CN114037352A

CN114037352A - Automatic container terminal multi-AGV dynamic scheduling method based on digital twinning

Info

Publication number: CN114037352A
Application number: CN202111456026.3A
Authority: CN
Inventors: 苌道方; 李玉; 孙苗苗; 徐国轩; 高银萍; 陆后军; 田宇; 凌强
Original assignee: Shanghai Maritime University
Current assignee: Shanghai Maritime University
Priority date: 2021-12-02
Filing date: 2021-12-02
Publication date: 2022-02-11

Abstract

The invention discloses a digital-twin-based automatic container terminal multi-AGV dynamic scheduling method which comprises a multi-AGV dynamic scheduling digital-twin frame construction process, an AGV scheduling model construction and scheme generation process and a digital-twin-based multi-AGV dynamic scheduling strategy implementation process. The invention introduces digital twin into the AGV dispatching problem of the current automatic container terminal, and utilizes the closed loop interaction and iterative optimization of the physical space and twin space of the terminal to improve the dispatching efficiency and reduce the dispatching deviation. On one hand, the scheme can be verified and optimized before scheduling execution, and the optimality of the scheduling scheme is ensured; on the other hand, the method can sense abnormal disturbance and analyze and process the abnormal disturbance in the process of scheduling execution, call an algorithm to update a scheduling scheme, ensure the consistency of the physical space and the twin space and ensure that the scheduling scheme can meet the actual production requirement.

Description

Automatic container terminal multi-AGV dynamic scheduling method based on digital twinning

Technical Field

The invention relates to the field of ship shipping, in particular to a digital twin-based automatic container terminal multi-AGV dynamic scheduling method.

Background

The container terminal is an important node for international transportation and is also an important hub for sea and land cargo transportation. With the deep construction of automated container terminals, the transition from automation to intelligence will be a great trend for container terminals.

However, in the face of wharf working environment with strong production continuity and fast environmental change, when a certain part of equipment fails, the whole working process is affected, and timely response of dynamic events in the working process is the most troublesome problem at present; secondly, the current wharf information has low transparency, and the making and implementation of the dispatching plan are seriously hindered when some uncertain factors appear in production. In addition, the current wharf lacks effective simulation tools, information models and independent decision-making mechanisms, so that the scheduling in the working process is extremely difficult to realize in a flexible manner. The occurrence of digital twins makes possible the interaction of twin spaces with physical spaces. Meanwhile, in the process of dispatching operation by the AGV, a twin space is established to realize iterative interaction with actual production and timely respond to dynamic events in dispatching, so that the production information of the wharf is necessary to be transparent. Therefore, the invention combines digital twin with wharf AGV scheduling, and provides an automatic container wharf multi-AGV dynamic scheduling method based on the digital twin. Compared with the current wharf operation, the digital twin improves the wharf operation efficiency, and is an important research direction for the future improvement of the automatic container wharf.

Currently, the research on the AGV dispatching of the automated container terminal has the following defects:

(1) AGV scheduling is a key link of operation of an automatic container terminal, and the efficiency of the AGV scheduling is also one of decisive factors influencing the overall operation efficiency of the terminal. When most scholars study the scheduling problem of the AGV, the modeling process depends on certain specific assumptions (such as equipment is available all the time, the operation time is constant, and the like), and most dynamic factors considered are not from actual production data and do not accord with the actual conditions of wharf operation; in addition, when some scholars study the AGV dynamic scheduling, the influence of the rescheduling stability on wharf operation is ignored, and the stability of the AGV dynamic scheduling cannot be ensured.

(2) In the research of AGV dynamic scheduling, a virtual scene interacting with iteration of actual production is lacked, so that the problem of dynamic scheduling cannot be well solved, when a certain part fails, the whole operation process is affected, and the problem of timely responding to a dynamic event in the operation is the most troublesome problem at present. In addition, the current wharf lacks effective simulation tools, information models and independent decision-making mechanisms, so that the scheduling in the working process is extremely difficult to realize in a flexible manner.

(3) In the current wharf scheduling operation process, the transparency of production information is low, and information islands exist among systems; scheduling plans are severely hampered in their formulation and implementation when some uncertainty arises in production. Moreover, in the actual production scheduling process, the uncertain events, information asymmetry and abnormal disturbance conditions can cause execution deviation, and the efficiency and stability of scheduling execution are affected.

Disclosure of Invention

The invention aims to provide a digital twin-based automatic container terminal multi-AGV dynamic scheduling method, which is used for researching three problems mentioned in the background technology and researching the dynamic scheduling problem of the automatic container terminal multi-AGV by using the virtual-real interaction characteristic of the digital twin under the condition of considering the influence of rescheduling stability and uncertain factors. The invention combines digital twin with wharf AGV dispatching, provides a multi-AGV dynamic dispatching digital twin framework, considers dispatching stability in dispatching problem modeling, and researches an automatic container code multi-AGV dynamic dispatching method based on digital twin. The digital twin and the genetic algorithm are combined and used for automatic container terminal AGV dispatching operation, the virtual-real interaction characteristic of the digital twin is utilized to respond to disturbance occurring in a physical space, a corresponding dispatching model and the genetic algorithm are called, a dispatching scheme is adjusted and updated, and the current operation of a terminal is guided.

In order to achieve the purpose, the technical scheme adopted by the invention is as follows:

a digital twin-based automatic container terminal multi-AGV dynamic scheduling method comprises the following steps: a multi-AGV dynamically schedules a digital twin frame construction process; an AGV scheduling model construction and scheme generation process; and a digital twin-based multi-AGV dynamic scheduling strategy implementation process. The invention introduces the digital twin technology into the AGV scheduling problem of the current automatic container terminal, and improves the scheduling efficiency and reduces the scheduling deviation by utilizing the closed loop interaction and the iterative optimization of the physical space and the twin space of the terminal. On one hand, the scheme can be verified and optimized before scheduling execution, and the optimality of the scheduling scheme is ensured; on the other hand, the method can sense abnormal disturbance and analyze and process the abnormal disturbance in the process of scheduling execution, call an algorithm to update a scheduling scheme, ensure the consistency of the physical space and the twin space and ensure that the scheduling scheme can meet the actual production requirement.

The invention relates to a digital twin-based automatic container terminal multi-AGV dynamic scheduling method, wherein a multi-AGV dynamic scheduling digital twin frame construction process specifically comprises the following steps:

s1: and constructing a wharf twin space, wherein the wharf twin space is a mirror image of a wharf physical space. The physical entities of the wharf physical space mainly comprise a shore bridge, a field bridge, an AGV, a ship, a storage yard, a container, a driving path and the like, a 3D MAX tool is used for establishing multi-dimensional digital twin models of the physical entities, and the multi-dimensional digital twin models are combined according to the real scene of the wharf to form a wharf twin space;

s2: constructing a dynamic digital twin model in the process of scheduling operation, wherein the dynamic digital twin model mainly comprises an AGV health state model, an AGV loading and unloading efficiency model and an AGV energy consumption model; the AGV health state modeling is to complete the mapping between data and states by acquiring data including motor current, voltage and speed in the AGV operation process and establishing equipment state health factors by utilizing a long-time memory network, as shown in a formula (1-5).

Y_i＝t_i*tanh(Z_i) (1)

f_i＝σ(W_c·[Y_i-1,X_i]+b_c) (3)

s_i＝σ(W_c·[Y_i-1,X_i]+b_c) (4)

Where i denotes the number of AGVs used, Y_iRepresenting the output result, X, of the i-th motor state data after operation_iRepresents the input corresponding current, voltage, speed data of the i-th AGV Motor, t_iIndicating the non-operating time, Z, of the ith AGV in the job_iMotor status data representing the ith AGV f_iFor a logistic regression function, it is decided how much information in the state of the machine at the previous moment can be transmitted to the next moment, σ is a logistic function for calculating a value between 0 and 1, W_cAs a weight matrix, b_cIs a bias that is a function of the bias,

representing to obtain a new motor state value; the AGV loading and unloading efficiency model is used for screening AGV transportation time data in scheduling by utilizing correlation analysis and establishing a mapping relation between time data characteristics and efficiency, and is shown in a formula (6).

Where V denotes the number of AGVs, p denotes the loading and unloading efficiency, M_cTotal number of containers to be loaded or unloaded, T_fiFor the ith AGV completion time, T_siFor the ith AGV start time, t_iThe non-working time of the ith AGV; the AGV energy consumption model is an equipment energy consumption model constructed according to data such as the AGV operation state and the number of tasks completed in unit time, and is shown in a formula (7).

Wherein E_VRepresenting total energy consumption of the AGV, N representing the number of container tasks, Q representing the number of shore bridges, B representing the number of yard bridges, θ_e、θ_d、θ_wThe average energy consumption per unit time t is the average energy consumption of the AGV in an idle running state, a load running state and a waiting state_ajj'time taken for AGV a to move from the end of container task j to the start of another container task j', T_ajThe time taken for the AGV a to travel from the start node to the end of task j' when transporting the container task j,

indicating the time it takes for the AGV a to wait for the shore bridge m while transporting the container task j,

represents the time it takes for the AGV a to wait for the bridge b while transporting the container task j; the AGV in operation can be simulated and optimized in real time by constructing a dynamic AGV health state model, an AGV loading and unloading efficiency model and an AGV energy consumption model;

s3: and (5) verifying the consistency of the virtual reality and the real reality of the wharf multi-dimensional digital twin model. The application refers to Tao Fei et al [1]]Henwei et al [2 ]]And (5) verifying the false-true verification rule of the digital twin model, and completing the false-true consistency verification of the multi-dimensional digital twin model in the twin space. CodeThe key point of the head twin space construction lies in the aspect of verifying the consistency of the virtual and the real of the digital twin model, and the validity, the correctness and the accuracy of the model are ensured by verifying the consistency of the virtual and the real. The invention completes the consistency verification of the model between the virtuality and the reality from two aspects. On the one hand, the evaluation index theta is passed^eEvaluating similarity of the model and the layout; on the other hand, whether the output of the wharf twin space can sufficiently reflect the behavior characteristics of the wharf physical space is verified, and the evaluation index theta is used^bAnd judging whether the same feature number of the dock physical space and the twin space and the importance degree of the features meet the requirements of virtual and real consistency. Model hierarchy analysis is adopted for calculating the similarity of the models. A shore bridge model, an AGV model, a field bridge model, a container model and other models are constructed on the automatic container terminal, and the similarity calculation of the models is specifically realized according to the formulas 8-13.

Q_i＝rank A_i (12)

In which ξ_ijIndicates the alternative xi of AGV or shore bridge model in the model base_iOf the jth aspect of (e.g., color, profile, surface roughness, etc.), beta_ijA value representing a wharf model alternative evaluation normalization; omega_jRepresenting the weight of each model attribute of the wharf; c_iAnd P_iRespectively representing alternatives xi_iDistance from the negative and positive optimal solutions. σ is the majority criterion weight, and when σ is greater than 0.5, A_iThe representation model has higher similarity; when σ is less than 0.5, A_iThe display model similarity is low. Q_iIs to rank the alternatives, θ^e1The model similarity evaluation index is represented, and the value thereof is 0 or 1.

The threshold is variable and can be modified according to actual needs. When Q is_iGreater than a threshold value

When theta is greater than theta^e1Is 1, otherwise is 0.

Layout similarity requires consideration of the position data of the berths, buffers and shellfishes. Thus will theta^e2As an index for layout similarity verification, as is the relative distance between the berths, buffers, and shelfs. QC (quasi-cyclic)_i(x_i,y_i)i∈[1,i^*]I ∈ Z denotes the coordinates of the ith berth, i^*The number of quay bridges; v_j(x_j,y_j)j∈[1,j^*]J ∈ Z denotes the jth buffer coordinate, j^*The number of buffer areas; b is_k(x_k,y_k)k∈[1,k^*]K ∈ Z denotes the coordinate of the kth beta-position, k^*Indicating the number of decibels. QV_ijRepresenting the distance between the ith berth and the jth buffer area in the wharf physical space; VB_ikThe distance between the ith berth and the kth berth in the wharf physical space is expressed by the following formula. The distance between the models in the twin space can be obtained by direct measurement.

VQV_ijAnd VVB_ikRespectively representing the distances between the berth and the buffer zone of the twin model and between the berth and the berth in the dock twin space. QPE_ijRepresenting a position error between a berth in the wharf twin space and a wharf physical system with respect to a buffer; BPE_ikThe position error between the berth and the dock physical system in the dock twin space is expressed by the following formula, wherein

The adjustable error threshold can be set according to actual requirements.

QPE_ij＝|QV_ij-VQV_ij| (16)

BPE_ik＝|VB_ik-VVB_ik| (17)

Thus, θ^e＝θ^e1∩θ^e2When theta is^eWhen 1, the wharf physical space and the twin space have the same model similarity and layout similarity.

And verifying whether the output of the twin space of the wharf can sufficiently reflect the behavior characteristics of the physical space of the wharf or not by adopting a high-order singular value decomposition method. Supposing that under the same production operation scene of the wharf, a scheduling scheme is used as input, equipment, container position data and the like generated by a wharf physical space and a twin space in the operation process are stored in data tensors, wherein the two space data tensors are respectively omega^pAnd Ω^vIs subjected to modulo-n expansion to obtain

And

then singular value decomposition is carried out to respectively obtain left singular matrixes

And

right singular matrix

And

wherein ∑ is ═ x₁,χ₂,…,χ_k,χ_k+1,…,χ_h) K is less than or equal to h is less than or equal to min (m, n), the elements on the sigma diagonal are singular values, which represent the importance of the corresponding feature, and the values are arranged from large to small. Since the values in Σ are decreasing and, to reduce the amount of computation, the matrix Σ can be approximately described with singular values of k terms,

judging the physical space characteristics of the wharf by solving the characteristic values of the physical space and the twin space

And twinning space characteristics

Whether or not equal. Sigma_allRepresenting the number of identical features in the first k features of the two spaces.

Wherein theta is^σIs an evaluation index for judging the same characteristic quantity of the consistency between the reality and the virtuality when theta is^σBy 1 is meant that both spaces have the same first k main features. Theta^γIn order to judge the evaluation index of the feature importance,

for judging the degree of importance of each feature,

the threshold value can be determined according to actual requirements, and is specifically realized as follows. Alpha is alpha_allRepresenting the number of features with the same importance ratio of the first k features in both spaces. If the number of the features with the same importance ratio in the first k main features is k, the dock physical space and the twin space have the same main features.

Therefore, when theta^b＝θ^σ∩θ^γWhen theta is^bWhen 1, the wharf physical space and the twin space have the same behavior characteristics.

In summary, the evaluation criterion of the consistency between real and virtual is θ^e∩θ^bWhen theta is^e∩θ^bWhen the number is 1, the constructed wharf twin space meets the verification of virtual and real consistency.

S4: and completing the construction of a multi-AGV dynamic dispatching digital twin frame of the automatic container terminal on the basis. The framework mainly comprises a wharf physical space, a data service platform, an AGV dispatching digital twin system and a connection part, and the formal expression of the framework is shown as a formula (27). The invention combines the twin space and the wharf service system to form an AGV dispatching digital twin system. The twin space completes model updating by combining historical data on the basis of real-time data, and virtual-real synchronization is realized. And transmitting the new data generated by updating to the data service platform for storage. Meanwhile, the service system can generate a scheduling scheme by combining a corresponding model and an algorithm library on the basis of real-time data, and realize services such as scheduling process information statistics, equipment monitoring and the like. And the generated scheduling scheme is subjected to simulation verification in a twin space, and after an optimal scheduling scheme is obtained, the AGV scheduling digital twin system drives the wharf physical space to operate through the data service platform.

DTF∷＝{TPS,DSP,VSS,CON} (27)

Wherein DTF represents the digital twin framework, and: ═ representation is defined such that TPS represents the dock physical space, DSP represents the data service platform, VSS represents the AGV dispatching digital twin system, and CON represents the connections between the parts.

The invention relates to a digital twin-based automatic container terminal multi-AGV dynamic scheduling method, wherein an AGV scheduling model construction process specifically comprises the following steps:

s5: and generating an AGV dispatching scheme. The initial input data is data such as container tasks, AGV, quay crane and field bridge number, influence factors of the self-adaptive crossing rate of the genetic algorithm, the population scale of the influence factors of the self-adaptive variation rate, the maximum iteration times and the like. First, scheduling with AGV completion time and stability as optimization objectivesModels, and adds constraints that an AGV can only process one container task at a certain point in time. Then according to formula

The time taken for each AGV to complete a transport container task is calculated,

indicating the time it takes for the AGV a to transport the container from the shore bridge q to the destination compartment,

indicating the time taken by the AGV a from the destination box area of one yard b to the next designated box area in the case of no load;

indicating the time it takes for the AGV a to transport a container from a certain block of the yard bridge b to the destination shore bridge q. According to the calculation result, the container task is allocated to T_iaThe lowest value AGV. And finally, on the basis of the distribution result of the container tasks, solving the scheduling model by using a genetic algorithm on the premise of meeting all constraint conditions, thereby obtaining the sequence of the AGV transporting the container tasks and further generating a scheduling scheme.

Establishing a scheduling model taking AGV completion time and stability as optimization targets, wherein parameters of the model are as follows: i represents an import container set, namely 1,2,3 … I belongs to I; e represents the set of export containers, namely 1,2,3 … j ∈ E; n denotes the set of all containers, N ═ I ═ E; q represents a set of quayside container QCs, and m, l belongs to Q; v represents an AGV set, and B represents a set of a bridge; n is a radical of_qRepresenting nodes of intersection of the shore bridge sides and the road network; n is a radical of_bRepresenting nodes where AGV partners intersect the road network; n 'represents other nodes except for nodes intersected with a shore bridge and an AGV partner, and c, d belongs to N'; n is a radical of^*Representing a set of all nodes, N^*＝N′∪N_q∪N_b；T₁Indicating the time it takes for the quayside crane to place or grasp a container to or from the AGV; t is₁Placing or receiving an AGV partner for AGV aThe time it takes for the partner to acquire the container; s_a,imRepresents the time when the AGV a starts to transport QC m containers i; f. of_a,imIndicating the completion time of the AGV a transporting the last container i; p is a radical of_a,imRepresenting the processing time during which the AGV a transports the container; h is_a,imRepresenting the running time from the AGV a to the AGV partner for transporting the QCm container i;

representing the time taken for the AGV a to transport the container i from the shore bridge to the yard bridge, q, b ∈ N^*；t_a,cdIndicating the travel time at nodes c and d when AGV a is empty,

or

M represents a very large positive number;

the values representing the changes can be modified according to actual requirements.

0-1 decision variables:

representing that the AGV a processes a QCm container i, a belongs to V, i belongs to N, and m belongs to Q; x_imjlRepresenting that the AGV a processes a QC m container i and then processes a QCl container j, wherein a belongs to V, i, j belongs to N, and m belongs to Q;

indicating the order in which AGV a processes container i and container i ', a ∈ V, i, i' ∈ N, and m ∈ Q. Non-0-1 decision variables: d_iaIndicating whether container i has changed the corresponding AGV a after rescheduling.

Focusing the dispatching problem of the AGV of a typical automatic container terminal, aiming at improving the operation efficiency of the terminal and ensuring the stability in the process of rescheduling, the invention takes the completion time and the stability of the AGV as targets, and ensures the stability of the AGV by minimizing the deviation degree of a rescheduling scheme on the basis of meeting the requirement of minimizing the completion time of the AGV, thereby embodying the stability of the digital twin-based multi-AGV dynamic dispatching;

the formula (28) is an objective function, and aims to ensure the stability of the dispatching system by minimizing the deviation degree of a rescheduling scheme on the basis of meeting the requirement of minimizing the completion time of the AGV; equation (29) indicates that the completion time for transporting each container is less than or equal to the completion time for the last container; expressions (30) to (31) indicate that each AGV has an initial task of 0 and an end task of a virtual task f; equation (32) represents the sequence of operation of the AGV, ensuring that there is only one container task before and after the container task currently being processed by the AGV; equation (33) indicates that any container can only be handled by one AGV; equation (34) each AGV can transport one container at a time, and both equations ensure uniqueness between AGVs and containers. Equations (35) (36) ensure that the AGV begins the loading and unloading ship sequence. The formula (35) is that the same AGV completes a ship unloading task of the shore bridge and then completes a ship loading task of the shore bridge; and (36) after the same AGV finishes the loading task of a shore bridge, the same AGV finishes the unloading task of the shore bridge. Equation (37) the time relationship of the AGV a from the start of unloading the container onto the container mate; equation (38) time constraints from one AGV mate to another AGV mate during an empty AGV condition; equation (39) the time relationship between the loading of the container by the AGV a from the AGV partner to the shore bridge; equation (40) the time constraint from one shore bridge to another shore bridge for the empty AGV a condition; equation (41) is a non-negative constraint in the AGV scheduling problem.

The invention relates to a digital twin-based automatic container terminal multi-AGV dynamic scheduling method, which specifically comprises the following steps of:

s6: on the basis of establishing a wharf digital twin frame and a multi-AGV dispatching model, in the face of influence of uncertain factors in the AGV dispatching process, the digital twin frame is dynamically dispatched by the AGV, and related services in a system are driven by fusing real-time data and historical data generated in AGV operation through a data service platform, so that multi-AGV dynamic dispatching is realized. Before the start of scheduling, the scheduling scheme generated by the dock service system is transmitted to the twin space for verification, and the twin space simulates the whole flow of the scheduling scheme, so that the completion time (f) of each AGV is obtained according to step S2 (f)_a,im-s_a,im) Loading and unloading efficiency p of each AGV and total AGV energy consumption E of scheduling scheme_V(ii) a Judging whether the twin space simulation result meets the production target, if so, finishing the evaluation of a scheduling scheme, wherein the scheduling scheme can guide the operation of a physical space; if not, the process returns to step S5 to regenerate the new scheduling scheme, and the regenerated new scheduling scheme is verified again in the twin space. And iterating and interacting in the above way until an optimal scheduling scheme is obtained to guide the operation of the wharf physical space. And the AGV dispatching system generates service through a dispatching scheme according to the operation requirements and the field resources to generate an initial dispatching scheme. The initial scheduling scheme is subjected to simulation verification in a twin space, and a verified result is fed back to the scheduling system. And the scheduling system calls a scheduling algorithm to modify the scheme according to the feedback result, generates a new scheme, and drives the field operation to execute after the verification is passed.

S7: and in the process of executing the scheduling scheme, sending the verified scheduling scheme to a physical space. Collecting motor current, voltage, speed, container position data and the like in the AGV operation process in the scheduling process, and forming real-time operation data with unified data structure and data type after the real-time data are processed, stored and mapped by the data service platform; meanwhile, the twin space and the wharf physical space are driven to keep synchronous by the real-time data, and the wharf scheduling operation can be monitored in real time by the twin space under the driving of the data.

S8: in the dispatching process, the AGV dispatching system can call the wharf production process state monitoring service by using real-time data to judge whether disturbance exists or not. And judging whether disturbance exists in the operation process. If the disturbance does not exist, judging whether the scheduling task is finished; if not, the current scheduling scheme is continuously executed, otherwise, the system calls production information statistical service to output information such as completion time, failure rate and the like in the scheduling process.

S9: when abnormal disturbance occurs in the wharf physical space, the system can update the current scheduling scheme according to the disturbance condition; meanwhile, the disturbance data is transmitted to the twin space through the data service platform, the twin space processes the abnormal data, analyzes the specific type of dynamic event, feeds back the result to the AGV scheduling system, and returns to the step S5 to regenerate the new scheduling scheme. In step S6, a simulation verification is performed on the new scheduling scheme in the twin space, and the verified scheduling scheme is compared with the target value of the currently executed scheme, so as to determine whether a rescheduling needs to be triggered. If the rescheduling is not required to be triggered, the current scheme is continuously executed until the task is completed; if necessary, rescheduling the remaining uncompleted tasks, replacing the updated scheduling scheme currently being executed with the real-time scheduling scheme, and returning to step S8 to continue execution. The operation is repeated in a circulating way until the task is completely finished. And after the task is finished, a Gantt chart of a scheduling scheme, an allocation table of container tasks in the scheduling scheme, the completion time of each AGV, the completion time of the total scheduling scheme, a stable value of the scheduling scheme and an algorithm performance iteration chart are output.

Compared with the prior art, the invention has the advantages or positive effects

The advantages are that:

(1) the invention provides a general architecture of digital twin-based automatic container terminal AGV multi-dynamic scheduling, and researches on the operation mechanisms of digital twin model construction, twin model consistency verification and digital twin-based automatic container terminal AGV multi-dynamic scheduling strategies.

(2) The method considers the influence of the scheduling stability on the scheduling process in the modeling of the scheduling problem, and finds the optimal scheduling scheme by taking the maximum completion time and the scheduling stability as targets; an AGV dynamic scheduling strategy based on a digital twin is designed to realize timely response to uncertain events in scheduling. The research of the invention can promote the improvement of the operation efficiency of the automatic container terminal.

(3) The invention considers the problem of information isolated island between the current wharf systems, fuses data between the systems by using the virtual-real interaction characteristic of digital twin, enhances the information interaction between the systems and provides a new thought for production information transparence.

Drawings

FIG. 1 is a flow chart of a digital twin based method for dynamic scheduling of multiple AGVs in an automated container terminal according to the present invention;

FIG. 2 is a digital twin frame for automatic container terminal multi-AGV dynamic scheduling based on the digital twin automatic container terminal multi-AGV dynamic scheduling method of the present invention;

FIG. 3 is a layout diagram of an automated container terminal of the digital twin-based automated container terminal multiple AGV dynamic scheduling method of the present invention;

FIG. 4 is a flow chart of the digital twin based multi-AGV dynamic scheduling method of the digital twin based automated container terminal of the present invention;

FIG. 5 is a schematic diagram of chromosome coding of the digital twin-based automatic container terminal multi-AGV dynamic scheduling method of the present invention;

FIG. 6 is a diagram of an initial schedule generation scheme for the digital twin based automated container terminal multiple AGV dynamic scheduling method of the present invention (without the use of digital twins);

FIG. 7 is a diagram of the digital twin-based initial scheduling scheme of the digital twin-based automatic container terminal multi-AGV dynamic scheduling method of the present invention;

FIG. 8 is a comparison of AGV completion times for two initial scheduling schemes of the digital twin-based automatic container terminal multiple AGV dynamic scheduling method of the present invention;

FIG. 9 is a diagram of a digital twin based final scheduling scheme of the digital twin based automatic container terminal multiple AGV dynamic scheduling method of the present invention;

FIG. 10 is an AGV completion time of the digital twin based final scheduling scheme of the digital twin based automated container terminal multiple AGV dynamic scheduling method of the present invention;

FIG. 11 is the performance of a genetic algorithm in an embodiment of the digital twin based automated container terminal multiple AGV dynamic scheduling method of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings 1 to 8 and the detailed description thereof. The advantages and features of the present invention will become more apparent from the following description. It is to be noted that the drawings are in a very simplified form and are all used in a non-precise scale for the purpose of facilitating and distinctly aiding in the description of the embodiments of the present invention. To make the objects, features and advantages of the present invention comprehensible, reference is made to the accompanying drawings. It should be understood that the structures, ratios, sizes, and the like shown in the drawings and described in the specification are only used for matching with the disclosure of the specification, so as to be understood and read by those skilled in the art, and are not used to limit the implementation conditions of the present invention, so that the present invention has no technical significance, and any structural modification, ratio relationship change or size adjustment should still fall within the scope of the present invention without affecting the efficacy and the achievable purpose of the present invention.

With reference to fig. 1, the invention discloses a digital twin-based automatic container terminal multi-AGV dynamic scheduling method, which comprises a multi-AGV dynamic scheduling digital twin frame construction process, an AGV scheduling model construction and scheme generation process, and a digital twin-based multi-AGV dynamic scheduling policy implementation process. The invention introduces the digital twin technology into the AGV scheduling problem of the current automatic container terminal, and improves the scheduling efficiency and reduces the scheduling deviation by utilizing the closed loop interaction and the iterative optimization of the physical space and the twin space of the terminal. On one hand, the scheme can be verified and optimized before scheduling execution, and the optimality of the scheduling scheme is ensured; on the other hand, the method can sense abnormal disturbance and analyze and process the abnormal disturbance in the process of scheduling execution, call an algorithm to update a scheduling scheme, ensure the consistency of the physical space and the twin space and ensure that the scheduling scheme can meet the actual production requirement.

In the operation of the automatic container terminal, AGV scheduling is a key link of the operation of the automatic container terminal, and the efficiency of the AGV scheduling is also one of the decisive factors influencing the overall operation efficiency of the terminal. The implementation mainly researches digital twin-based automatic container terminal multi-AGV dynamic scheduling, and aims to improve scheduling efficiency and reduce scheduling deviation through closed-loop interaction of a physical space and a twin space of a terminal. Before the scheduled job starts. The twin space can verify the scheduling scheme generated by the code head service system, and the verified result can be transmitted to the service system so as to modify the scheduling scheme until the optimal scheduling scheme is generated. In the actual production operation process of the automatic container terminal, when a certain device in the physical space of the terminal breaks down or is disturbed, real-time data can be collected by the data service platform and transmitted to the twin space, the twin space analyzes a specific type of dynamic event through abnormal data analysis and feeds the dynamic event back to an AGV scheduling system, and the system calls a corresponding scheduling model and a genetic algorithm to adjust and update a previous scheduling scheme so as to respond to an uncertain event in the operation and guarantee the continuity, stability and efficiency of production to be maximized; meanwhile, real-time data can update the display in the twin space to realize virtual-real synchronization.

With reference to fig. 2, a multi-AGV dynamic dispatching digital twin frame for an automated container terminal is described, which mainly comprises a terminal physical space, a data service platform, an AGV dispatching digital twin system and a connection part. The formal expression is shown in formula 1. The dock physical space is the basis of the framework and is mainly responsible for providing basic information such as equipment current, electric quantity and the like, information of tasks being executed, equipment running state and data information related to scheduling. And the data information is processed by the data service platform and then transmitted to the AGV dispatching digital twin system. The AGV dispatching digital twin system mainly comprises a service system and a twin space. The twin space completes model updating by combining historical data on the basis of real-time data, and virtual-real synchronization is realized. And transmitting the new data generated by updating to the data service platform for storage. Meanwhile, the service system can generate a scheduling scheme by combining a corresponding model and an algorithm library on the basis of real-time data, and realize services such as scheduling process information statistics, equipment monitoring and the like. And the generated scheduling scheme is subjected to simulation verification in a twin space, and the operation of the wharf physical space is driven through a data service platform after the optimal scheduling scheme is obtained. The interaction between each part is carried out through EDI or a predefined interface. Through the interaction of the four parts, an AGV dynamic scheduling process with continuous iteration and optimization between the virtual part and the real part is formed. The above steps are repeated, so that an AGV dynamic scheduling process of virtual-real interaction iteration is realized.

DTF∷＝{TPS,DSP,VSS,CON} (1)

With reference to fig. 3, the AGV scheduling problem of a typical automatic container terminal is focused, the automation equipment related to the AGV scheduling problem mainly comprises a quay bridge, AGVs and a yard bridge, fig. 2 is a simplified wharf model in the research of the application, the purpose of the AGV scheduling problem is to assist in constructing the scheduling model, a line segment in the graph is a directed line segment, a peripheral line segment is clockwise, the directions of adjacent middle longitudinal line segments are opposite, and the AGVs can complete transportation tasks according to nodes in the graph.

With the aid of FIG. 2, the AGV scheduling problem is modeled.

Model parameters:

(1) parameter(s)

I denotes the import container set (1,2,3 … I ∈ I); e represents the set of export containers (1,2,3 … j ∈ E); n denotes the set of all containers, N ═ I ═ E; q represents a set of quayside container QCs, and m, l belongs to Q; v represents an AGV set, and a belongs to V; b represents a set of bridges; n is a radical of_qRepresenting nodes of intersection of the shore bridge sides and the road network; n is a radical of_bRepresenting nodes where AGV partners intersect the road network; n 'represents other nodes except for nodes intersected with a shore bridge and an AGV partner, and c, d belongs to N';N^*representing a set of all nodes, N^*＝N′∪N_q∪N_b；T₁Indicating the time it takes for the quayside crane to place or grasp a container to or from the AGV; t is₁Placing an AGV partner for the AGV a or acquiring the time used by the container from the AGV partner; s_a,imRepresents the time at which the AGV a starts to transport QCm containers i; f. of_a,imIndicating the completion time of the AGV a transporting the last container i; p is a radical of_a,imRepresenting the processing time during which the AGV a transports the container; h is_a,imRepresenting the running time from the AGV a to the AGV partner for transporting the QCm container i;

or

M represents a very large positive number;

(2) Decision variables

0-1 decision variables

indicating the order in which AGV a processes container i and container i ', a ∈ V, i, i' ∈ N, and m ∈ Q.

Non-0-1 decision variables

d_iaIndicating whether container i has changed the corresponding AGV a after rescheduling.

In order to improve wharf operation efficiency and guarantee stability in the rescheduling process, the method takes AGV completion time and stability as targets, and guarantees the stability of the rescheduling scheme by minimizing the deviation degree of the rescheduling scheme on the basis of meeting the requirement of minimizing the AGV completion time, so that the stability of the digital twin-based multi-AGV dynamic scheduling is reflected;

the formula (1) is an objective function, and aims to ensure the stability of a dispatching system by minimizing the deviation degree of a rescheduling scheme on the basis of meeting the requirement of minimizing the completion time of the AGV; the formula (2) represents that the completion time of transporting each container is less than or equal to the completion time of the last container; formulas (3) to (4) show that each AGV has an initial task of 0 and an end task of a virtual task f; equation (5) represents the sequence of operations of the AGV, ensuring that there is only one container task before and after the container task currently being processed by the AGV; equation (6) indicates that any container can only be handled by one AGV; formula (7) every AGV can transport a container at every turn, and two formulas guarantee the uniqueness between AGV and container. Equations (8) (9) ensure that the AGV begins the loading and unloading ship sequence. The formula (8) is that the same AGV completes a ship unloading task of the shore bridge and then completes a ship loading task of the shore bridge; and (9) after the same AGV finishes the ship loading task of a shore bridge, the same AGV finishes the ship unloading task of the shore bridge. Equation (10) the time relationship from the start of the unloading of the container by AGV a onto the container mate; equation (11) time constraint from one AGV mate to another AGV mate in an empty AGV state; equation (12) the time relationship between the loading of the container by the AGV a from the AGV mate to the shore bridge; equation (13) the time constraint from one shore bridge to another shore bridge in the empty state of AGV a; equation (14) is a non-negative constraint in the AGV scheduling problem.

With reference to fig. 4, in the implementation of the digital-twin-based multiple-AGV dynamic scheduling policy of the automatic container terminal in this embodiment, on the basis of the establishment of the digital twin frame and the scheduling model, in the face of the influence of uncertain factors in the AGV scheduling process, the digital twin frame is dynamically scheduled by the AGVs, and the data service platform fuses real-time data and historical data generated in the AGV operation to drive related services in the system, so that the multiple-AGV dynamic scheduling is implemented. The specific process is as follows:

before the dispatching is started, the AGV dispatching system generates a service through a dispatching scheme according to the operation requirements and the field resources to generate an initial dispatching scheme. The initial scheduling scheme is subjected to simulation verification in a twin space, and a verified result is fed back to the scheduling system. And the scheduling system calls a scheduling algorithm to modify the scheme according to the feedback result, generates a new scheme, and drives the field operation to execute after the verification is passed. In the process of executing the scheduling scheme, the AGV scheduling system calls a wharf production process state monitoring service to judge whether disturbance exists or not. If the disturbance does not exist, judging whether the scheduling task is finished; if not, the current scheduling scheme is continuously executed, otherwise, production information statistical service is called, and information such as completion time, failure rate and the like in the scheduling process is output. If the disturbance exists, updating the current scheduling scheme according to the disturbance condition; meanwhile, the twin space analyzes specific types of dynamic events through abnormal data analysis, feeds the dynamic events back to the AGV scheduling system, and the system calls a corresponding scheduling model and a genetic algorithm to generate a real-time scheduling scheme. And performing simulation verification on the new scheduling scheme in the twin space by utilizing the characteristic of the compressed space-time ratio, and comparing the verified scheduling scheme with the target value of the currently executed scheme so as to judge whether rescheduling needs to be triggered. If the rescheduling does not need to be triggered, the current scheme is continuously executed; and if necessary, rescheduling the remaining uncompleted tasks, and replacing the updated scheduling scheme which is currently executed by using the real-time scheduling scheme. And repeating the steps in a circulating way until the AGV completes all tasks.

The problem of AGV scheduling of an automatic container terminal is always an NP difficult problem, and for an emergency occurring in the production operation of the terminal, the current system is difficult to respond to the sudden disturbance in a short time. Therefore, the application provides a wharf scheduling element heuristic adaptive genetic algorithm based on the digital twin, which is used for responding to uncertain events in scheduling. The genetic algorithm is widely applied to the scheduling field due to the characteristics of global search capability, strong flexibility, simple and convenient operation and the like. Therefore, in conjunction with fig. 4 and fig. 5, the optimization of the real-time scheduling scheme of the present application is to be implemented by using an adaptive genetic algorithm with multi-layer coding. The improvement point of the self-adaptive genetic algorithm is that the genetic parameters are self-adaptively adjusted, so that the convergence of the algorithm is ensured while the diversity of the population is kept. The adaptive strategy is to perform adaptive adjustment in the evolution process: the rough searching process is favorable for keeping population diversity, and the later stage is adjusted to be a smaller value to carry out careful searching, so that the optimal solution is prevented from being broken, and the convergence speed is accelerated. The problem is coded by a real number coding mode according to the characteristics of researching the AGV scheduling problem. To better address the problem, chromosomal genes were encoded in multiple layers, as shown in FIG. 5. During transport of the AGV, the AGV unloads a container and then loads another container. After a load is completed, the AGV may continue to transport the next unloaded container. A multi-level chromosome of length 2 x N is designed, where N represents the number of container tasks. The first layer of the chromosome represents the container number, the first N/2 represents unloading the container, and the last N/2 represents loading the container; the second level of chromosomes represents the AGV number. The most common roulette method is adopted for selection operation in the algorithm, and because container tasks are not matched with AGV, only cross variation operation is needed to be carried out on chromosomes, and the cross operation adopts a multipoint cross mode. Cross probability (P)_c) And mutation probability (P)_m) Dynamic tuning in an adaptive mannerThe whole algorithm is expressed by equations (15) and (16) to improve the convergence rate and accuracy of the algorithm. The algorithm of the present application relates to the elite retention and stopping criterion. Since the crossover, mutation, etc. operations may destroy the gene structure of the optimal individual obtained in the early stage, the optimal individual obtained by population evolution up to now using elite retention is directly retained without the subsequent crossover, mutation, etc. operations. For the termination condition of the algorithm, the maximum evolution algebra allowed by the genetic algorithm is adopted as a stopping criterion.

Wherein f' is the fitness value of the preferred individual, f_avgIs a population mean fitness value, f_maxIs the maximum fitness value of the population, c₁、c₂、m₁、m₂Is constant and has a value ranging from 0 to 1.

This section demonstrates examples of the models developed in the present application and the algorithms proposed, all experiments run on the window10 operating system, Intel (R) core (TM) i7-8750U CPU @2.40GHz, 12GB memory computer Visual Studio (2017) and Unity (2019) versions. The average was taken as the final result for 10 experiments for each case.

Setting model parameters:

the number of containers varies from 20 to 200, 10 AGVs, 4 shore bridges, 4 yard bridges. And the processing time of each shore bridge and the travel time of the AGV in the operation are obtained by performing multiple times of simulation on the conventional actual operation.

Setting algorithm parameters:

the influence factor c1 of the adaptive crossing rate of the genetic algorithm is 0.8, c2 is 0.3, the influence factor m1 of the adaptive variation rate is 0.6, m2 is 0.2, the population size is 20, and the maximum iteration number is 200.

In this embodiment, a comparative experiment is performed by taking an actual loading and unloading task of an automatic dock as an example, and advantages and disadvantages of two scheduling schemes based on digital twin generation and direct generation (without using digital twin) of a current dock scheduling system are compared.

The experimental results are as follows:

(1) before starting scheduling, two initial scheduling schemes are compared under the same test environment. First, an initial scheduling plan is directly generated by the AGV scheduling service system, as shown in FIG. 6. The task allocation and the task start and end times are shown in table 1.

TABLE 1 AGV initial scheduling scheme task assignment without digital twinning

Secondly, under the same test condition, an initial scheduling scheme based on a digital twin is generated by utilizing the interaction of the AGV scheduling service system and the three-dimensional twin space, and the figure 7 can be seen. The task allocation and the task start and end times are shown in table 2.

TABLE 2 AGV initial scheduling scheme task assignment based on digital twinning

The present application compares AGV completion times of the initial scheduling schemes in the two modes before scheduling begins. As can be seen in connection with fig. 8, the time-out of the initial scheduling scheme based on the digital twin is much smaller than the time-out of the scheduling scheme directly generated by the system. The reason is that the initial scheduling scheme generated by the system can be subjected to simulation verification in the twin space, the verification result can be fed back to the system, and the system can adjust the scheme according to the feedback result until the optimal scheme meeting the constraint is found.

(2) During the execution of the scheduled experiment, the job process status monitoring service detects that AGV9 failed at 106s, which is a three minute maintenance time. And the AGV dispatching system based on the digital twin can prejudge the disturbed execution scheme and whether the disturbed execution scheme meets the current constraint. If not, the rescheduling is triggered, namely, the subsequent uncompleted tasks are redistributed, the generated new scheme is verified in the twin space, and the verified scheme replaces the current execution scheme. With reference to fig. 9, after the rescheduling is completed, a final scheduling scheme based on the digital twin is obtained. With reference to fig. 10 and 11, the final scheduling scheme AGV based on the digital twin has a completion time of 521s, and the algorithm converges in 164 generations with a better best fitness value. Table 3 shows the AGV scheduling schemes based on the two modes of digital twin and without digital twin in the whole experimental process comparing the maximum completion time and the scheduling stability.

TABLE 3 comparison of the two methods

In summary, the experimental results in this embodiment show that, when an uncertain time occurs in the scheduling process, the digital twin-based AGV scheduling method reschedules an incomplete container task, that is, a task on a failed AGV is allocated to the earliest completed AGV. Through the result analysis of the AGV dispatching digital twin system, the system can respond to uncertain events in the wharf dispatching operation process in time, and meanwhile, the stability of the dispatching scheme is greatly improved. Therefore, the existence of the digital twin enables the interaction between the twin space and the physical space to be possible, and the interaction between the twin space and the physical space triggers the rescheduling in time, so that the idle time of the AGV is reduced, the transportation efficiency of the AGV is improved, and the maximum completion time is further reduced. In addition, parameters in AGV dispatching are more and more accurate and dispatching stability is higher and higher through the closed loop iterative interaction of the service system and the twin space.

While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.

The blocks or flows in the figures are not necessarily required to practice the invention.

Reference to the literature

[1] The digital twin model construction theory and application [ J ] computer integrated manufacturing system 2021,27(01):1-15.

[2]QIAN,W W,GUO,Y,CUI,K et al.Multidimensional Data Modeling and Model Validation for Digital Twin Workshop.Journal of Computing and Information Science in Engineering,2021,21(3): 031005。

Claims

1. A digital twin-based automatic container terminal multi-AGV dynamic scheduling method is characterized by comprising the following steps:

s1: creating a twin space corresponding to physical entities in a wharf physical space, wherein the physical entities in the wharf physical space mainly comprise a shore bridge, a field bridge, an AGV, a ship, a storage yard, a container and a driving path, establishing multi-dimensional digital twin models of the physical entities by using a 3D MAX tool, and combining the multi-dimensional digital twin models according to a wharf real scene to form the wharf twin space;

s2: according to the AGV dispatching process, a dynamic digital twinning model in the operation process is constructed, wherein the dynamic digital twinning model comprises an AGV health state model, an AGV loading and unloading efficiency model and an AGV energy consumption model; the AGV health state modeling is to complete the mapping between data and states by acquiring data including motor current, voltage and speed in the AGV operation process and establishing equipment state health factors by utilizing a long-time memory network, as shown in a formula (1-5).

Y_i＝t_i*tanh(Z_i) (1)

f_i＝σ(W_c·[Y_i-1,X_i]+b_c) (3)

s_i＝σ(W_c·[Y_i-1,X_i]+b_c) (4)

Wherein E_VRepresenting total energy consumption of the AGV, N representing the number of container tasks, Q representing the number of shore bridges, B representing the number of yard bridges, θ_e、θ_d、θ_wThe average energy consumption per unit time t is the average energy consumption of the AGV in an idle running state, a load running state and a waiting state_ajj′The time taken for the AGV a to travel from the end of a container task j to the start of another container task j', T_ajThe time taken for the AGV a to travel from the start node to the end of task j' when transporting the container task j,

s3, performing model similarity calculation on the multidimensional digital twin model constructed in the step S1, wherein the model similarity calculation is specifically according to the following formula 8-13:

Q_i＝rank A_i (12)

in which ξ_ijRefers to the ith component xi of AGV or shore bridge model in the model base_iThe ranking of the jth aspect of (1), wherein a splice between components can constitute a complete single model; beta is a_ijRepresents a normalized value; the jth aspect includes geometry, metallic luster, degree of wear; omega_jRepresenting the weight of each model attribute; c_iAnd P_iRespectively representing the ith constituent element xi of the model_iDistance from negative and positive optimal solutions; σ is the majority criterion weight, and when σ is greater than 0.5, A_iThe representation model has higher similarity; when σ is less than 0.5, A_iDisplaying the negative attitude of most people; q_iIs to order the ith constituent element, θ^e1Represents a model similarity evaluation index, and has a value of 0 or 1;

the threshold is variable and can be set according to actual needsPlacing; when Q is_iGreater than a threshold value

When theta is greater than theta^e1Is 1, otherwise is 0;

performing layout similarity verification on the multi-dimensional digital twin model constructed in the step S1; the layout similarity needs to consider the position data of the berths, the buffer areas and the berths; thus will theta^e2As an index for verification of layout similarity, the relative distance between the berth, the buffer and the berth; QC (quasi-cyclic)_i(x_i,y_i)i∈[1,i^*]I ∈ Z denotes the coordinates of the ith berth, i^*The number of quay bridges; v_j(x_j,y_j)j∈[1,j^*]J ∈ Z denotes the jth buffer coordinate, j^*The number of buffer areas; b is_k(x_k,y_k)k∈[1,k^*]K ∈ Z denotes the coordinate of the kth beta-position, k^*Representing the number of shellfish; QV_ijRepresenting the distance between the ith berth and the jth buffer area in the wharf physical space; VB_ikRepresenting the distance between the ith berth and the kth berth in the wharf physical space, as shown in formulas 14 and 15; the distance between the models in the virtual space is obtained through direct measurement;

VQV_ijand VVB_ikRespectively representing the distances between the twin model berth and the buffer zone and between the berth and the berth in the virtual space of the wharf; QPE_ijRepresenting a position error between the berth and a wharf physical system in the wharf virtual space with respect to the buffer; BPE_ikRepresenting the position error between the berth and the physical system of the wharf in the virtual space of the wharf, the related formula is as (16-18), wherein

The adjustable error threshold can be set according to actual requirements;

QPE_ij＝|QV_ij-VQV_ij| (16)

BPE_ik＝|VB_ik-VVB_ik| (17)

thus theta^e＝θ^e1∩θ^e2As shown in formula (19); when theta is^eWhen the quay physical space and the twin space have the same model similarity and layout similarity, the quay physical space and the twin space are represented as 1;

then verifying whether the output of the wharf twin space constructed in the step S1 can sufficiently reflect the behavior characteristics of the wharf physical space; the high-order singular value decomposition method is adopted for verification, and the evaluation index theta is obtained^bJudging whether the feature number and the feature importance degree of the physical space and the virtual space of the wharf meet the requirements of virtual-real consistency or not; because the production data generated by the physical space is complex, the production data is stored by adopting tensor; assuming that a wharf is in the same production operation scene, a scheduling scheme is used as input, and device and container position data and the like generated in a physical space and a virtual space of the wharf in the operation process are stored in data tensors, wherein the two space data tensors are respectively omega^pAnd Ω^vIs subjected to modulo-n expansion to obtain

And

then singular value decomposition is carried out to respectively obtainLeft singular matrix

And

right singular matrix

And

as shown in formula (20);

wherein ∑ is ═ x₁,χ₂,…,χ_k,χ_k+1,…,χ_h) K is less than or equal to h is less than or equal to min (m, n), elements on the sigma diagonal are singular values, represent the importance of the corresponding features, and the values are arranged from large to small; since the values in Σ are decreasing and, to reduce the amount of computation, the matrix Σ can be approximately described with singular values of k terms,

judging the physical space characteristics of the wharf by solving the characteristic values of the physical space and the virtual space

And virtual space characteristics

Whether they are equal; sigma_allRepresenting the number of identical features in the first k features of the two spaces, as shown in equations 21-23;

wherein theta is^σIs an evaluation index for judging the same characteristic quantity of the consistency between the reality and the virtuality when theta is^σ1 means that both spaces have the same first k main features; theta^γTo determine another evaluation index of the importance of the feature,

for judging the degree of importance of each feature,

the threshold value can be changed and can be determined according to actual requirements, and the specific implementation is shown in formulas 24-27; alpha is alpha_allRepresenting the number of features with the same importance ratio in the first k features in two spaces; if the number of the features with the same importance ratio in the first k main features is k, the main features of the dock physical space and the twin space are the same;

thus, θ^b＝θ^σ∩θ^γWhen θ is expressed as in formula (28)^b1 means that the wharf physics and the twin space have the same behavior characteristics;

that is, the evaluation criterion of the consistency of the actual and actual is theta^e∩θ^bWhen theta is^e∩θ^bWhen the number is 1, the constructed wharf twin space meets the verification of consistency of virtual and real;

s4: on the basis of the step S3, constructing a multi-AGV dynamic dispatching digital twin frame of the automatic container terminal; the frame mainly comprises a wharf physical space, a data service platform, an AGV dispatching digital twin system and a connection part, and the formal expression of the frame is shown as a formula (29); combining the twin space and the wharf service system to form an AGV dispatching digital twin system; the twin space completes model updating by combining historical data on the basis of real-time data, and virtual-real synchronization is realized; the new data generated by updating can be transmitted to the data service platform for storage; meanwhile, the service system can generate a scheduling scheme by combining a corresponding model and an algorithm library on the basis of real-time data, and realize services such as scheduling process information statistics, equipment monitoring and the like; the generated scheduling scheme is subjected to simulation verification in a twin space, and after an optimal scheduling scheme is obtained, the AGV scheduling digital twin system drives the wharf physical space to operate through a data service platform;

wherein DTF is represented as a digital twinning frame,

the representation is defined as that TPS represents wharf physical space, DSP represents a data service platform, VSS represents AGV dispatching digital twin system, and CON represents connection between each part;

s5: and (3) generating an AGV dispatching scheme: the initial input data comprises container tasks, AGV, the number of shore bridges and field bridges, influence factors of the self-adaptive crossing rate of the genetic algorithm, the population scale of the influence factors of the self-adaptive variation rate and the maximum iteration times; firstly, a scheduling model taking AGV completion time and stability as optimization targets is added with a constraint that one AGV can only process one container task at a specific time point; then according to formula

representing the time taken for the AGV a to transport the container from a certain box area of the bridge b to the destination shore bridge q; according to the calculation result, the container task is allocated to T_iaThe smallest value AGV; finally, on the basis of the allocation result of the container tasks, on the premise of meeting all constraint conditions, solving a scheduling model by using a genetic algorithm so as to obtain the sequence of the AGV transporting the container tasks and generate a scheduling scheme;

establishing a scheduling model taking AGV completion time and stability as optimization targets, wherein parameters of the model are as follows: i represents an import container set, namely 1,2,3 … I belongs to I; e represents the set of export containers, namely 1,2,3 … j ∈ E; n represents a collection of all containersAnd then, N ═ ie ═ oe; q represents a set of quayside container QCs, and m, l belongs to Q; v represents an AGV set, and B represents a set of a bridge; n is a radical of_qRepresenting nodes of intersection of the shore bridge sides and the road network; n is a radical of_bRepresenting nodes where AGV partners intersect the road network; n 'represents other nodes except for nodes intersected with a shore bridge and an AGV partner, and c, d belongs to N'; n is a radical of^*Representing a set of all nodes, N^*＝N′∪N_q∪N_b；T₁Indicating the time it takes for the quayside crane to place or grasp a container to or from the AGV; t is₁Placing an AGV partner for the AGV a or acquiring the time used by the container from the AGV partner; s_a,imIndicating the time at which AGV a began to transport QCm container i; f. of_a,imIndicating the completion time of the AGV a transporting the last container i; p is a radical of_a,imRepresenting the processing time during which the AGV a transports the container; h is_a,imIndicating QCm travel time for AGV a to transport container i to the AGV mate;

or

(ii) a M represents a very large positive number;

0-1 decision variables:

representing that the AGV a processes QCm container i, a belongs to V, i belongs to N, and m belongs to Q; x_imjlThe AGV a processes QCm container i and then processes QCl container j, wherein a belongs to V, i, j belongs to N, and m belongs to Q;

indicating the order in which AGV a processes container i and container i ', a ∈ V, i, i' ∈ N, and m ∈ Q. Non-0-1 decision variables: d_iaIndicating whether container i has changed the corresponding AGV a after rescheduling;

the scheduling model is specifically represented as follows:

an objective function:

and (3) constraint:

the formula (30) is an objective function, and aims to ensure the stability of the dispatching system by minimizing the deviation degree of a rescheduling scheme on the basis of meeting the requirement of minimizing the completion time of the AGV; equation (31) indicates that the completion time for transporting each container is less than or equal to the completion time for the last container; equations (32) - (33) indicate that each AGV is 0 for the initial task and f for the final task; equation (34) represents the sequence of operation of the AGV, ensuring that there is only one container task before and after the container task currently being processed by the AGV; equation (35) indicates that any container can only be handled by one AGV; formula (36) each AGV can transport one container at a time, and the two formulas ensure the uniqueness between the AGV and the container; equations (37), (38) ensure that the AGV begins the loading and unloading ship sequence; the formula (37) is that the same AGV completes the ship loading task of a shore bridge after completing the ship unloading task of the shore bridge; in the formula (38), after the same AGV completes the ship loading task of a shore bridge, the ship unloading task of the shore bridge is completed; equation (39) the time relationship from the start of the unloading of the container by AGV a onto the container mate; equation (40) time constraints from one AGV mate to another AGV mate in an empty AGV state; equation (41) the time relationship between the loading of the container by the AGV a from the AGV mate to the shore bridge; equation (42) the time constraint from one shore bridge to another shore bridge for AGV a in the idle state; equation (43) is a non-negative constraint in the AGV scheduling problem;

s6: before the start of scheduling, the scheduling scheme generated by the dock service system is transmitted to the twin space for verification, and the twin space simulates the whole flow of the scheduling scheme, so that the completion time (f) of each AGV is obtained according to step S2 (f)_a,im-s_a,im) Loading and unloading efficiency p of each AGV and total AGV energy consumption E of scheduling scheme_V(ii) a Judging whether the twin space simulation result meets the production target, if so, finishing the evaluation of a scheduling scheme, wherein the scheduling scheme can guide the operation of a physical space; if not, returning to the step S5, regenerating a new scheduling scheme, and verifying the regenerated new scheduling scheme to the twin space again; performing iterative interaction until an optimal scheduling scheme is obtained to guide the operation of the wharf physical space;

s7: sending the verified scheduling scheme to a physical space, and starting scheduling operation; the data collected in the dispatching process comprises motor current, voltage, speed and position of a container in the AGV operation process, and the data service platform processes, stores and maps the real-time data to form real-time operation data with unified data structures and data types; meanwhile, the real-time data can drive the twin space and the dock physical space to keep synchronous, and the twin space can monitor the dock scheduling operation in real time under the drive of the data;

s8: in the scheduling process, judging whether disturbance exists in the operation process by using real-time data; if the disturbance does not exist, judging whether the scheduling task is finished; if not, continuing to execute the current scheduling scheme, otherwise, calling production information statistical service by the system, and outputting completion time and failure rate in the scheduling process;

s9: when abnormal disturbance occurs in the wharf physical space, the system can update the current scheduling scheme according to the disturbance condition; meanwhile, the disturbance data are transmitted to a twin space through the data service platform, the twin space processes abnormal data, specific types of dynamic events are analyzed, results are fed back to the AGV scheduling system, and the step S5 is returned to regenerate a new scheduling scheme; performing simulation verification on the new scheduling scheme in the twin space in step S6, and comparing the verified scheduling scheme with a target value of the currently executed scheme, thereby determining whether rescheduling needs to be triggered; if the rescheduling is not required to be triggered, the current scheme is continuously executed until the task is completed; if necessary, rescheduling the remaining uncompleted tasks, replacing the updated scheduling scheme currently being executed with the real-time scheduling scheme, and returning to the step S8 for continuous execution; the steps are repeated in a circulating way until all tasks are completed; and after the task is finished, a Gantt chart of a scheduling scheme, an allocation table of container tasks in the scheduling scheme, the completion time of each AGV, the completion time of the total scheduling scheme, a stable value of the scheduling scheme and an algorithm performance iteration chart are output.