are regarded as resource exchange between nodes, and the Langevin diffusion equation of the improved network congestion diffusion model for a specific area i is defined as follows:
Figure FDA0003012536920000021
wherein
Figure FDA0003012536920000022
AGV Density for region i, Ai(t) and Bi(t) static and dynamic parts of the Langevin diffusion equation, Ai(t) is the difference between the number of AGVs entering the area i and the number of AGVs exiting the area i at the predetermined time t, and is defined as follows:
Figure FDA0003012536920000023
Bi(t) is the AGV exchange number of the region set Ne to which the region i is connected, and the random Gaussian white noise matrix xi with the average value of 0i(t) the dynamic error, B, representing the number of AGVs in region i after multiplicationi(t) is specifically defined as follows:
Figure FDA0003012536920000024
wherein, Ci(t) density of free space resources in region i, Cj(t) is the density of idle space resources in region j;
Figure FDA0003012536920000031
the AGV density is the AGV density which is common to the area i at the time t and the area j to be driven into at the time (t + 1); ne (line of contact)iThe set of regions connected for region i, i.e. the reachable regions for all AGVs at the next time, has
Figure FDA0003012536920000032
K(i，j)(t) is calculated from the following formula:
Figure FDA0003012536920000033
wherein k isiIndicating the degree of importance of the region i, the region set Ne being connected by the region iiThe number of middle zones is related to the geographic location; k(j，i)(t) represents the proportion of area i where the AGV can smoothly move to area j at time t,the method is used for evaluating the congestion degree of one area, and if the AGV moves to the area j completely, the value is 1; if the area is congested, the value is less than 1.
6. The method of claim 5, wherein A is a single channel for the path, and A is a global planning method for multiple AGVs based on the network congestion modeli(t) and BiThe formula of (t) is:
Figure FDA0003012536920000034
Figure FDA0003012536920000035
wherein the content of the first and second substances,
Figure FDA0003012536920000036
the AGV can move to the area of the area i;
Figure FDA0003012536920000037
representing the target area of area i.
7. The method of claim 6, wherein the estimated value P of the density of the area i is a global planning value for multiple AGVs based on the network congestion modeliThe formula of (t) is:
Figure FDA0003012536920000038
wherein T isiAverage waiting time for AGV in time neighborhood to pass through zone i:
Figure FDA0003012536920000039
wherein V is useIf the number of the AGV passes through the history recording area i is less than m, namely the number of the data is less than m, the value of V is beta, otherwise, the value of V is m,
Figure FDA00030125369200000310
α -th element of the V-order row vector:
Figure FDA0003012536920000041
wherein ET(i，α)For the moment alpha AGV enters zone i, IT(i，α)NF being the time when the alpha AGV enters the area i(i，α)The time it takes for the alpha AGV to completely pass through zone i.
Wherein with respect to the time unit t: the time consumed for passing through an area under the condition of no congestion is taken as a time unit, namely, t is increased by 1 every time an area is driven.
8. The method of claim 7, wherein the cost value G of the area i is a cost value of multiple AGVsiThe formula of (t) is:
Gi(t)＝Pi(t)+Fi+H(r，i)
the regional value GiAnd (t) substituting into the A-algorithm to obtain an optimal region set moving from the current region to the target region.
9. The method according to claim 8, wherein step 5) is to output an optimal path region set by using an a-algorithm, and the cost value g (t) of each region is used as a new estimated cost value of the a-algorithm.
10. The method according to claim 8, wherein in step 5), the optimal path is planned by using a D-x algorithm, and an area path set which reaches the target area and is optimal at the current time is found.
CN202110379885.0A 2021-04-08 2021-04-08 Multi-AGV global planning method based on network congestion model Active CN113516429B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110379885.0A CN113516429B (en) 2021-04-08 2021-04-08 Multi-AGV global planning method based on network congestion model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110379885.0A CN113516429B (en) 2021-04-08 2021-04-08 Multi-AGV global planning method based on network congestion model

Publications (2)

Publication Number Publication Date
CN113516429A true CN113516429A (en) 2021-10-19
CN113516429B CN113516429B (en) 2024-03-29

Family

ID=78062652

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110379885.0A Active CN113516429B (en) 2021-04-08 2021-04-08 Multi-AGV global planning method based on network congestion model

Country Status (1)

Country Link
CN (1) CN113516429B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114967711A (en) * 2022-07-04 2022-08-30 江苏集萃清联智控科技有限公司 Multi-AGV collaborative path planning method and system based on dynamic weighting map
CN115646818A (en) * 2022-12-28 2023-01-31 江苏智联天地科技有限公司 AGV intelligence letter sorting system
CN116360378A (en) * 2023-06-02 2023-06-30 北京中鼎昊硕科技有限责任公司 AGV trolley safety scheduling method based on data analysis

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016169290A1 (en) * 2015-04-21 2016-10-27 华南理工大学 Decision-making supporting system and method oriented towards emergency disposal of road traffic accidents
CN109242179A (en) * 2018-08-31 2019-01-18 心怡科技股份有限公司 A kind of intelligent dispatching algorithm based on flow control
CN109544922A (en) * 2018-11-27 2019-03-29 华南理工大学 A kind of traffic network Distributed Predictive Control method based on region division
CN109839935A (en) * 2019-02-28 2019-06-04 华东师范大学 The paths planning method and equipment of more AGV
CN110780671A (en) * 2019-10-30 2020-02-11 华南理工大学 Storage navigation intelligent vehicle scheduling method based on global vision
WO2020206798A1 (en) * 2019-04-08 2020-10-15 中国电子科技集团公司第二十八研究所 Method for assessing degree of route blockage in convective weather using optimal traversal path
CN111932000A (en) * 2020-07-28 2020-11-13 太原科技大学 Multi-AGV (automatic guided vehicle) scheduling method applied to large logistics system
CN112149555A (en) * 2020-08-26 2020-12-29 华南理工大学 Multi-storage AGV tracking method based on global vision
CN112444256A (en) * 2019-08-27 2021-03-05 扬州盛世云信息科技有限公司 Method for time shortest path based on road traffic flow

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016169290A1 (en) * 2015-04-21 2016-10-27 华南理工大学 Decision-making supporting system and method oriented towards emergency disposal of road traffic accidents
CN109242179A (en) * 2018-08-31 2019-01-18 心怡科技股份有限公司 A kind of intelligent dispatching algorithm based on flow control
CN109544922A (en) * 2018-11-27 2019-03-29 华南理工大学 A kind of traffic network Distributed Predictive Control method based on region division
CN109839935A (en) * 2019-02-28 2019-06-04 华东师范大学 The paths planning method and equipment of more AGV
WO2020206798A1 (en) * 2019-04-08 2020-10-15 中国电子科技集团公司第二十八研究所 Method for assessing degree of route blockage in convective weather using optimal traversal path
CN112444256A (en) * 2019-08-27 2021-03-05 扬州盛世云信息科技有限公司 Method for time shortest path based on road traffic flow
CN110780671A (en) * 2019-10-30 2020-02-11 华南理工大学 Storage navigation intelligent vehicle scheduling method based on global vision
CN111932000A (en) * 2020-07-28 2020-11-13 太原科技大学 Multi-AGV (automatic guided vehicle) scheduling method applied to large logistics system
CN112149555A (en) * 2020-08-26 2020-12-29 华南理工大学 Multi-storage AGV tracking method based on global vision

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114967711A (en) * 2022-07-04 2022-08-30 江苏集萃清联智控科技有限公司 Multi-AGV collaborative path planning method and system based on dynamic weighting map
CN115646818A (en) * 2022-12-28 2023-01-31 江苏智联天地科技有限公司 AGV intelligence letter sorting system
CN116360378A (en) * 2023-06-02 2023-06-30 北京中鼎昊硕科技有限责任公司 AGV trolley safety scheduling method based on data analysis
CN116360378B (en) * 2023-06-02 2023-09-19 北京中鼎昊硕科技有限责任公司 AGV trolley safety scheduling method based on data analysis

Also Published As

Publication number Publication date
CN113516429B (en) 2024-03-29

Similar Documents

Publication Publication Date Title
CN113516429A (en) Multi-AGV global planning method based on network congestion model
Hwang et al. Q (λ) learning-based dynamic route guidance algorithm for overhead hoist transport systems in semiconductor fabs
Lyu et al. Approach to integrated scheduling problems considering optimal number of automated guided vehicles and conflict-free routing in flexible manufacturing systems
Jamrus et al. Hybrid particle swarm optimization combined with genetic operators for flexible job-shop scheduling under uncertain processing time for semiconductor manufacturing
Fan et al. An improved genetic algorithm for flexible job shop scheduling problem considering reconfigurable machine tools with limited auxiliary modules
Guo et al. Decomposition-based classified ant colony optimization algorithm for scheduling semiconductor wafer fabrication system
Dang et al. Scheduling of mobile robots for transportation and manufacturing tasks
WO2019141220A1 (en) Mobile robot path planning method and system
Majd et al. Integrating learning, optimization, and prediction for efficient navigation of swarms of drones
WO2019141223A1 (en) Path planning method and system for mobile robot
Finke et al. Stable task load balancing strategies for cooperative control of networked autonomous air vehicles
Le et al. Fast reactive scheduling to minimize tardiness penalty and energy cost under power consumption uncertainties
CN109709985A (en) A kind of unmanned plane task optimization method, apparatus and system
Zhang et al. Dynamic fleet management with rewriting deep reinforcement learning
Huang et al. A novel priority dispatch rule generation method based on graph neural network and reinforcement learning for distributed job-shop scheduling
Samra et al. A linear time and space algorithm for optimal traffic-signal duration at an intersection
Qin et al. Dynamic dispatching for interbay automated material handling with lot targeting using improved parallel multiple-objective genetic algorithm
Ferreira et al. Distributed allocation and scheduling of tasks with cross-schedule dependencies for heterogeneous multi-robot teams
Schmidt et al. Research on decentralized control strategies for automated vehicle-based in-house Transport systems–a survey
Liu et al. A new knowledge-guided multi-objective optimisation for the multi-AGV dispatching problem in dynamic production environments
Zaidi et al. Optimisation and simulation of transportation tasks in flexible job shop with muti-robot systems
Reveliotis et al. Min-time coverage in constricted environments: Problem formulations and complexity analysis
Lin et al. A multi-fidelity model approach for simultaneous scheduling of machines and vehicles in flexible manufacturing systems
Park et al. Design of Experiment and Analysis Method for the Integrated Logistics System Using Orthogonal Array
Haiming et al. Algorithm of path planning based on time window for multiple mobile robots in warehousing system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant