Design and research of automated warehouse simulation platform based on virtual visualization framework

Warehouse conveying control system

The warehouse transportation control system uses a scheduling mode, which will detect the movement of the equipment in real-time and constantly feed back the movement in the process of transportation. The system adopts an automatic control mode. In this way, the virtual host computer sends instructions to the controller of the equipment. After receiving the instructions, the controller automatically controls the equipment to complete the operation and feeds back the motion state of the equipment to the virtual host computer. This part of the system is the connecting bridge between the automated finished product palletizing system and the automated warehouse system.

In this virtual simulation scenario, high-rise shelves will be used to allow goods to enter and exit automatically and improve work efficiency. In this article, the stacker is used to access the goods, and the warehouse area and the storage area are connected by the roller conveyor line. The automated warehouse is mainly composed of a roller conveyor line, stacker, shelf, mechanical arm, AGV car, and other parts. The working mode of the automated warehouse is as follows: when entering the warehouse, a worker first places the goods on a roller conveying line, and the roller conveying line sends the goods to a warehouse area. Then, a stacker takes down the goods from the roller conveying line, and finally, the stacker transports the goods and places the goods in a designated bin. When going out of the warehouse, the stacker takes out the goods in the designated bin, then places the goods on the roller conveyor line and sends them to the automatic sorting area. The identification device of the automatic sorting system identifies which lane the goods should belong to through the bar code, the conveyor will send the goods to the corresponding lane. Then, the mechanical arm will put the goods neatly on the pallet. Finally, the AGV car delivers the goods and pallets to the goods storage area (Yan, Himan & Yangkun, 2020). The overall planning of the automated warehouse virtual simulation is shown in Fig. 1.

Overall architecture of simulation system

The 3D models of shelves, stackers, conveyor belts, mechanical arms, pallets, automatic sorting equipment, and AGV cars in the automated warehouse are built by 3DMAX. Then, the 3D models of the automated warehouse are imported into the engine for rendering and mapping. Actions are added to the stacker, cargo, AGV trolley, and mechanical arm in the three-dimensional model of the automated warehouse to present the virtual simulation design more perfectly. The overall framework is shown in Fig. 2.

By creating, accessing, reading, and writing the shared memory, the variables in the PLC ladder diagram are analyzed and assigned to the three-dimensional model of the automated warehouse. The purpose is to control the access of the automated warehouse through the PLC ladder diagram, which is more in line with the realistic operation mode (Xin, Jinsong & Junlin, 2019).

Overall composition of the warehousing system

In this article, the composition of the automated warehouse is redesigned, which can make the warehousing more rapid and accurate. The system mainly includes four aspects: the automated warehouse system, the automated finished product palletizing system, the conveying system, and the system controller. The warehousing system framework is shown in Fig. 3.

In Fig. 3, two-way data communication can be realized between the automated warehouse and each component to realize the timely feedback of data. For example, after receiving the instructions from the conveying system, the conveyor can feed back the data from the start to the completion of the goods conveying.

Overall scheme of control system

Whether the whole system runs smoothly and efficiently depends on the design of the control system. The control system needs to be designed according to certain characteristics of the goods and the frequency of goods access. It should conform to the actual production operation, which not only realizes the required system as a whole but also improves the overall work efficiency so that the system can be more in line with reality (Longman Ryan et al., 2023). The control of the system mainly includes the control of the automatic palletizing system, automated warehouse system, and storage and transportation system. The following is an introduction to these three aspects:

• (1) Automatic palletizing control system

The control of the automatic finished product palletizing system mainly includes: the controller controls the palletizing robot to complete the actions that should be completed, and the goods are neatly stacked on the pallet or conveyor belt (Zhang & Lv, 2022). The controller controls the conveyor belt, and its main function is to transport the goods. The controller detects the operation of the goods on the conveyor belt. Location detects whether the goods arrive at the place where they should arrive (Changpu & Yichun, 2019). The virtual host computer sends job instructions to each part. The controller controls the relevant equipment, and all aspects of each equipment are fed back to the virtual host computer. The structure of the automatic palletizing control system is shown in Fig. 4.

• (2) Automated warehouse system

The automated warehouse system mainly controls the movement of the stacker. The correct movement of the stacker is the most critical part of the project. The control of the stacker mainly includes receiving operation instructions and specifying whether the goods are warehoused or warehoused (Ammouri, 2023). If the control design of the stacker is reasonable, the overall work efficiency will be improved, mainly including the control of vertical lifting, horizontal movement, and forks of the stacker, as well as the detection of accurate arrival at the bin and whether there are goods in the bin (Alizadehsalehi & Yitmen, 2023). The structure of the automated warehouse control system is shown in Fig. 5.

Design of improved Tabu Search algorithm

In this article, an improved Tabu Search algorithm is designed to solve the optimization problem of space allocation. The pseudocode for the improved algorithm is as follows:

Input: Functions at the port of entry and exit.

Output: Objective function of warehousing efficiency.

Batches now = seed.seed(C, X, Y,A).

Batches now = Batches now[1:].

if ratio * C <= (batches now[−1]) <= C:

batches.extend(batches now).

else:

Orders left.extend(batches now[]).

del batches now[]

batches.extend(batches now).

if ratio * C <= orders item number() <=

Batches.append ().

if orders item number(,) < ratio *

Orders left.extend().

orders = orders left.

if orders item number(orders) <=

batches.append(orders).

The candidate solution set may be equal to the neighborhood solution set or belong to the neighborhood solution set, that is, as long as the candidate solution set itself is a proper subset of the neighborhood solution set, the algorithm may not correctly select the optimal solution of the neighborhood solution set. But this situation will still reduce the search time. A better solution may still be obtained after many iterations. When there is no better solution in the candidate solution set selected in this article, the best inferior solution in the current situation is selected as the initial solution of the next iteration.

While fully retaining the local depth search ability of the traditional Tabu Search algorithm and the good global depth search ability of the genetic algorithm, the two algorithms overcome their respective shortcomings when they are used together. The step flow of the improved Tabu Search algorithm is shown in Fig. 6.

The number of goods that should be allocated is obtained through the dynamic programming algorithm. It can be seen that the core and difficulty of the algorithm lies in whether the objective function (penalty evaluation function) can meet the requirements of the algorithm.

The data center module is used to store and manage data. It is also a piece of system memory, but it is closed and does not allow other processes to access it. The configuration file was called once during initialization. A configuration file is a file containing all variables (Fardad & Ali, 2023). All variable data is read here. The tool can be used to write configuration files interfacially and intuitively, and the establishment of configuration files can effectively eliminate hard coding so that the modification operation does not need to be in the source file (Wei et al., 2019). Some important codes of the data center module are as follows:

Xml Document doc = new Xml Document();

doc.Load(System.App Domain.Current Domain.Base Directory +

config Full Name);

data Array = new Data[doc.Last Child.Child Nodes.Count];

for (int i = 0; i < data Array.Length; i++)

Data temp = new Data();

foreach(Xml Node clone Child in doc.Last Child.Child Nodes[i].Child Nodes).

switch (clone Child.Name).

temp.length_PLC =

ushort.Parse(clone Child.Inner Text);

catch

temp.length_PLC = 1;

data Array[i] = temp;

The read-write data module is used for realizing the functions of establishing a link, reading data, writing data, breaking the link, and the like (Ekren, 2020). First, call the method that can be linked to the PLC hardware device through the IP address and port number. Through this method, a socket object with a long connection will be created to link with the physical object. The subsequent read and write operations will use this connection object.

Experimental data

The data used in the experiment are all from a production line in the production workshop, which is a product assembly line. The weight of each part is measured, as well as the process arrangement and operation time measurement in the product assembly process. According to the daily production plan of the day, under the existing warehouse layout, the size of the automated warehouse, the parameters of the shelves, and the stacker are designed according to the number of parts required in a certain cycle (Rahman, Janardhanan & Nielsen, 2020). The existing warehouse layout is shown in Fig. 7.

Data collection

• (1) Collection of warehouse and facility equipment data

The large automatic warehouse is equipped with automatic shelves, tunnel stackers, chain conveyors and car Kunzi conveyors. Select the shelf used for storing assembly line parts in the warehouse to carry out the operation of goods in and out of the warehouse. The height of the shelf is 6.00 m. There are 20 goods compartments in four layers and five rows. The size of each unit goods compartment is $1\times 0.6\times 1.5\mathrm{m}$, and the maximum load is $1,000.00\mathrm{k}\mathrm{g}$. After being palletized, they are stored on shelves, and one pallet can only be equipped with one compartment. The operator of the parts in and out of the warehouse is equipped with a small lane-type stacker.

• (2) Assembly production data collection

Workers are scheduled to work 8 h a day and plan to produce 1,250 units. Firstly, the structure, assembly process and process flow of this kind of assembly parts are analyzed. The production time of each assembly part is calculated, and the operation process of each production beat and production line balance is calculated (Li et al., 2023).

• ① Product structure analysis of assembly parts

Obtain the assembly BOM and WI manual from the production line Kanban, and arrange the relevant data as shown in Tables 1 and 2 (Li, Shen & Zhou, 2023).

• ② Analysis of assembly process flow

Based on the WI of the assembly, the assembly sequence and time of each process are given, and the relevant data are shown in Table 3.

• ③ Determine the takt time

The takt time R of each product is calculated by producing 1,250 assemblies in a working time of 8 H/day.

Takt Time:

(1) $$R={\displaystyle \frac{T}{C}={\displaystyle \frac{8\times 1,800s}{1,000}}}$$

(2) $$S_{min}={\displaystyle \frac{S}{C}}$$where: T is the working time/s, C is the production quantity, and S_min is the minimum number of stations.

The operation flow of each process is shown in Fig. 8.

Algorithm solution and analysis of cargo location assignment optimization model

When calculating the allocation scheme of assembly parts, it is assumed that the start time, acceleration and deceleration time, braking time, and picking time are all o during the operation of the stacker, that is, the stacker always runs at a uniform speed in the horizontal and vertical directions (Arbex Valle & Beasley, 2020). The parts warehousing platform is located at the same end of the shelf, and a coordinate system with the position of the goods near the warehousing platform as the coordinate origin is established. The X-axis direction is the length direction of the shelf. The Y-axis direction is the height direction of the shelf, and the Z-axis is the width direction of the rack.

The adaptability of the optimal candidate solution in this article is much higher than that of the historical optimal solution. If a certain number of candidate solutions are selected in the neighborhood of the current solution, the solution will be accepted by the algorithm regardless of whether it is in the Tabu list (Hosamo et al., 2023). If it is uncertain that a candidate solution of this type exists, replace the current solution by selecting a candidate solution that does not exist in the Tabu list, and add its corresponding move to the Tabu list (Ward et al., 2021). Through repeated iterations until the requirements of the stop criterion are met.

Before solving with the algorithm, determine the specific values of the algorithm parameters. The algorithm parameters are shown in Table 4.

Comparison of algorithm solution results

Input the pseudo-code in the MATLAB software and run the code to solve the location-allocation scheme, and finally get the sorting sequence of parts. According to the pre-set matrix definition, get the location coordinates of each part on the shelf through the sequence. At the same time, the iteration graph of the traditional Tabu Search algorithm is obtained by running the MATLAB software. The iterative results of the algorithm are shown in Fig. 9.

After using the improved Tabu Search algorithm, that is, Tabu Search improves the genetic search population so that the performance of the algorithm has been significantly improved. The value of the objective function is 0.107 < 0.1381 (the result of the traditional Tabu Search algorithm). The algorithm has good convergence. The strategy of Tabu Search to improve the population makes the algorithm search for a considerable proportion of individuals in the population, which can effectively promote the population to develop in a better direction and make the results of the algorithm tend to the optimal solution.

Comparative analysis of storage location-allocation schemes and algorithms

The detailed results of the parts location-allocation optimization scheme of the assembly parts are shown in Tables 5 and 6.

Table 5:

Location coordinates of goods location-allocation by traditional Tabu Search algorithm.

Number	Codes	Coordinates	Number	Codes	Coordinates
1	B01_10	(3, 1, 1)	11	B01_110	(3, 3, 1)
2	B01_20	(3, 4, 1)	12	B01_120	(1, 3, 1)
3	B01_30	(2, 2, 1)	13	B01_130	(4, 4, 1)
4	B01_40	(2, 5, 1)	14	B01_140	(4, 3, 1)
5	B01_50	(3, 5, 1)	15	B01_150	(4, 5, 1)
6	B01_60	(2, 3, 1)	16	B01_160	(4, 2, 1)
7	B01_70	(1, 5, 1)	17	B01_170	(2, 1, 1)
8	B01_80	(4, 1, 1)	18	B01_180	(1, 4, 1)
9	B01_90	(1, 1, 1)	19	B01_190	(2, 4, 1)
10	B01 100	(3, 2, 1)	20	B01 200	(1, 2, 1)

DOI: 10.7717/peerj-cs.1809/table-5

Table 6:

Location coordinates of improved Tabu Search algorithm for location-allocation.

Number	Number of subrogation location-allocation		Delegate location assignment
	Code	Coordinates	Code	Coordinates
1	B01_10	(1, 3, 1)	B01_110	(3, 2, 1)
2	B01_20	(4, 1, 1)	B01_120	(2, 2, 1)
3	B01_30	(2, 5, 1)	B01_130	(1, 5, 1)
4	B01_40	(4, 2, 1)	B01_140	(1, 2, 1)
5	B01_50	(4, 5, 1)	B01_150	(4, 3, 1)
6	B01_60	(2, 3, 1)	B01_160	(3, 1, 1)
7	B01_70	(4, 4, 1)	B01_170	(3, 5, 1)
8	B01_80	(2, 4, 1)	B01_180	(3, 3, 1)
9	B01_90	(3, 4, 1)	B01_190	(2, 1, 1)
10	B01 100	(1, 4, 1)	B01 200	(1, 1, 1)

DOI: 10.7717/peerj-cs.1809/table-6

The coordinates of the cargo allocation scheme obtained by the traditional Tabu Search algorithm and the improved Tabu Search algorithm are substituted into the objective function to obtain the objective function value.

• (1) Warehouse-in and warehouse-out efficiency of assembled parts

The coordinates of the cargo allocation scheme obtained by the traditional Tabu Search algorithm and the improved Tabu Search algorithm are substituted into the objective function to obtain the objective function value.

• (2) Storage efficiency of assembled parts

(3) $$T_1(x,y)=2\times \sum _m^{x=1}\sum _n^{y=1}(D_{xy}\times t_{xy})$$

(4) $$D_{xy}=D_{xy}^{\prime }\times Volum$$

(5) $$t_{xy}=max\{(x-0.5)\times {\displaystyle \frac{L}{v_x},(y-0.5)\times {\displaystyle \frac{H}{v_y}\}}}$$

Substitute the weight $D_{xy}$ of the parts, the running speed $v_x=1.2m/s,v_y=1m/s$ of the stacker and the coordinates of the parts in Table 2 into the formula to calculate the efficiency $T=1629<8h$ of parts in and out of the warehouse.

• (3) Shelf stability

Equivalent center of gravity $G_x$ in the overall horizontal direction of the shelf is as follows:

(6) $$G_x=\sum _m^{x=1}\sum _n^{y=1}{D}_{xy}\times (x-0.5)\times L/\sum _m^{x=1}\sum _n^{y=1}{D}_{xy}$$

Equivalent center of gravity $G_y$ in the vertical direction of the overall shelf is as follows:

(7) $$G_y=\sum _m^{x=1}\sum _n^{y=1}{D}_{xy}\times (y-0.5)\times H/\sum _m^{x=1}\sum _n^{y=1}{D}_{xy}$$

The equivalent center of gravity $G_{TS}=(3.31,2.75)$, $G=\left(2.18,0.92\right)$are obtained by substituting the shelf size $L=4m,H=6m$, the weight of the parts and the location-allocation coordinates of the parts into Formula (6).

The objective function values obtained by the traditional Tabu Search algorithm and the improved Tabu Search algorithm are compared, and the comparison results are shown in Table 7.

Table 7:

Solution results of traditional Tabu Search algorithm and the improved Tabu Search algorithm.

Objective function	Traditional tabu algorithm	Improved algorithm	Indications
In and out storage efficiency	2,565.5 s	1,727 s	Reflect the storage operation capacity, the smaller the value, the better
Overall equivalent center of gravity of the rack	(3.43, 2.86)	(2.39, 0.79)	Reflecting the shelf stability index, the closer the equivalent center of gravity is to (2.5, 0.5), the safer the shelf is

DOI: 10.7717/peerj-cs.1809/table-7

According to the objective function value results of the two algorithms in Table 7, the results obtained by the improved Tabu Search algorithm are better than those obtained by the traditional Tabu Search algorithm. The operation time of the stacker is shorter. The overall equivalent center of gravity of the shelf is closer to the geometric center of gravity of the shelf in the horizontal direction and closer to the bottom position in the vertical direction.

In addition, when the traditional Tabu Search algorithm is used to solve the problem of cargo allocation, there is a possibility that the algorithm will converge in advance when the results are not mature. In the improved Tabu Search algorithm, the hybrid strategy of Tabu Search improved the diversity of the population and expanded the search area. To eliminate the accidental operation of the algorithm, the traditional Tabu Search algorithm and the improved Tabu Search algorithm are respectively run five times in the MATLAB software, and the multiple iterations of the two algorithms are obtained as shown in Fig. 10.

Stability coefficient refers to the ratio of the number of repeated runs of the algorithm and the number of times and the final run result is the same as the total number of runs of the algorithm. The higher the ratio is, the more stable the algorithm is. The stability factor is as follows:

(8) $$sta={\displaystyle \frac{n}{N}}$$

It can be seen from the trend of the fitness value curve and the final fitness value in the running iteration chart of the two algorithms in Fig. 10 that the fitness value of each of the five repeated operations of the traditional Tabu Search algorithm is different. The stability coefficient is 0, while the results of four of the five operations of the improved Tabu Search algorithm are the same. The stability coefficient is 0.8, which is much higher than that of the traditional Tabu Search algorithm.

Discussion

Through experimental analysis, it is verified that the Tabu list of the Tabu direct search can automatically block the current Tabu area filtered by the Tabu searcher in real-time, which can effectively avoid the model outflanking the direct search again. Then, build a benchmark that can release the area where the Tabu direct search is located to ensure the diversity of relevant information more effectively in the process of the second direct search.

According to basic knowledge, the Tabu machine learning system relies on the structural framework of neural network parallel search, combines the powerful potential of the Tabu machine learning system, avoids bypassing direct search and the diversity of genetic factors in genetic evolution, and promotes rapid global optimization. It fully retains the local organization depth direct search of the traditional Tabu machine learning system and the good global depth direct search potential of neural network database data. It also overcomes the shortcomings of traditional research content. The effectiveness of the algorithm is verified by an example, which can reduce the time and distance of picking to a certain extent and improve the efficiency of order picking. The research results can be used to guide the picking operation under the multi-block warehouse layout in practice and help to improve the timeliness of customer order fulfillment.