CN114548861A - Warehouse management method based on digital twin - Google Patents
Warehouse management method based on digital twin Download PDFInfo
- Publication number
- CN114548861A CN114548861A CN202210108546.3A CN202210108546A CN114548861A CN 114548861 A CN114548861 A CN 114548861A CN 202210108546 A CN202210108546 A CN 202210108546A CN 114548861 A CN114548861 A CN 114548861A
- Authority
- CN
- China
- Prior art keywords
- goods
- warehouse
- data
- model
- task
- 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.)
- Pending
Links
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Operations Research (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
Abstract
The invention discloses a warehouse management method based on digital twins, and belongs to the technical field of intelligent warehousing. According to the method, on the basis of model construction and data mapping, task data and warehouse basic source data are combined, time prediction of each task is achieved through a relevant prediction algorithm, a data basis is provided for task scheduling and planning, and decision making of management personnel is assisted. The method is based on the warehouse place environment, a simulation virtual model of the automatic warehouse space is established, and information such as the position of an automatic conveying platform, the position of goods, a goods inlet, a goods outlet, a goods inlet road, a goods outlet road and the like is considered; the model has good reconfigurability and can be flexibly set according to specific requirements. Meanwhile, roads are planned and defined, vehicle collision and path conflict are avoided through intelligent scheduling and planning simulation, route optimization is achieved, and operation efficiency is improved.
Description
Technical Field
The application relates to the technical field of warehouse management and control, in particular to a warehouse management method based on digital twins.
Background
The current warehouse operation and warehouse management are very complex, and only manual input and operation are needed, so that time and labor are wasted, and errors are easy to occur; meanwhile, the problems that warehouse goods information is difficult to control in real time, the state of the warehousing equipment in the operation process is difficult to monitor in real time, the warehouse management efficiency is low and the like exist. The invention constructs an intelligent warehouse management system and method based on a digital twin technology, and improves the operation efficiency and the comprehensive management level of the warehouse through the technologies of visual management and control, transportation time prediction, intelligent scheduling, path planning and the like.
Disclosure of Invention
The invention aims to solve the technical problem of the prior art and provides a novel warehouse management method based on digital twins.
The technical problem to be solved by the present invention is achieved by the following technical means. The invention relates to a warehouse management method based on digital twins, which is characterized by comprising the following steps:
constructing a digital twin body for a physical entity of the warehouse: the construction objects mainly comprise an automatic conveying platform, a goods shelf, goods and other physical entity equipment of the warehouse; twin display of the digital twin warehouse is completed by three-dimensional modeling of various constructed objects and combining large-screen display;
establishing a mapping from a physical repository to a virtual repository: the method mainly comprises a physical entity, a virtual entity, a service, twin data and mapping connection among all components; by adopting a data driving method, the physical model can be updated and corrected by utilizing the history and real-time operation data of the system, and the real-time operation state of the warehouse is displayed;
on the basis of model construction and data mapping, time prediction of each task is realized by combining task data and warehouse basic source data through a related prediction algorithm, a data basis is provided for task scheduling and planning, and decision making of managers is assisted;
establishing a simulation virtual model of an automatic warehouse space based on the warehouse place environment, and considering the information of the position of an automatic conveying platform, the position of goods, a goods inlet, a goods outlet, a goods inlet road and a goods outlet road; meanwhile, the road is planned and defined, and vehicle collision and path conflict are avoided through intelligent scheduling and planning simulation.
The invention relates to a warehouse management method based on digital twin, which further adopts the following preferable technical scheme that: the method comprises the steps of forecasting the time required by goods to leave the warehouse, wherein the demand forecasting is positioned at the bottommost layer in the whole warehouse management system and plays a supporting role, a plurality of decision optimization systems on the upper layer are supported, the decision optimization systems obtain the optimal decision by combining accurate forecasting data with an operation research technology, and the result is provided for a service execution system on the upper layer.
The invention relates to a warehouse management method based on digital twin, which further adopts the following preferable technical scheme that: the method for predicting the time required by goods to leave the warehouse comprises the following steps: acquiring required service data from an external data source, processing and cleaning basic data, processing and analyzing the data through a time sequence or machine learning or other artificial intelligence technologies, and finally calculating a prediction result and pushing the prediction result to a downstream module for use through multiple ways;
the whole framework of the cargo ex-warehouse demand prediction module is sequentially as follows from top to bottom: the system comprises a data source input layer, a basic data processing layer, a core service layer, a data output layer and a downstream execution module;
(1) data source input layer: the data source of warehouse management comprises most of needed business data, including goods information, warehousing information and historical ex-warehouse information; the data of the cargo consignment plan are mostly input by operators through a storage and transportation system;
(2) core service layer: the layer is a core part and mainly comprises feature construction, a prediction algorithm and prediction result processing; meanwhile, the system can also comprise a plurality of service lines which do not intersect;
the characteristic construction is to convert the cleaned basic data into characteristic data in a standard format through further processing and provide the characteristic data for a subsequent algorithm model; the core algorithm mainly utilizes time sequence analysis, machine learning or other artificial intelligence technologies to predict the goods delivery requirements; the predicted result processing is that the format and some special requirements cannot meet the requirements of downstream modules, so that the predicted result processing also needs to be processed according to actual conditions, such as additional information of standard deviation and goods delivery identification;
the goods ex-warehouse demand prediction core technology is mainly divided into a basic layer, a frame layer, a tool layer and an algorithm layer; the base layer uses HDFS for data storage, and the Yarn is used for resource scheduling and developing a related task scheduling module; the framework layer mainly comprises Spark RDD, Spark SQL and Hive; the tool layer directly uses mature and stable algorithms and models, the algorithms are all packaged in a third party Python package, and the third party Python package is selected from xgboost, numpy, pandas, sklern, scipy or hyper; the algorithm layer is mainly divided into a time sequence and a machine learning algorithm;
processing an input sequence with any time sequence by using the memory inside the RNN; the time series task prediction of the consignment mainly comprises ARIMA and Holt winters; the ARIMA is an autoregressive integral moving average model and is mainly used for predicting a stable sequence like the warehouse goods delivery single quantity; holt winters is also called cubic exponential smoothing algorithm and is used for predicting ex-warehouse cargos with seasonality and obvious trend.
The invention relates to a warehouse management method based on digital twin, which further adopts the following preferable technical scheme that: the algorithm layer is an iterative decision tree algorithm, GBDT, and the algorithm is composed of a plurality of decision trees, the decision results of all the trees are accumulated to make a final decision, and the final decision is used for predicting goods which are higher than the warehouse but have unobvious historical rules.
The invention relates to a warehouse management method based on digital twin, which further adopts the following preferable technical scheme that: the general flow of goods delivery demand prediction based on machine learning algorithm is as follows:
(1) the method comprises the following steps: determining main characteristics through data analysis and model tests, and generating characteristic data in a standard format through a series of tasks;
(2) selecting a model: different goods have different characteristics, so different algorithm models can be distributed according to factors such as the historical ex-warehouse demand of the goods, the sensitivity of special training days and the like;
(3) selecting characteristics: screening a batch of characteristics to filter out the characteristics which are not needed, wherein the characteristics of different types of goods are different;
(4) sample partitioning: grouping training data into a plurality of groups of samples, and generating a model file for each group of samples during training; generally, the same type of goods are divided into a group;
(5) model parameters: selecting the optimal model parameters, wherein the proper parameters can improve the accuracy of the model, and the model training and prediction are respectively carried out on different parameter combinations;
(6) model training: after the characteristics, the model and the sample are determined, model training is carried out, and after training, a model file is generated and stored in the HDFS;
(7) model prediction: reading the model file for prediction execution;
(8) multi-model optimization: in order to improve the prediction accuracy, a plurality of algorithm models are used, and an optimal prediction result is selected after the prediction result of each model is output;
(9) abnormal interception of a predicted value: the more complex and difficult-to-interpret algorithm is, the more easily the condition that the extremely individual predicted value is abnormally high occurs, and the abnormal high prediction cannot be interpreted by combining historical goods ex-warehouse demand data, so that the abnormal values can be intercepted by some rules and replaced by a more conservative numerical value;
(10) and (3) evaluating a model: calculating the prediction accuracy;
(11) and (3) error analysis: the distribution of an error in different dimensions is obtained by analyzing the prediction accuracy so as to provide a reference basis for algorithm optimization.
The invention relates to a warehouse management method based on digital twin, which further adopts the following preferable technical scheme that: the method carries out intelligent scheduling and path planning design of an automatic conveying platform on a plurality of cargos which are simultaneously delivered into and delivered out of the warehouse,
firstly, establishing a simulation virtual model of an automatic warehouse space; mainly considering the vehicle goods parking position, namely idle or goods shelf occupation, a goods inlet, a goods outlet, a goods inlet road and a goods outlet road; the warehouse model has good reconfigurability, and parameters of length and width of a warehouse space, the number and density of shelves and vehicles, the number and position of goods outlets of a goods inlet, the number and density of transportation order tasks are flexibly set according to specific requirements;
in a warehouse space, when a plurality of automatic conveying platforms operate simultaneously, in order to avoid collision conflict, simplify the operation rules of the automatic conveying platforms and improve the operation safety of the system, transverse roads and longitudinal roads among shelf areas are set to be one-way roads;
for the established warehouse space model, the task is in a form of conveying a certain shelf from a parking position to a certain goods inlet/outlet position, and after the goods inlet/outlet task is completed, the shelf is conveyed back to the position of the shelf area;
the task can be broken down into the following 3 steps:
step 1: the automatic conveying platform moves to a task corresponding shelf;
step 2: after the automatic conveying platform is filled, the automatic conveying platform moves to a goods inlet/outlet;
step 3: the automatic conveying platform moves to a proper vacant position to wait for unloading;
aiming at the 1 st step of the logistics task, evaluating all automatic conveying platforms which can execute the task according to an evaluation function, selecting the most appropriate carrier from the evaluation functions, and defining the evaluation function as follows:
gn＝w*tn1*tn2
the above equation represents the total cost, tn, for the nth automated transport platform to perform this task1The time which is expected to be consumed by the nth automatic conveying platform to complete the currently running task is represented; w represents a congestion coefficient used for reflecting the congestion degree of the system; set up w>1 may reflect that the actual time taken to complete the current task is more than expected; tn2Indicating the time spent by the nth automated conveying platform to run to the task-requiring shelf;
dividing all automatic conveying platforms into two types of executing the current task and being in an idle state according to an evaluation function, respectively calculating waiting cost and path cost, evaluating all automatic conveying platforms according to the sum of the waiting cost and the path cost, and selecting the automatic conveying platform with the minimum total cost gn to undertake a conveying task;
aiming at the step 3, the goods shelves carried by the automatic conveying platform are parked in the idle area as soon as possible, so that the automatic conveying platform can continue to execute the next task; therefore, selecting an idle position closest to the starting point for parking; cost estimation using manhattan distance:
gn＝abs(cur.x-n.x)+abs(cur.y-n.y)
in the formula: gnRepresenting the cost of parking to the nth storage location;
cur.x represents the abscissa of the current point, and cur.y represents the ordinate of the current point;
x denotes the abscissa of the nth storage position, n.y denotes the ordinate of the nth storage position; abs represents a function for absolute value;
in the designed warehouse space structure, the method applied to the motion planning of a plurality of automatic conveying platforms is a modified and improved A-algorithm; the basic flow of the A-algorithm is that the nodes are selectively expanded from the starting point according to the estimated cost until the target point is expanded; the key is to select a suitable merit function:
f(n)＝g(n)+h(n)
in the formula: (n) represents the estimated cost from the origin to the destination via node n, (g) (n) represents the true cost from the origin to node n, and h (n) represents the estimated cost from node n to the destination;
if the specific positions of all automatic conveying platforms at each moment are known in advance, possible road conflicts can be predicted and avoided during path planning; setting a space-time running map, recording the running track of each robot along with the time, and recording the planned robot track to form a three-dimensional map of horizontal-vertical-time as a reference for track planning; meanwhile, on the basis of considering the distance of the vehicle moving to a certain position, the waiting time cost increased by collision prevention is further added;
the route planning correcting method corrects the space-time operation map in time after each unplanned avoidance occurs, so that the map conforms to the reality.
The invention relates to a warehouse management method based on digital twin, which further adopts the following preferable technical scheme that: in order to avoid possible collision, an exclusive point is set; only one automatic conveying platform can occupy a certain position at a certain moment; other automated transport platforms at the same time have to wait to enter this position.
Compared with the prior art, the invention has the following beneficial effects:
the method disclosed by the invention is based on model construction and data mapping, combines task data and warehouse basic source data, realizes time prediction of each task through a related prediction algorithm, provides a data basis for task scheduling and planning, and assists management personnel in making decisions. The method is based on the warehouse place environment, a simulation virtual model of the automatic warehouse space is established, and information such as the position of an automatic conveying platform, the position of goods, a goods inlet, a goods outlet, a goods inlet road, a goods outlet road and the like is considered; the model has good reconfigurability and can be flexibly set according to specific requirements. Meanwhile, roads are planned and defined, vehicle collision and path conflict are avoided through intelligent scheduling and planning simulation, route optimization is achieved, and operation efficiency is improved.
Drawings
FIG. 1 is a logic diagram of the method of the present invention;
FIG. 2 is a flow chart of warehousing task demand prediction;
fig. 3 is a flow chart of the intelligent scheduling and path planning service.
Detailed Description
The technical solution of the present invention will be described in detail with reference to the accompanying drawings and embodiments. And are not to be construed as limiting the claims.
Referring to fig. 1, a digital twin-based warehouse management method:
constructing a digital twin body for a physical entity of the warehouse: the construction objects mainly comprise an automatic conveying platform, a goods shelf, goods and other physical entity equipment of the warehouse; twin display of the digital twin warehouse is completed by three-dimensional modeling of various constructed objects and combining large-screen display;
establishing a mapping from a physical repository to a virtual repository: the system mainly comprises a physical entity, a virtual entity, a service, twin data and mapping connection among all components; by adopting a data driving method, the physical model can be updated and corrected by utilizing the history and real-time operation data of the system, and the real-time operation state of the warehouse is displayed;
on the basis of model construction and data mapping, time prediction of each task is realized by combining task data and warehouse basic source data through a related prediction algorithm, a data basis is provided for task scheduling and planning, and decision making of managers is assisted;
establishing a simulation virtual model of an automatic warehouse space based on the warehouse place environment, and considering information of a goods parking position, a goods inlet, a goods outlet, a goods inlet road and a goods outlet road of an automatic conveying platform; meanwhile, the road is planned and defined, and vehicle collision and path conflict are avoided through intelligent scheduling and planning simulation.
As shown in fig. 2, the present embodiment predicts the time required for goods delivery, and the demand prediction is at the lowest layer in the whole warehouse management system and plays a role of supporting, and supports multiple upper-layer decision optimization systems, and these decision optimization systems use accurate prediction data in combination with an operation research technology to obtain an optimal decision, and provide the result to a higher-layer business execution system. The whole framework of the cargo ex-warehouse demand prediction module is sequentially as follows from top to bottom: the system comprises a data source input layer, a basic data processing layer, a core service layer, a data output layer and a downstream execution module.
Firstly, acquiring required service data from an external data source, then processing and cleaning basic data, then processing and analyzing the data through artificial intelligence technologies such as time series or machine learning, and finally calculating a prediction result and pushing the prediction result to a downstream module for use through multiple ways.
Fig. 2 shows a core content introduction of the ex-warehouse demand prediction technology, which mainly includes the following contents:
(1) data source input layer: the data sources for warehouse management include most of the business data that is needed, such as cargo information, warehousing information, historical warehouse-out information, and so on. Most of the data of the cargo consignment plan come from operators and are input by the storage and transportation system;
(2) core service layer: the layer is a core part and mainly comprises feature construction, a prediction algorithm and prediction result processing; and may contain multiple non-intersecting service lines. The characteristic construction is to convert the cleaned basic data into characteristic data in a standard format through further processing and provide the characteristic data for a subsequent algorithm model; the core algorithm mainly utilizes artificial intelligence technologies such as time series analysis and machine learning to predict the goods delivery demand, and is the most core part in the prediction system; the predicted result processing cannot meet the requirements of downstream modules on format and some specificity, so that the predicted result processing needs to be processed according to actual conditions, such as adding additional information such as standard deviation, goods warehouse-out identification and the like.
The core technology for forecasting the goods delivery demand is mainly divided into a basic layer, a framework layer, a tool layer and an algorithm layer. The base layer uses HDFS for data storage, and the Yarn is used for resource scheduling and developing a related task scheduling module; the framework layer mainly comprises Spark RDD, Spark SQL and Hive; the tool layer directly uses mature and stable algorithms and models, the algorithms are all packaged in a third-party Python package, and the common packages are xgboost, numpy, pandas, sklern, scipy, hyperopt and the like; the algorithm layer is mainly divided into a time series algorithm and a machine learning algorithm, such as an iterative decision tree algorithm, GBDT, and the algorithm consists of a plurality of decision trees, decision results of all the trees are accumulated to make a final decision, and the final decision is used for predicting goods which are high out of a warehouse but have unobvious historical laws. In addition, the RNN internal memory is used to handle input sequences at arbitrary timing, which makes it easier to handle tasks such as timing prediction. The time-series task prediction of cargo consignment mainly comprises ARIMA and Holt witters. The ARIMA is an autoregressive integral moving average model and is mainly used for predicting a stable sequence like warehouse goods delivery single quantity. Holt winters is also called cubic exponential smoothing algorithm and is used for predicting ex-warehouse goods with seasonality and obvious trend.
The general flow of goods delivery demand prediction based on machine learning algorithm is as follows:
1) the method comprises the following steps: the main characteristics are determined through data analysis and model tests, and characteristic data in a standard format are generated through a series of tasks.
2) Selecting a model: different goods have different characteristics, so different algorithm models can be distributed according to factors such as historical ex-warehouse demand of the goods, special training day sensitivity and the like.
3) Selecting characteristics: a batch of characteristics is screened to filter out unwanted characteristics, different types of goods being characterized differently.
4) Sample partitioning: the training data is grouped into a plurality of groups of samples, and a model file is generated for each group of samples during real training. Typically the same type of goods are grouped together, which is done to account for parallelization and model accuracy.
5) Model parameters: the optimal model parameters are selected, and the appropriate parameters improve the accuracy of the model because different parameter combinations need to be trained and predicted respectively.
6) Model training: and after the characteristics, the model and the sample are determined, model training can be carried out, the training usually takes a long time, and a model file is generated after the training and is stored in the HDFS.
7) Model prediction: and reading the model file for prediction execution.
8) Multi-model optimization: in order to improve the prediction accuracy, a plurality of algorithm models may be used, and after the prediction result of each model is output, the system selects an optimal prediction result through some rules.
9) Abnormal interception of a predicted value: the more complex and difficult to interpret algorithms are, the more likely the extreme individual predicted values are abnormally high, and the high predicted values cannot be interpreted in combination with historical cargo ex-warehouse demand data, so that the abnormal values can be intercepted by some rules and replaced by a more conservative value.
10) And (3) evaluating a model: the prediction accuracy is calculated.
11) And (3) error analysis: the distribution of an error in different dimensions is obtained by analyzing the prediction accuracy so as to provide a reference basis for algorithm optimization.
As shown in fig. 3, in this embodiment, the intelligent scheduling and the path planning design of the automatic transportation platform are performed on a plurality of goods simultaneously entering and exiting the warehouse, and a simulation virtual model of the automatic warehouse space is first established. The vehicle cargo parking position (free or occupied shelf), the cargo inlet, the cargo outlet, the cargo inlet road and the cargo outlet road are mainly considered. The warehouse model has good reconfigurability, and parameters such as the length and width of the warehouse space, the number and density of the shelves and the vehicles, the number and the density of the goods inlets and outlets, the position, the number and the density of transportation order tasks and the like can be flexibly set according to specific needs. In a warehouse space, when a plurality of automatic conveying platforms operate simultaneously, in order to avoid collision conflict, simplify the operation rules of the automatic conveying platforms and improve the safety of system operation, transverse roads and longitudinal roads among shelf areas are set to be one-way roads. For the established warehouse space model, the task is in the form of conveying a certain shelf from a parking position to a certain goods inlet/outlet position, and after the goods inlet/outlet task is completed, the shelf area position is conveyed back. The task can be decomposed into that the automatic conveying platform moves to a corresponding goods shelf of the task, the automatic conveying platform moves to the goods inlet/outlet after the automatic conveying platform is filled, and the automatic conveying platform moves to a proper vacant position to wait for unloading.
Aiming at the 1 st step of the logistics task, evaluating all automatic conveying platforms which can execute the task according to an evaluation function, selecting the most appropriate carrier from the evaluation functions, and defining the evaluation function as follows:
gn＝w*tn1*tn2
the above equation represents the total cost, tn, for the nth automated transport platform to perform this task1Indicating the time it is expected for the nth automated transport platform to complete the task currently running (this is 0 if it is currently idle). w represents a congestion coefficient to reflect the congestion level of the system. Set up w>A 1 may reflect that the actual time taken to complete the current task is more than expected. tn2Indicating the time spent by the nth automated transport platform to run to the task requiring shelf. And according to an evaluation function, dividing all automatic conveying platforms into two types of executing the current task and being in an idle state, respectively calculating waiting cost and path cost, evaluating all automatic conveying platforms according to the sum of the waiting cost and the path cost, and selecting the automatic conveying platform with the minimum total cost gn to bear the transportation task.
In the step 3, the goods shelves conveyed by the automatic conveying platform should be parked in the idle area as soon as possible, so that the automatic conveying platform can continue to execute the next task. Therefore, an idle position closest to the starting point is selected for parking. Cost estimation using manhattan distance:
gn＝abs(cur.x-n.x)+abs(cur.y-n.y)
in the formula: gnIndicating the cost of parking to the nth storage location. cur.x denotes the abscissa of the current point, and cur.y denotes the ordinate of the current point. X denotes the abscissa of the nth storage position and n.y denotes the ordinate of the nth storage position. abs denotes a function for absolute value.
In a designed warehouse space structure, at present, algorithms such as an artificial potential field method, a neural network, fuzzy logic, a-x and the like are applied to a motion planning method of a plurality of automatic conveying platforms. Among them, the a-algorithm has been widely used and can ensure that an optimal solution path is found. And (4) correcting and improving on the basis of an A-algorithm by considering the constraint of the unidirectional operation of the road in the storage space structure. The basic flow of the a-algorithm is to selectively expand nodes from a starting point according to an estimated cost until a target point is expanded. The key is to select a suitable merit function:
f(n)＝g(n)+h(n)
in the formula: f (n) represents the estimated cost of arriving at the destination point from the origin via node n, g (n) represents the true cost of arriving at node n from the origin, and h (n) represents the estimated cost of arriving at the destination point from node n.
If the specific positions of all the automatic conveying platforms at each moment are known in advance, the possible road conflicts can be predicted and avoided in the path planning. Therefore, a 'space-time running map' can be set, the running track of each robot along with the time is recorded, and the planned robot track is recorded to form a 'horizontal-vertical-time' three-dimensional map which is used as a reference for track planning. Meanwhile, on the basis of considering the distance of moving to a certain position, the waiting time cost increased by collision prevention is further added. Therefore, the conflict delay is considered in path planning, and the conflict generated by the planned path can be reduced.
Due to the unidirectional property of the road, the planning sequence cannot guarantee the priority of the operation. If a task planned later is inserted in front of a previously planned task road, it can only be waited by a previously planned task behind the road. This results in the actual trajectory not conforming exactly to the original plan. In this case, the task can still be completed according to the planned path, except that time may not be in time. If the map continues to be run according to time-biased spatio-temporal maps to plan new tasks, errors accumulate and the path is not optimal, although the task can still be completed. Therefore, the route planning correction method corrects the space-time operation map in time after each unplanned avoidance occurs, so that the map is consistent with the reality.
Further, in order to avoid a collision that may occur, an exclusive point is set. A certain position can only be occupied by one automatic conveying platform at a certain moment. Other automated transport platforms at the same time have to wait to enter this position.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A warehouse management method based on digital twin is characterized in that:
constructing a digital twin body for a physical entity of the warehouse: the construction objects mainly comprise an automatic conveying platform, a goods shelf, goods and other physical entity equipment of the warehouse; twin display of the digital twin warehouse is completed by three-dimensional modeling of various constructed objects and combining large-screen display;
establishing a mapping from a physical repository to a virtual repository: the system mainly comprises a physical entity, a virtual entity, a service, twin data and mapping connection among all components; by adopting a data driving method, the physical model can be updated and corrected by utilizing the history and real-time operation data of the system, and the real-time operation state of the warehouse is displayed;
on the basis of model construction and data mapping, time prediction of each task is realized by combining task data and warehouse basic source data through a related prediction algorithm, a data basis is provided for task scheduling and planning, and decision making of managers is assisted;
establishing a simulation virtual model of an automatic warehouse space based on the warehouse place environment, and considering the information of the position of an automatic conveying platform, the position of goods, a goods inlet, a goods outlet, a goods inlet road and a goods outlet road; meanwhile, the road is planned and defined, and vehicle collision and path conflict are avoided through intelligent scheduling and planning simulation.
2. A digital twin based warehouse management method according to claim 1, wherein: the method comprises the steps of forecasting the time required by goods to leave the warehouse, wherein the demand forecasting is positioned at the bottommost layer in the whole warehouse management system and plays a supporting role, a plurality of decision optimization systems on the upper layer are supported, the decision optimization systems obtain the optimal decision by combining accurate forecasting data with an operation research technology, and the result is provided for a service execution system on the upper layer.
3. A method for digital twin based warehouse management as claimed in claim 2 wherein:
the method for predicting the time required by goods to leave the warehouse comprises the following steps: acquiring required service data from an external data source, processing and cleaning basic data, processing and analyzing the data through a time sequence or machine learning or other artificial intelligence technologies, and finally calculating a prediction result and pushing the prediction result to a downstream module for use through multiple ways;
the whole framework of the cargo ex-warehouse demand prediction module is sequentially as follows from top to bottom: the system comprises a data source input layer, a basic data processing layer, a core service layer, a data output layer and a downstream execution module;
(1) data source input layer: the data source of warehouse management comprises most of needed business data, including goods information, warehousing information and historical ex-warehouse information; the data of the cargo consignment plan are mostly input by operators through a storage and transportation system;
(2) core service layer: the layer is a core part and mainly comprises feature construction, a prediction algorithm and prediction result processing; meanwhile, the system can also comprise a plurality of service lines which do not intersect;
the characteristic construction is to convert the cleaned basic data into characteristic data in a standard format through further processing and provide the characteristic data for a subsequent algorithm model; the core algorithm mainly utilizes time sequence analysis, machine learning or other artificial intelligence technologies to predict the goods delivery requirements; the predicted result processing is that the format and some special requirements cannot meet the requirements of downstream modules, so that the predicted result processing also needs to be processed according to actual conditions, such as additional information of standard deviation and goods delivery identification;
the goods ex-warehouse demand prediction core technology is mainly divided into a basic layer, a frame layer, a tool layer and an algorithm layer; the base layer uses HDFS for data storage, and the Yarn is used for resource scheduling and developing a related task scheduling module; the framework layer mainly comprises Spark RDD, Spark SQL and Hive; the tool layer directly uses mature and stable algorithms and models, the algorithms are all packaged in a third party Python package, and the third party Python package is selected from xgboost, numpy, pandas, sklern, scipy or hyper; the algorithm layer is mainly divided into a time sequence and a machine learning algorithm;
processing an input sequence with any time sequence by using the memory inside the RNN; the time series task prediction of the consignment mainly comprises ARIMA and Holt winters; the ARIMA is an autoregressive integral moving average model and is mainly used for predicting a stable sequence like the warehouse goods delivery single quantity; holt winters is also called cubic exponential smoothing algorithm and is used for predicting ex-warehouse cargos with seasonality and obvious trend.
4. A method for digital twin based warehouse management as claimed in claim 3 wherein: the algorithm layer is an iterative decision tree algorithm, GBDT, and the algorithm is composed of a plurality of decision trees, the decision results of all the trees are accumulated to make a final decision, and the final decision is used for predicting goods which are higher than the warehouse but have unobvious historical rules.
5. A method for digital twin based warehouse management as claimed in claim 3 wherein: the general flow of goods delivery demand prediction based on machine learning algorithm is as follows:
(1) the method comprises the following steps: determining main characteristics through data analysis and model tests, and generating characteristic data in a standard format through a series of tasks;
(2) selecting a model: different goods have different characteristics, so different algorithm models are distributed according to factors such as the historical ex-warehouse demand of the goods, the special training day sensitivity and the like;
(3) selecting characteristics: screening a batch of characteristics to filter out the characteristics which are not needed, wherein the characteristics of different types of goods are different;
(4) sample partitioning: grouping training data into a plurality of groups of samples, and generating a model file for each group of samples during training; generally, the same type of goods are divided into a group;
(5) model parameters: selecting the optimal model parameters, wherein the proper parameters can improve the accuracy of the model, and the model training and prediction are respectively carried out on different parameter combinations;
(6) model training: after the characteristics, the model and the sample are determined, model training is carried out, and after training, a model file is generated and stored in the HDFS;
(7) model prediction: reading the model file for prediction execution;
(8) multi-model optimization: in order to improve the prediction accuracy, a plurality of algorithm models are used, and an optimal prediction result is selected after the prediction result of each model is output;
(9) abnormal interception of a predicted value: the more complex and difficult-to-interpret algorithm is, the more easily the condition that the extremely individual predicted value is abnormally high occurs, and the abnormal high prediction cannot be interpreted by combining historical goods ex-warehouse demand data, so that the abnormal values can be intercepted by some rules and replaced by a more conservative numerical value;
(10) and (3) evaluating a model: calculating the prediction accuracy;
(11) and (3) error analysis: the distribution of an error in different dimensions is obtained by analyzing the prediction accuracy so as to provide a reference basis for algorithm optimization.
6. A digital twin based warehouse management method according to claim 1, wherein: the method carries out intelligent scheduling and path planning design of an automatic conveying platform on a plurality of cargos which are simultaneously delivered into and delivered out of the warehouse,
firstly, establishing a simulation virtual model of an automatic warehouse space; mainly considering the vehicle goods parking position, namely idle or goods shelf occupation, a goods inlet, a goods outlet, a goods inlet road and a goods outlet road; the warehouse model has good reconfigurability, and parameters of length and width of a warehouse space, the number and density of shelves and vehicles, the number and position of goods outlets of a goods inlet, the number and density of transportation order tasks are flexibly set according to specific requirements;
in a warehouse space, when a plurality of automatic conveying platforms operate simultaneously, in order to avoid collision conflict, simplify the operation rules of the automatic conveying platforms and improve the operation safety of the system, transverse roads and longitudinal roads among shelf areas are set to be one-way roads;
for the established warehouse space model, the task is in a form of conveying a certain shelf from a parking position to a certain goods inlet/outlet position, and after the goods inlet/outlet task is completed, the shelf is conveyed back to the position of the shelf area;
the task can be broken down into the following 3 steps:
step 1: the automatic conveying platform moves to a task corresponding shelf;
step 2: after the automatic conveying platform is filled, the automatic conveying platform moves to a goods inlet/outlet;
step 3: the automatic conveying platform moves to a proper vacant position to wait for unloading;
aiming at the 1 st step of the logistics task, evaluating all automatic conveying platforms which can execute the task according to an evaluation function, selecting the most appropriate carrier from the evaluation functions, and defining the evaluation function as follows:
gn＝w*tn1*tn2
the above equation represents the total cost, tn, for the nth automated transport platform to perform this task1Indicating the nth automated conveying platformThe time expected to be consumed for completing the task currently running; w represents a congestion coefficient used for reflecting the congestion degree of the system; set up w>1 may reflect that the actual time taken to complete the current task is more than expected; tn2Indicating the time spent by the nth automated conveying platform to run to the task-requiring shelf;
dividing all automatic conveying platforms into two types of executing the current task and being in an idle state according to an evaluation function, respectively calculating waiting cost and path cost, evaluating all automatic conveying platforms according to the sum of the waiting cost and the path cost, and selecting the automatic conveying platform with the minimum total cost gn to bear the transportation task;
aiming at the step 3, the goods shelves carried by the automatic conveying platform are parked in the idle area as soon as possible, so that the automatic conveying platform can continue to execute the next task; therefore, selecting an idle position closest to the starting point for parking; cost estimation using manhattan distance:
gn＝abs(cur.x-n.x)+abs(cur.y-n.y)
in the formula: gnRepresenting the cost of parking to the nth storage location;
cur.x represents the abscissa of the current point, and cur.y represents the ordinate of the current point;
x denotes the abscissa of the nth storage position, n.y denotes the ordinate of the nth storage position; abs represents a function for absolute value;
in the designed warehouse space structure, the method applied to the motion planning of a plurality of automatic conveying platforms is a modified and improved A-algorithm; the basic flow of the A-algorithm is that the nodes are selectively expanded from the starting point according to the estimated cost until the target point is expanded; the key is to select a suitable merit function:
f(n)＝g(n)+h(n)
in the formula: (n) represents the estimated cost from the origin to the destination via node n, (g) (n) represents the true cost from the origin to node n, and h (n) represents the estimated cost from node n to the destination;
if the specific positions of all automatic conveying platforms at each moment are known in advance, possible road conflicts can be predicted and avoided during path planning; setting a space-time running map, recording the running track of each robot along with the time, and recording the planned robot track to form a three-dimensional map of horizontal-vertical-time as a reference for track planning; meanwhile, on the basis of considering the distance of the vehicle moving to a certain position, the waiting time cost increased by collision prevention is further added;
the route planning correcting method corrects the space-time operation map in time after each unplanned avoidance occurs, so that the map conforms to the reality.
7. A method for digital twin based warehouse management as claimed in claim 6 wherein: in order to avoid possible collision, an exclusive point is set; at a certain moment, a certain position can only be occupied by one automatic conveying platform independently; other automated transport platforms at the same time have to wait to enter this position.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210108546.3A CN114548861A (en) | 2022-01-28 | 2022-01-28 | Warehouse management method based on digital twin |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210108546.3A CN114548861A (en) | 2022-01-28 | 2022-01-28 | Warehouse management method based on digital twin |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114548861A true CN114548861A (en) | 2022-05-27 |
Family
ID=81673673
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210108546.3A Pending CN114548861A (en) | 2022-01-28 | 2022-01-28 | Warehouse management method based on digital twin |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114548861A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115421500A (en) * | 2022-11-04 | 2022-12-02 | 北自所(北京)科技发展股份有限公司 | Automatic loading and unloading vehicle control method and system based on digital twin model |
| CN116703301A (en) * | 2023-06-15 | 2023-09-05 | 山东盛太锆业资源有限公司 | Zirconia waste material management and control method based on information of multiple recovery warehouses |
| CN117196150A (en) * | 2023-09-19 | 2023-12-08 | 西湾智慧(广东)信息科技有限公司 | Intelligent park center platform system based on digital twinning |
| CN117196480A (en) * | 2023-09-19 | 2023-12-08 | 西湾智慧(广东)信息科技有限公司 | Intelligent logistics park management system based on digital twinning |
| CN117236855A (en) * | 2023-11-13 | 2023-12-15 | 山东朝辉生物科技有限公司 | Biological feed warehouse management system and method |
| CN117333114A (en) * | 2023-11-30 | 2024-01-02 | 广州欧派集成家居有限公司 | Household product intelligent sorting, storing and managing system and method |
| CN117829723A (en) * | 2023-12-18 | 2024-04-05 | 慧之安信息技术股份有限公司 | Warehouse management simulation method and system based on industrial meta universe |
| CN117910929A (en) * | 2024-03-14 | 2024-04-19 | 浙江菜鸟供应链管理有限公司 | Storage system all-link processing method and storage system all-link simulation platform |
| CN118485369A (en) * | 2024-07-12 | 2024-08-13 | 山东港源管道物流有限公司 | Liquid cargo safety analysis method based on digital twin technology |
-
2022
- 2022-01-28 CN CN202210108546.3A patent/CN114548861A/en active Pending
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115421500A (en) * | 2022-11-04 | 2022-12-02 | 北自所(北京)科技发展股份有限公司 | Automatic loading and unloading vehicle control method and system based on digital twin model |
| CN116703301B (en) * | 2023-06-15 | 2023-12-08 | 山东盛太锆业资源有限公司 | Zirconia waste material management and control method based on information of multiple recovery warehouses |
| CN116703301A (en) * | 2023-06-15 | 2023-09-05 | 山东盛太锆业资源有限公司 | Zirconia waste material management and control method based on information of multiple recovery warehouses |
| CN117196150B (en) * | 2023-09-19 | 2024-04-16 | 西湾智慧(广东)信息科技有限公司 | Intelligent park center platform system based on digital twinning |
| CN117196480A (en) * | 2023-09-19 | 2023-12-08 | 西湾智慧(广东)信息科技有限公司 | Intelligent logistics park management system based on digital twinning |
| CN117196150A (en) * | 2023-09-19 | 2023-12-08 | 西湾智慧(广东)信息科技有限公司 | Intelligent park center platform system based on digital twinning |
| CN117196480B (en) * | 2023-09-19 | 2024-05-03 | 西湾智慧(广东)信息科技有限公司 | Intelligent logistics park management system based on digital twinning |
| CN117236855A (en) * | 2023-11-13 | 2023-12-15 | 山东朝辉生物科技有限公司 | Biological feed warehouse management system and method |
| CN117333114A (en) * | 2023-11-30 | 2024-01-02 | 广州欧派集成家居有限公司 | Household product intelligent sorting, storing and managing system and method |
| CN117333114B (en) * | 2023-11-30 | 2024-04-02 | 广州欧派集成家居有限公司 | Household product intelligent sorting, storing and managing system and method |
| CN117829723A (en) * | 2023-12-18 | 2024-04-05 | 慧之安信息技术股份有限公司 | Warehouse management simulation method and system based on industrial meta universe |
| CN117910929A (en) * | 2024-03-14 | 2024-04-19 | 浙江菜鸟供应链管理有限公司 | Storage system all-link processing method and storage system all-link simulation platform |
| CN118485369A (en) * | 2024-07-12 | 2024-08-13 | 山东港源管道物流有限公司 | Liquid cargo safety analysis method based on digital twin technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114548861A (en) | Warehouse management method based on digital twin | |
| Iglesias et al. | Data-driven model predictive control of autonomous mobility-on-demand systems | |
| Cattaruzza et al. | The multi-trip vehicle routing problem with time windows and release dates | |
| US20220156693A1 (en) | Computerized system and method for developing optimized cargo transportation solutions | |
| Rizzoli et al. | Ant colony optimization for real-world vehicle routing problems: from theory to applications | |
| Powell | An operational planning model for the dynamic vehicle allocation problem with uncertain demands | |
| Sáez et al. | Hybrid adaptive predictive control for the multi-vehicle dynamic pick-up and delivery problem based on genetic algorithms and fuzzy clustering | |
| CN112001064B (en) | Full-autonomous water transport scheduling method and system between container terminals | |
| Goodarzi et al. | A new bi-objective vehicle routing-scheduling problem with cross-docking: Mathematical model and algorithms | |
| Lee et al. | Vehicle capacity planning system: A case study on vehicle routing problem with time windows | |
| Ghannadpour et al. | A multi-objective vehicle routing and scheduling problem with uncertainty in customers’ request and priority | |
| US20080010357A1 (en) | System, computer-implemented method, and software for vessel scheduling for product distribution | |
| Li et al. | Learning to optimize industry-scale dynamic pickup and delivery problems | |
| Kilic et al. | A mathematical model and a heuristic approach for periodic material delivery in lean production environment | |
| Ancele et al. | Toward a more flexible VRP with pickup and delivery allowing consolidations | |
| Jiang et al. | Picking-replenishment synchronization for robotic forward-reserve warehouses | |
| Chargui et al. | Berth and quay crane allocation and scheduling with worker performance variability and yard truck deployment in container terminals | |
| Chen et al. | A variable neighborhood search algorithm with reinforcement learning for a real-life periodic vehicle routing problem with time windows and open routes | |
| Zou et al. | An effective discrete artificial bee colony algorithm for scheduling an automatic-guided-vehicle in a linear manufacturing workshop | |
| CN115619198B (en) | Library displacement dynamic programming method, device, electronic equipment and storage medium | |
| CN113848970A (en) | Multi-target collaborative path planning method for vehicle and unmanned aerial vehicle | |
| Wu et al. | A disruption recovery model for time-dependent vehicle routing problem with time windows in delivering perishable goods | |
| EL Bouyahyiouy | The selective full truckload multi-depot vehicle routing problem with time windows: Formulation and a genetic algorithm | |
| CN116976069A (en) | Logistics park and warehouse decision support software platform system | |
| Touzout et al. | An assign-and-route matheuristic for the time-dependent inventory routing problem |
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 |