CN113359702B - Intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search - Google Patents
Intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search Download PDFInfo
- Publication number
- CN113359702B CN113359702B CN202110510257.1A CN202110510257A CN113359702B CN 113359702 B CN113359702 B CN 113359702B CN 202110510257 A CN202110510257 A CN 202110510257A CN 113359702 B CN113359702 B CN 113359702B
- Authority
- CN
- China
- Prior art keywords
- agv
- solution
- picking
- search
- population
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000005457 optimization Methods 0.000 title claims abstract description 19
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 19
- 238000009826 distribution Methods 0.000 claims abstract description 6
- 238000003860 storage Methods 0.000 claims description 14
- 230000001174 ascending effect Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims 1
- 241000531116 Blitum bonus-henricus Species 0.000 description 3
- 235000008645 Chenopodium bonus henricus Nutrition 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 239000002245 particle Substances 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000010410 layer Substances 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
-
- 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
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0219—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory ensuring the processing of the whole working surface
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
An optimized dispatching method for AGV operation of an intelligent warehouse based on water wave optimization-tabu search not only considers the planning of the sorting paths of the AGVs, but also considers the distribution of the sorting tasks among a plurality of the AGVs; the method comprises the steps that a grid method is adopted to describe a warehouse environment, a simple water wave optimization algorithm is used as a main algorithm to distribute a picking subset for the AGV, a propagation and wave breaking operator uses a local search operation based on bit reversal, and the propagation and wave following operation well combines global search and local search, so that the global search capability of the algorithm is guaranteed, and the search is prevented from falling into local optimum by a fixed degree; the sub-algorithm for planning the picking path adopts tabu search; the method introduces a collision avoidance strategy, can better minimize the maximum goods picking completion time of all the AGV and avoid collision in the goods picking process. The invention reasonably schedules a plurality of AGV in the warehousing operation, improves the warehouse operation efficiency, shortens the maximum goods picking time and improves the operation efficiency of the warehousing system.
Description
Technical Field
The invention relates to an intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search.
Background
The rapid development of the internet has prompted the emergence of new sales modes such as online retail and the like, so that the order transaction amount faced by the warehouse logistics industry is heavier and heavier nowadays, the warehouse is larger in scale and more complex in roadway structure than the past, and the picking operation is harder and harder. The traditional warehouse adopting the manual management mode consumes labor cost and operation time, cannot meet the operation requirements in the current warehousing system, and can greatly reduce the warehousing operation efficiency. The picking operation is the core of the warehousing operation, occupies about 55-75% of the total operation cost of the warehouse, and in the picking operation, the running time of goods transported between the warehouses occupies more than 50% of the total picking time, so the key point of the optimization picking operation lies in optimizing the transportation time of the goods in the warehouse, namely optimizing the picking path, thereby reducing the picking distance and shortening the picking time. Therefore, appropriate methods are needed to address the pick path planning problem in warehouses.
The intelligent storage robot represented by the AGV (automatic Guided vehicle) has the characteristics of high automation degree, sensitivity and the like, can flexibly change a driving path according to the requirement of a goods taking position, and has low requirement on the site environment; on the other hand, the AGV can enable the warehousing system to efficiently finish the goods picking operation under the condition that people do not need to directly participate, and labor and time cost is reduced. Thus, more and more warehouses are introducing intelligent warehouses of AGVs. In the face of heavy workload, multiple AGVs are usually required to complete picking operation cooperatively in a warehouse, dispatching of the AGVs is very complicated under the conditions of large warehouses and a large amount of goods, and the dispatching efficiency of the AGVs is very important for the warehousing operation. Pinkam et al plan an optimal picking route for an AGV using a greedy algorithm for picking in a warehouse, and use Dijkstra's algorithm for path planning and obstacle avoidance, and use local search to find the nearest target goods. Vivaldini et al introduced three key routines in their study that determine AGV behavior in 2010: the method comprises a Dijkstra-based shortest path method for calculating a global path, an A-algorithm for local path planning and an AGV automatic positioning algorithm based on an extended Kalman filter. In 2016, Vivaldini et al further proposed an estimation method to determine the number of AGVs required to execute a given transport order within a specific time window, comparing the shortest job priority algorithm with the Tabu Search (TS) algorithm in the task allocation problem, and planning the path using Dijkstra shortest path method. Tang 26107An et al solved the AGV path planning problem in automated stereoscopic warehouse using a modified PSO (particle Swarm optimization) algorithm, which introduces a Gaussian operator to mutate the particles and prevents the search from falling into local optima. In the methods, the Dijkstra algorithm is mainly characterized in that the Dijkstra algorithm is expanded outwards from a starting point to all possible nodes layer by layer until a target node is reached, the solution searched by the algorithm is possibly not optimal, and the searching speed is low; the algorithm efficiency is reduced when the number of the path nodes is large, because the calculation amount of the algorithm on each path node is large, the searching speed is low, the searching range is small, and the obtained solution is not the optimal solution; the PSO algorithm is searched in a solution space by a group of random individuals, and the particles are self-adjusted in the solution space according to self historical experiences and experiences of other examples, so that the whole population is continuously close to the optimal solution, Tougo 26107, and the like do not consider the collision avoidance problem of the AGV. In addition, when the problem of scheduling multiple AGVs is considered, it needs to consider that the picking tasks are reasonably distributed to each AGV and the picking paths are planned for the AGVs.
Disclosure of Invention
In order to overcome the defects, the method decomposes the problem into two parts of goods distribution and path planning, wherein a simple Water Wave Optimization algorithm (S-WWO) is used for iteratively evolving a goods distribution solution, a propagation operation and a Wave breaking operation are used for teaching that the whole local search and the local search are well coordinated, and a population reduction strategy is adopted for accelerating algorithm convergence to avoid the algorithm convergence from falling into local optimum; the route planning strategy uses a TS algorithm, an optimal picking route is iteratively searched through neighborhood search, the solution is allowed to be deteriorated in a certain degree in the neighborhood search, so that the search jumps out of local optimization, a new search area is searched, a taboo table is used for avoiding the search from being trapped in endless loop, and the search efficiency is improved; the collision avoidance strategy considers the collision problem which may occur in the goods picking operation of the multiple AGVs, and based on the backtracking idea, a collision-free goods picking route is planned for the AGVs. The method can reasonably distribute the picking operation to each AGV, and considers the problem of possible conflict among the AGVs, and aims to reasonably schedule the AGVs in the warehousing operation, improve the warehouse operation efficiency, shorten the maximum picking time and improve the operation efficiency of the warehousing system.
The technical scheme adopted for realizing the purpose of the invention is as follows:
an intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search comprises the following steps:
step 2, constructing a warehouse environment model by using a grid method, namely establishing a coordinate system based on warehouse entry coordinates, wherein storage positions are all located at grid intersections, and an AGV runs in a goods space roadway;
the entrance and the exit of the warehouse are not positioned at the same position, but positioned at two ends of the diagonal line of the warehouse respectively, so that the AGV can directly leave from the exit of the warehouse without returning to the starting point after completing the picking task; each shelf is a single layer, and each shelf only stores one type of goods;
step 3, randomly generating a group of initial solution population P, wherein each individual in the population represents the distribution condition of all cargos by a solution X of a simple water wave optimization main algorithm distributed for the cargos, and the X is represented by a m multiplied by n dimensional 0-1 vector;
for each cargo, if it is demanded quantity q i If the capacity Q of the AGVs is exceeded, the goods are required to be picked and sent by the AGVs, each AGV needs to be fully loaded with the goods, and the goods surplus is reserved for other AGVs to be picked and sent until the surplus of the goods is smaller than the capacity of the AGVs; for those full AGVs, their task is to load the load and then transport it directly to the warehouse exit, thus simplifying the problem by performing the following steps for each load whose demand exceeds the capacity of the AGV:
3.1) making
q′ i The remainder is the remaining amount after the goods are distributed to the AGV;
3.2) q' i When the load is 0, arranging k AGVs to be respectively full of the load, and removing the AGVs from all the AGV sets and removing the load from the load set when subsequently considering the dispatching of a plurality of AGV jobs;
3.3) otherwise, arranging k-1 AGVs to be respectively full of the load, removing the AGVs from all the AGVs in the set when dispatching the AGVs subsequently, and enabling q to be used i ＝q′ i ；
After the simplification process, the capacity of each cargo cannot exceed the capacity of the AGV in a batch of picking tasks;
each solution X in the population needs to satisfy the following constraint conditions:
each load is assigned to only one AGV, i.e. there is no intersection between the subsets of loads assigned to all AGVs, defined as:
wherein, X j And X j′ Respectively, the goods subsets distributed to the jth and jth AGV, wherein m is the number of AGV;
all picking operations need to be completed completely, namely, the goods collection A cannot have omission, and the omission is expressed as:
wherein, A is all goods set in one picking task, m is AGV quantity, X j A subset of the loads assigned to each AGV;
for each AGV, the total demand for the goods allocated to it needs to meet its capacity constraint, expressed as:
wherein n is j For the number of goods allocated to the jth robot,
for a subset X of goods assigned to the robot j In the specification, Q is the capacity of the AGV, and the required quantity of the ith goods is the capacity of the AGV;
and 4, planning a picking sequence for each AGV by using a TS sub-algorithm based on each solution generated in the initial population, wherein the method comprises the following steps of:
step 4.1, deriving an instance for each subproblem based on X, namely determining goods distributed to each AGV;
step 4.2, setting a null taboo table tabu and setting a taboo length TabuLen;
4.3, generating a picking path for the AGV by using a greedy method to serve as an initial solution y;
4.4, if the termination condition is met, returning to the optimal picking route, and ending the algorithm;
step 4.5, randomly selecting a point in the initial solution y as a tabu object d best ；
Step 4.6, perform neighborhood search on y, generate NbSize new solutions, and update the best candidate solution y' and the best move d best The neighborhood search method used here generates a new solution for two adjacent points in the random exchange path each time, and updates the optimal candidate solution by using the best improved solution priority strategy, if the optimal candidate solution generated by neighborhood search is not superior to the current solution, the optimal candidate solution is also used to replace the current solution, although the quality of the solution is poor, the solution can be prevented from falling into local optimum, and the search process can be balanced;
step 4.7, if the final best candidate solution y' is better than y or
The following steps are carried out:
step 4.7.1, mixing d best Adding the tail of the tab;
step 4.7.2, if the tabu.length is larger than TabuLen, removing the taboo object added with the tabu at the earliest time;
step 4.7.3, using y' to update y;
step 4.8, if y is superior to the current global optimal solution y * Updating y using come y * ；
And 5, planning conflict-free picking routes for all AGVs by using a collision avoidance strategy for each solution in the population, and comprising the following steps of:
step 5.1, generating a detailed route set S of all AGVs by using a row-first method;
the row-first method, i.e. when an AGV is going to travel from (x, y) to (x ', y'), which first travels from (x, y) to (x ', y) and then from (x', y) to (x ', y'), the column-first method is the reverse of the previous method, but both methods enable the AGV to reach the target point from the current position, where we use the row-first method to generate the initial position for all AGVsInitial detailed route set S ═ S 1 ，S 2 ，...，S m }。
Step 5.2, finding out all collision points, namely the points where the position and the time of the AGV conflict, and sequencing the points according to the ascending order of the time;
step 5.3, for each collision point c and for the two AGVs colliding therewith, the following steps are performed:
step 5.3.1, according to the two AGV closest target points, selecting one AGV as b according to the following rules:
1) the target points are positioned in the same column, and the AGV with the larger number of the target point rows is selected as b;
2) the target points are located in the same row, and an AGV is randomly selected as b;
3) otherwise, selecting the AGV with the smaller column where the target point is positioned as b;
step 5.3.2, order
The intersection which is before c and is closest to c in the detailed path of b;
step 5.3.3, if b is
Changing to a column-first mode to drive without collision at the next point, replacing the original path segment, and updating the detailed path of b;
step 5.3.4, otherwise, continuously backtracking the previous intersection, judging whether the path segment can be updated in a column-first mode, and if the path segment can be updated in the final mode
If the collision is still generated when the starting point p of the b is traced back, the standby route is not changed, one of the two AGVs with shorter picking completion time is selected to wait for the other AGV to pass through, then the operation is continued, and the picking completion time of the AGV is updated;
step 6, calculating the fitness function value of each solution X in the population, comprising the following steps:
step 6.1, calculating the maximum picking completion time of each AGVInter T j The AGV comprises the goods picking travel time and the goods picking time of the AGV;
step 6.2, calculating fitness function values of the solutions X, wherein the fitness function value of each solution is the maximum goods picking completion time T of all the AGV, and comprises goods picking time and running time; since for each AGV, the total demand for the goods allocated to it needs to meet its capacity constraint, in calculating the fitness of each solution, the constraint violation is added to the objective function in the encoding, defined as:
wherein, T j The maximum pick completion time for the jth AGV, and P is a very large positive integer used as a constraint capacity violation, x ji Indicates whether the ith load is assigned to the jth AGV, q i The required quantity of the ith goods is Q, and the capacity of the AGV is Q;
step 7, evolving a solution of the main algorithm using the S-WWO algorithm, comprising the steps of:
step 7.1, for each solution X in the population P, the following operations are performed:
step 7.1.1, a propagation operation is performed on solution X, i.e. r local searches based on bit inversion are performed randomly to generate a new solution X', where r ═ rand (1, λ) X )，λ X Is the wavelength of solution X;
step 7.1.2, for each newly found optimal solution X * Performing a wave breaking operation, namely performing neighborhood search on the wave breaking operation to generate k neighborhood solutions, where the neighborhood search is also performed by using local search in the wave breaking operation, and in addition, the value of k can be expressed as follows:
wherein n is the number of goods, k max Is a predefined parameter, epsilon is a very small positive integer avoiding zero;
and to propagation operatorsLocal search and neighborhood search operations in the wave breaking operator are modified, the modified search operations are based on bit reversal, and one element X is randomly selected from the solution X each time ji (j is more than or equal to 1 and less than or equal to m, i is more than or equal to 1 and less than or equal to n), and the value is modified into x ji ＝1-x ji And then randomly selecting an element X from X j′i (j ≠ j'), if it has a value corresponding to y ji If the original value is reversed, the value is modified to x j′i ＝1-x j′i ；
Step 7.1.3, update the optimal solution X * ；
Step 7.2, updating the population scale, and removing the worst solution in the population if the population scale is reduced;
reference is made here to the Population reduction strategy mentioned in the group Size Variable Water Wave Optimization algorithm (Water Wave Optimization with Variable Size amplification, V-WWO) by zhangjeff et al, expressed as:
wherein, NP max And NP min Maximum and minimum size of population, g and g respectively max Current and maximum allowed iteration times (or fitness calculation times) respectively;
and 8, if the termination condition is met, returning the maximum completion time of all the AGVs, finishing the algorithm, otherwise, returning to the step 4.
The beneficial effects of the invention are as follows: a plurality of AGV in the operation of rationally dispatching storage improve the warehouse operating efficiency, shorten the biggest time of choosing goods, promote storage system operating efficiency.
Drawings
Fig. 1 is a schematic diagram of a warehouse environment.
Fig. 2 is an overall flow chart of the hybrid algorithm of the present invention.
FIG. 3 is a detailed flow chart of an AGV cargo allocation algorithm based on S-WWO
Fig. 4 is a detailed flowchart of the algorithm for AGV path planning based on TSA.
Detailed Description
The technical scheme of the invention is further explained by combining the attached drawings.
Referring to fig. 1 to 4, a method for optimizing and scheduling an AGV operation in an intelligent warehouse based on water wave optimization-tabu search includes the following steps:
assume that there is a group of n items in a picking job: a ═ A 1 ，A 2 ，...，A n There are a set of m AGVs in the warehouse for picking: b ═ B 1 ，B 2 ，...，B m }. The problem of scheduling multiple AGVs is that the group of goods is distributed to all the AGVs, an optimal picking route is planned for each AGV, and the jth AGV starts from the starting point p j Starting (j is more than or equal to 1 and less than or equal to m), picking all goods distributed to the goods, and then leaving the warehouse from an outlet;
step 2, constructing a warehouse environment model by using a grid method, namely establishing a coordinate system based on warehouse entry coordinates, wherein storage positions are all located at grid intersections;
and 3, randomly generating a group of initial solution population P, wherein each individual in the population represents the distribution condition of all goods by using the solution X of the simple water wave optimization main algorithm for distributing the goods, and X is { X ═ X } 1 ，X 2 ，...，X m Expressed using m × n-dimensional 0-1 vectors;
the subset of loads assigned to the jth AGV may be represented as
Wherein when x ji When 1, the item A is represented i Assigned to the jth AGV, otherwise x ji ＝0；
Step 4, based on each solution generated in the initial population, a picking sequence is planned for each AGV using the TS sub-algorithm as shown in FIG. 4
Comprises the following stepsThe following steps:
step 4.1, deriving an instance for each subproblem based on X, namely determining goods distributed to each AGV;
step 4.2, setting a null taboo table tabu and setting a taboo length TabuLen;
4.3, generating a picking path for the AGV by using a greedy method to serve as an initial solution y;
4.4, if the termination condition is met, returning to the optimal picking route, and ending the algorithm;
step 4.5, randomly selecting a point in the initial solution y as a tabu object d best ；
Step 4.6, perform neighborhood search on y to generate NbSize new solutions, and update the best candidate solution y' and the best move d best ；
Step 4.7, if the final best candidate solution y' is better than y or
The following steps are carried out:
step 4.7.1, mixing d best Adding the tail of the tab;
step 4.7.2, if the tabu.length is larger than TabuLen, removing the taboo object added with the tabu at the earliest time;
step 4.7.3, using y' to update y;
step 4.8, if y is better than the current global optimal solution y * Updating y using come y * ；
4.9, planning and obtaining a current optimal picking path for each AGV;
during the neighborhood search, we use the best improved solution precedence strategy to update the current best candidate solution y ', i.e., if each neighborhood move yields a new solution y "that is better than the current best candidate solution y', y" is used to update y ', and d whose movement can improve y' is used to update the tabu object d best (ii) a Otherwise y' is retained. After the neighborhood search operation is executed, if the optimal candidate solution y 'is superior to the current solution y, and the taboo object d corresponding to the y' is best If not in tabu, the new d best Add the end of the TabuAnd updates the current solution y with y' as the start of the next iteration of the algorithm. It should be noted here that if the length of the tabu in the tabu table exceeds the set maximum tabu length tabu len, the first element in the tabu is removed, and if the optimal candidate solution y 'generated by the neighborhood search is not superior to y, the current solution y is also replaced by the optimal candidate solution y', which may cause the quality of the solution to be poor, but may prevent the solution from falling into local optimality, and balance the search process;
finally, the global optimal solution y is updated using the current solution y * In the whole process, the sub-algorithm iteratively executes neighborhood search operation on the current solution, updates the tabu table until a termination condition is met, and plans an optimal picking path with the picking travel distance as small as possible for the AGV;
and 5, planning conflict-free picking routes for all AGVs by using a collision avoidance strategy for each solution in the population, and comprising the following steps of:
step 5.1, generating a detailed route set S of all AGVs by using a row-first method;
step 5.2, finding out all collision points, and sequencing the points according to the ascending order of time;
step 5.3, for each collision point c and for the two AGVs colliding therewith, the following steps are performed:
step 5.3.1, according to the two AGV closest target points, selecting one AGV as b according to the following rules:
1) the target points are positioned in the same column, and the AGV with the larger number of the target point rows is selected as b;
2) the target points are positioned in the same row, and an AGV is randomly selected as a b;
3) otherwise, selecting the AGV with the smaller column where the target point is positioned as b;
step 5.3.2, order
The intersection which is before c and is closest to c in the detailed path of b;
step 5.3.3, if b is
Changing to a column-first mode to drive without collision at the next point, replacing the original path segment, and updating the detailed path of b;
step 5.3.4, otherwise, continuously backtracking the previous intersection, judging whether the path segment can be updated in a column-first mode, and if the path segment can be updated in the final mode
If the collision still occurs when the starting point p of the second AGV is traced back to the starting point p of the second AGV, the standby route is not changed, the AGV is selected to continue the operation after the other AGV passes the standby route, and the sorting completion time of the AGV is updated;
step 6, calculating the fitness function value of each solution X in the population, and comprising the following steps of:
step 6.1, calculating the maximum picking completion time T of each AGV j The AGV comprises the goods picking driving time and the goods picking time of the AGV;
for each AGV, it arrives and departs from the pick route Y j The time of the first storage location may be expressed as:
wherein, t j (1) Indicating the time when the jth AGV arrives at the first lot,
indicates the starting point of the AGV
From the first storage position y j1 V is the AGV speed; t' j Indicating the time the AGV left the first slot,
indicating the amount of goods taken at the first storage location,
representing the time taken to take each unit of the good;
for pick-up route Y j Other positions in (1 < k ≦ n) that the AGV arrives and departs from j ) The time of each cargo space is respectively as follows:
t j (k)＝t′ j (k-1)+d(y jk-1 ，y jk )/v
wherein, t j (k) Represents the time, t ', that the jth AGV arrived at the kth slot' j (k-1) represents the time of departure from the k-1 st cargo space, d (y) jk-1 ，y jk ) Indicating the distance from the kth-1 to the kth cargo space of the AGV, and v is the AGV speed; t' j (k) Indicating the time the AGV left the kth slot,
indicating the quantity of the goods taken at the kth slot,
representing the time taken to take each unit of the good;
and finally, the AGV finishes the picking task, and the time of leaving the warehouse is as follows:
wherein n is j Is the amount of cargo, t ', allocated to the jth AGV' j (n j ) To leave the n-th j The time of each cargo space is determined,
is the n-th j The distance from each cargo space to the warehouse exit, and v is the AGV speed;
step 6.2, calculating fitness function values of the solutions X, wherein the fitness function value of each solution is the maximum picking completion time T of all the AGV;
for each AGV, the total demand for the goods assigned to it needs to meet its capacity constraint, so the constraint violation is added to the objective function in the code, defined as:
wherein, T j The maximum pick completion time for the jth AGV, and P is a very large positive integer used to constrain capacity violations, x ji Indicates whether the ith load is assigned to the jth AGV, q i The required quantity of the ith goods is Q, and the capacity of the AGV is Q;
step 7, using the S-WWO algorithm shown in fig. 3 to evolve the main algorithm, comprising the steps of:
step 7.1, for each solution X in the population P, the following operations are performed:
step 7.1.1, a propagation operation is performed on solution X, i.e. r local searches based on bit inversion are performed randomly to generate a new solution X', where r ═ rand (1, λ) X )，λ X Is the wavelength of solution X;
step 7.1.2, for each newly found optimal solution X * Performing a wave breaking operation, i.e. on the new optimal solution X * Neighborhood searching is performed to generate k neighborhood solutions, where the value of k can be expressed as follows max For one predefined parameter:
wherein n is the number of goods, k max Is a predefined parameter, epsilon is a very small positive integer avoiding zero;
step 7.1.3, if the generated optimal neighborhood solution is better than X * Then update X * Otherwise, X is retained * The change is not changed;
step 7.2, updating the population scale, and removing the worst solution in the population if the population scale is reduced;
an S-WWO algorithm which abandons a refraction operator is adopted, and a population reduction strategy is used at the same time, and is expressed as follows:
wherein g and g max Current and maximum allowed number of iterations (or fitness calculation times), respectively, and NP max And NP min Maximum and minimum population numbers, respectively, population-sized NPs are removed from NPs by iteratively deleting the current worst solution max Reduced to NP min ；
And 8, if the termination condition is met, returning the maximum picking completion time of all the AGVs, finishing the algorithm, otherwise, returning to the step 4.
To facilitate understanding of the effects of the present invention, the following experimental description is provided:
the experimental data adopts 10 groups of experimental data sets generated based on picking work in the intelligent warehouse under the actual condition of three scales. Picking operations with different quantities of goods in warehouses with warehouse sizes of 50 × 50, 100 × 100 and 200 × 200 were experimentally tested.
The hybrid Algorithm proposed in the present description is compared with five hybrid algorithms, namely, currently popular Genetic Algorithm (GA), Biogeography-based optimization (BBO), ecography-based optimization (EBO), Differential Evolution (DE), and PSO, as main algorithms, to test the performance of the hybrid Algorithm proposed in the present description, and each Algorithm uses TS Algorithm as a route planning sub-Algorithm.
Each algorithm was run 30 times on each experimental example, and the Number of Fitness function Evaluations (NFEs) was used as a termination condition, with the NFEs of each experimental example being 500m (m is the Number of AGVs). In the experiment, we set v 2, Δ t 1, and Q50.
The experimental results show the maximum value, the minimum value, the median value and the standard deviation (expressed in seconds) of the results obtained by each algorithm in the current experimental example, and the minimum median value and the minimum value in the experimental results are shown in a bold mode.
The experimental results are shown in tables 1 and 2 below:
TABLE 1
TABLE 2
According to the experimental results, the S-WWO algorithm obtains the minimum median value and the minimum value in all experimental examples; in addition, the median of the maximum picking completion time of all AGVs obtained by the mixed algorithm of the S-WWO algorithm as the main algorithm on the data sets 1 to 10 is about 6.35 to 10.10 percent smaller than GA, about 12.28 to 18.16 percent smaller than BBO, about 9.90 to 15.26 percent smaller than EBO, about 3.85 to 6.71 percent smaller than DE and about 8.26 to 12.59 percent smaller than PSO. Therefore, the AGV operation optimization scheduling method based on the water wave optimization-tabu search algorithm can effectively solve the problems, shorten the maximum goods picking time and improve the storage operation efficiency.
The invention has the beneficial effects that: a plurality of AGV in the operation of rationally scheduling storage improve warehouse operating efficiency, shorten the biggest completion time of choosing goods, promote storage system operating efficiency. The content of the example of the present specification is only an illustration of one embodiment of the present application, and a detailed problem solving process is described, but the scope of the present invention is not limited to the specific form described in the example, and several variations and modifications can be made without departing from the basic concept of the present application.
Claims (4)
1. An intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search is characterized by comprising the following steps:
step 1, inputting a goods picking operation data set of an intelligent AGV, and analyzing the data set to obtain information such as warehouse scale, goods picking amount and AGV number;
step 2, constructing a warehouse environment model by using a grid method, namely establishing a coordinate system based on warehouse entry coordinates, wherein storage positions are all located at grid intersections;
step 3, randomly generating a group of initial solution population P, wherein each individual in the population represents the distribution condition of all cargos by a solution X of a simple water wave optimization main algorithm distributed for the cargos, and the X is represented by a m multiplied by n dimensional 0-1 vector;
step 4, planning a picking sequence for each AGV by using a tabu search sub-algorithm based on each solution generated in the initial population;
step 5, for each solution in the population, planning a collision-free picking route for all the AGV by using a collision avoidance strategy;
step 6, calculating a fitness function value of each solution X in the population;
step 7, evolving the solution of the main algorithm by using a simple water wave optimization main algorithm; the method comprises the following steps:
step 7.1, for each solution X in the population P, the following operations are performed:
step 7.1.1, a propagation operation is performed on solution X, i.e. r local searches based on bit inversion are performed randomly to generate a new solution X', where r ═ rand (1, λ) X )，λ X Is the wavelength of solution X;
step 7.1.2, for each newly found optimal solution X * Performing a wave breaking operation, namely performing neighborhood search on the wave breaking operation to generate k neighborhood solutions, wherein the neighborhood search is also performed by using local search in the wave breaking operation, and the value of k is represented as follows:
wherein n is the number of goods, k max Is aA predefined parameter, epsilon being a very small positive integer avoiding zero;
and the local search and neighborhood search operation in the propagation operator and the wave breaking operator are modified, the modified search operation is based on bit reversal, and one element X is randomly selected from the solution X every time ji J is more than or equal to 1 and less than or equal to m, i is more than or equal to 1 and less than or equal to n, and the value is modified into x ji ＝1-x ji And then randomly selecting an element X in X j′i J ≠ j', if its value is equal to y ji If the original value is reversed, the value is modified to x j′i ＝1-x j′i ；
Step 7.1.3, update the optimal solution X * ；
Step 7.2, updating the population scale, and removing the worst solution in the population if the population scale is reduced;
the population reduction strategy is represented as:
wherein g and g max Current and maximum allowed number of iterations, respectively, and NP max And NP min Maximum and minimum population numbers, respectively, population-sized NPs are removed from NPs by iteratively deleting the current worst solution max Reduced to NP min ；
And 8, if the termination condition is met, returning the maximum completion time of all the AGVs, finishing the algorithm, otherwise, returning to the step 4.
2. The method for optimizing and scheduling AGV operation in an intelligent warehouse based on water wave optimization-tabu search as claimed in claim 1, wherein in said step 4, the construction process is as follows:
step 4.1, deriving an instance for each subproblem based on X, namely determining the goods allocated to each AGV;
step 4.2, setting a null taboo table tabu and setting a taboo length TabuLen;
4.3, generating a goods picking path for the AGV as an initial solution y by using a greedy algorithm;
step 4.4, if the termination condition is met, returning to the optimal picking route, and ending the algorithm;
step 4.5, randomly selecting a point in the initial solution y as a tabu object;
step 4.6, perform neighborhood search on y, update the optimal candidate solution using the best improved solution priority strategy, generate NbSize new solutions, and update the optimal candidate solution y' and the optimal move d best ；
Step 4.7, if the final best candidate solution y' is better than y or
The following steps are carried out:
step 4.7.1, mixing d best Adding the tail of the tab;
step 4.7.2, if the tabu.length is greater than TabuLen, removing the taboo object added with the tabu at the earliest time;
step 4.7.3, using y' to update y;
step 4.8, if y is superior to the current global optimal solution y * Updating y with y * 。
3. The method for optimizing and scheduling AGV operation in intelligent warehouse based on water wave optimization-tabu search as claimed in claim 1 or 2, wherein in said step 5, for each solution in the population, a collision-free picking route is planned for all AGVs using collision-avoidance strategy, comprising the following steps:
step 5.1, generating a detailed route set S of all AGVs by using a row-first method;
step 5.2, finding out all collision points, namely the points where the position and the time of the AGV conflict, and sequencing the points according to the ascending order of the time;
step 5.3, for each collision point c and for the two AGVs colliding therewith, the following steps are performed:
step 5.3.1, according to the two AGV closest target points, selecting one AGV as b according to the following rules:
1) the target points are positioned in the same column, and the AGV with the larger number of the target point rows is selected as b;
2) the target points are located in the same row, and an AGV is randomly selected as b;
3) otherwise, selecting the AGV with the smaller column where the target point is positioned as b;
step 5.3.2, order
The intersection which is before c and is closest to c in the detailed path of b;
step 5.3.3, if b is
If the vehicle does not collide at the next point when the vehicle is driven in a column-first mode, replacing the original path segment and updating the detailed path of the b;
step 5.3.4, if not, continuing to backtrack the last intersection, judging whether the path segment can be updated in a column-first mode, and if so, finally
If the collision still occurs when the starting point p of the b is traced back, the operation of changing the standby route is not carried out, and one of the two AGVs with shorter picking completion time is selected to wait for the other AGV to pass through, then the operation is continued, and the picking completion time of the AGV is updated.
4. The method for optimizing and scheduling AGV operation in an intelligent warehouse based on water wave optimization-tabu search as claimed in claim 1 or 2, wherein the step 6 of calculating the fitness function value of each solution X in the population specifically comprises the following steps:
step 6.1, calculating the maximum picking completion time T of each AGV j The AGV comprises the goods picking travel time and the goods picking time of the AGV;
step 6.2, calculating the fitness function value of the solution X, and for each AGV, the total demand of the goods distributed to the AGV needs to meet the capacity constraint, so when calculating the fitness of each solution, adding the constraint violation condition into the target function in the coding, and defining the constraint violation condition as:
wherein, T j The maximum pick completion time for the jth AGV, and P is a very large positive integer used to constrain capacity violations, x ji Indicates whether the ith load is assigned to the jth AGV, q i Q is the capacity of the AGV for the ith load demand.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110510257.1A CN113359702B (en) | 2021-05-11 | 2021-05-11 | Intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110510257.1A CN113359702B (en) | 2021-05-11 | 2021-05-11 | Intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113359702A CN113359702A (en) | 2021-09-07 |
| CN113359702B true CN113359702B (en) | 2022-09-23 |
Family
ID=77526146
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110510257.1A Active CN113359702B (en) | 2021-05-11 | 2021-05-11 | Intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113359702B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114115259B (en) * | 2021-11-18 | 2024-04-16 | 青晨(广州)电子商务科技有限公司 | AGV real-time path gauge and anti-collision method and system |
| CN114326621B (en) * | 2021-12-25 | 2023-11-14 | 长安大学 | Group intelligent airport consignment car scheduling method and system based on layered architecture |
| CN115759407B (en) * | 2022-11-18 | 2023-08-01 | 宝开(上海)智能物流科技有限公司 | Multi-dimensional scheduling scheme optimization method, system and storage medium based on sorting system |
| CN117151596B (en) * | 2023-11-01 | 2023-12-29 | 领先未来科技集团有限公司 | Logistics management method, system and storage medium for storage AGVs (automatic guided vehicle) through Internet of things |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4138541B2 (en) * | 2003-03-13 | 2008-08-27 | 独立行政法人科学技術振興機構 | Distributed path planning apparatus and method, and distributed path planning program |
| CN106779153B (en) * | 2016-11-15 | 2021-08-03 | 浙江工业大学 | Intelligent stereoscopic warehouse goods space allocation optimization method |
| CN112686474B (en) * | 2021-01-22 | 2022-04-22 | 华南理工大学 | Parallel assembly line balancing method based on improved water wave optimization algorithm |
-
2021
- 2021-05-11 CN CN202110510257.1A patent/CN113359702B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113359702A (en) | 2021-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113359702B (en) | Intelligent warehouse AGV operation optimization scheduling method based on water wave optimization-tabu search | |
| WO2019154445A2 (en) | Warehouse entry/exit control method for shelf array, and transportation system | |
| Van Laarhoven et al. | Job shop scheduling by simulated annealing | |
| CN107036618A (en) | A kind of AGV paths planning methods based on shortest path depth optimization algorithm | |
| CN113075927A (en) | Storage latent type multi-AGV path planning method based on reservation table | |
| CN107657364A (en) | A kind of overloading AGV tasks towards tobacco plant material transportation distribute forming method | |
| Zou et al. | Efficient multiobjective optimization for an AGV energy-efficient scheduling problem with release time | |
| Zhang et al. | A Building‐Block‐Based Genetic Algorithm for Solving the Robots Allocation Problem in a Robotic Mobile Fulfilment System | |
| Hu et al. | Conflict-free scheduling of large-scale multi-load AGVs in material transportation network | |
| Li et al. | A simulation study on the robotic mobile fulfillment system in high-density storage warehouses | |
| Sun et al. | An autonomous vehicle interference-free scheduling approach on bidirectional paths in a robotic mobile fulfillment system | |
| Zou et al. | An effective discrete artificial bee colony algorithm for scheduling an automatic-guided-vehicle in a linear manufacturing workshop | |
| CN116523165B (en) | Collaborative optimization method for AMR path planning and production scheduling of flexible job shop | |
| CN114595607B (en) | Digital twin spinning barrel conveying method and system | |
| Zhu et al. | A decision model on human-robot collaborative routing for automatic logistics | |
| Teck et al. | A bi-level memetic algorithm for the integrated order and vehicle scheduling in a RMFS | |
| CN116523221A (en) | Optimal scheduling method and system for intelligent warehouse picking task | |
| CN116224926A (en) | Dynamic scheduling optimization method and device for single-piece small-batch flexible manufacturing workshops | |
| Shi et al. | Task allocation and path planning of many robots with motion uncertainty in a warehouse environment | |
| Farahani et al. | Online multimodal transportation planning using deep reinforcement learning | |
| Nishida et al. | Just-in-time routing and scheduling for multiple automated guided vehicles | |
| Jiao et al. | Online joint optimization of pick order assignment and pick pod selection in robotic mobile fulfillment systems | |
| Qin et al. | Dynamic dispatching for interbay automated material handling with lot targeting using improved parallel multiple-objective genetic algorithm | |
| Li et al. | A reinforcement learning-based hyper-heuristic for AGV task assignment and route planning in parts-to-picker warehouses | |
| Miao et al. | Joint scheduling of parallel machines and AGVs with sequence-dependent setup times in a matrix workshop |
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 |