Ship segmented logistics scheduling method and system based on cross-system business data fusion

Abstract

The embodiment of the invention provides a ship segmented logistics scheduling method and system based on cross-system business data fusion, and belongs to the technical field of ship segmented logistics scheduling. The method comprises the steps of obtaining storage yard state information; obtaining path selection information according to the storage yard state information; obtaining a target function of a stock dump state prediction sequence according to the path selection information; obtaining ship storage position prediction information according to a target function of the storage yard state prediction sequence; and according to the ship storage position prediction information, completing ship segmented logistics scheduling. According to the method, effective training of a large-scale storage yard layout optimization model can be realized, and storage yard layout optimization is realized while shipyard data are protected.

Description

Technical Field

[0001] The present application relates to the technical field of ship segment logistics scheduling, and in particular to a ship segment logistics scheduling method and system with cross-system business data fusion. Background Art

[0002] In the field of smart manufacturing, warehouses play a vital role in the supply chain process. Warehouses balance costs and service quality to customers by improving space utilization and optimizing space layout. Warehouses include four main functions: receiving, storage, order picking and shipping. Storage is the most common function of warehouses. In the storage process of items, problems such as regional item allocation, cross-regional item allocation and storage location allocation may be involved. Disadvantages of related technologies:

[0003] 1. A large amount of data is needed to train the model, and the currently available public data sets make modeling difficult.

[0004] 2. Existing deep reinforcement learning models can only be used for small and medium-sized storage yards and cannot solve the state space explosion problem caused by large-scale storage yards. Summary of the invention

[0005] The main purpose of the embodiments of the present application is to provide a method and system for ship segment logistics scheduling with cross-system business data integration.

[0006] The technical solution adopted by the present invention is:

[0007] On the one hand, an embodiment of the present invention provides a ship segmentation logistics scheduling method for cross-system business data fusion, the method comprising the following steps:

[0008] Get the yard status information;

[0009] Acquiring path selection information according to the yard status information;

[0010] According to the path selection information, an objective function of a yard state prediction sequence is obtained;

[0011] According to the objective function of the yard state prediction sequence, obtaining ship storage location prediction information;

[0012] According to the predicted information of the storage location of the ship, the logistics scheduling of the ship is completed in sections.

[0013] Furthermore, the obtaining of the storage yard status information comprises the following steps:

[0014] Obtain ship position information;

[0015] Obtain obstacle location information;

[0016] Get the schedule information of the ships in the yard;

[0017] The ship position information, the obstacle position information, and the schedule information of the ships in the yard are used as the yard status information.

[0018] Furthermore, obtaining the path selection information according to the storage yard status information comprises the following steps:

[0019] Get the new ship's starting position;

[0020] Get the current position of the new ship;

[0021] Get the candidate target position of new ship;

[0022] Calculating path loss information according to the starting position of the new ship, the current position of the new ship and the candidate target position of the new ship;

[0023] According to the path loss information, a storage position sequence is calculated using a WNGO algorithm;

[0024] Path selection information is obtained according to the storage location sequence and the storage yard status information.

[0025] Furthermore, the path loss information is calculated based on the starting position of the new ship, the current position of the new ship and the candidate target position of the new ship, and the formula used includes:

[0026]

[0027] in, is the total path loss in the yard state at time t, is the actual path loss from the starting position of the new ship to the current position of the new ship, wherein the actual path loss includes the loss of moving the obstacle block. is the path loss from the current position of the new ship to the candidate target position of the new ship.

[0028] Further, the step of calculating the storage position sequence using the WNGO algorithm according to the path loss information comprises the following steps:

[0029] The Whale Optimization Algorithm and the Neighborhood Gradient Optimization Algorithm are integrated to obtain the WNGO algorithm.

[0030] According to the path loss information and the WNGO algorithm, the state transfer hysteresis and predictability optimization information of the storage yard is obtained;

[0031] A storage location sequence is obtained based on the state transfer hysteresis and predictability optimization information of the storage yard.

[0032] Furthermore, the objective function of the yard state prediction sequence is obtained according to the path selection information, and the formula used includes:

[0033]

[0034] Among them, R t The reward for selecting the storage location of the yard state at time t is W = [ω1, ω2, …, ω N ] is the weight of the predicted state of the yard; MaxF(R t ,W) is the objective function of the yard status prediction sequence;

[0035] is the path loss of the new block entering the yard, is the average path loss of all rearranged blocks transported out of the yard in the next rearrangement task, is the average path loss for emptying the yard;

[0036] The number of times new blocks are moved into the yard, is the average number of times all rearranged blocks are moved in the yard, is the average number of moves to empty the yard.

[0037] Furthermore, obtaining the ship storage location prediction information according to the objective function of the yard status prediction sequence comprises the following steps:

[0038] According to the objective function of the yard state prediction sequence, state fitness evaluation information is obtained;

[0039] According to the state fitness evaluation information, repeated states are eliminated to obtain ship storage location prediction information.

[0040] On the other hand, an embodiment of the present invention further provides a ship segment logistics scheduling system with cross-system business data fusion, the system comprising:

[0041] The first module is used to obtain the yard status information;

[0042] The second module is used to obtain path selection information according to the yard status information;

[0043] The third module is used to obtain the objective function of the yard state prediction sequence according to the path selection information;

[0044] The fourth module is used to obtain the ship storage location prediction information according to the objective function of the yard status prediction sequence;

[0045] The fifth module is used to complete the ship segment logistics scheduling according to the ship storage location prediction information.

[0046] On the other hand, an embodiment of the present invention also provides a ship segmented logistics scheduling device with cross-system business data integration, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, it implements the method for ship segmented logistics scheduling with cross-system business data integration as described above.

[0047] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method described above.

[0048] The embodiments of the present application include at least the following beneficial effects: The present application provides a method and system for ship segmented logistics scheduling with cross-system business data fusion. The present invention obtains yard status information; obtains path selection information based on the yard status information; obtains the objective function of the yard status prediction sequence based on the path selection information; obtains ship storage location prediction information based on the objective function of the yard status prediction sequence; completes ship segmented logistics scheduling based on the ship storage location prediction information. The present invention can realize the effective training of large-scale yard layout optimization models, and realize yard layout optimization while protecting shipyard data. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of a ship segmented logistics scheduling method for cross-system business data fusion provided by an embodiment of the present invention;

[0050] Figure 2 It is a schematic diagram of the shipyard yard layout and the task re-arrangement process provided by an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of a decision process for a new ship block to enter a storage yard provided by an embodiment of the present invention;

[0052] Figure 4 is a schematic diagram of optimizing the layout of a large storage yard based on deep reinforcement learning provided by an embodiment of the present invention;

[0053] Figure 5 is a schematic diagram of a storage yard storage state prediction provided by an embodiment of the present invention;

[0054] Figure 6 It is a schematic diagram of an example of a process of storing and rearranging new ship blocks using SOA provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0056] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0057] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0059] Before describing the embodiments of the present application in detail, some nouns and terms involved in the embodiments of the present application are first described. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0060] 1) WOA (Whale Optimization Algorithm) is a heuristic optimization algorithm based on the hunting behavior of humpback whales in nature.

[0061] 2) NGO (Neighborhood Gradient Optimization) is an optimization algorithm that combines gradient information and neighborhood search.

[0062] 3) PSLAP (Partial-State Linear Approximation Problem) is a partial-state linear approximation problem, which involves how to effectively perform state approximation when dealing with large-scale state spaces to reduce computational complexity and storage requirements.

[0063] 4) PPO (Proximal Policy Optimization) is an optimization algorithm used to solve reinforcement learning problems.

[0064] 5) DRL (Deep Reinforcement Learning) is a technology that combines deep learning and reinforcement learning to solve complex decision-making and control problems.

[0065] 6) SOA (Stack Optimization Algorithm) is an algorithm used to optimize stack storage.

[0066] 7) WNGO is the similarity coded whale-hawk optimization algorithm of the present invention, which combines the characteristics of the whale optimization algorithm (WOA) and the eagle optimization algorithm (EOA) to solve the storage location optimization problem of the yard.

[0067] The present invention takes into account that the use of deep reinforcement learning to solve large-scale PSLAP must solve the modeling difficulties caused by state space explosion and lack of public data sets. The present invention achieves the modeling of large-scale PSLAP problems by standardizing storage rules. Then, according to the hysteresis of the yard state transfer, a heuristic algorithm is used to obtain the optimal Top-k storage coordinates in the current state of the yard, and then the corresponding yard state is generated according to the Top-k coordinates for pre-training of the deep reinforcement learning model.

[0068] The main purposes of the present invention are as follows:

[0069] ① The heuristic algorithm is used to simulate manual experience to select the storage location of the ship blocks in the yard to alleviate the problem of insufficient training data. The data set constructed by the heuristic algorithm has better results than that achieved by manual experience and is more conducive to model training;

[0070] ② Based on the above heuristic method of generating training data and combined with the hysteresis of the yard state transition, a deep reinforcement learning training method for large yards is proposed to solve the state space explosion problem in the model training process;

[0071] ③ Combined with the PPO algorithm, a deep reinforcement learning optimization model for yard layout optimization is designed.

[0072] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0073] On the one hand, the embodiment of the present invention provides 1. A ship segmented logistics scheduling method for cross-system business data fusion, referring to Figure 1 , the method comprises the following steps:

[0074] S100, obtaining yard status information;

[0075] S200, obtaining path selection information according to the yard status information;

[0076] S300, obtaining an objective function of a yard status prediction sequence according to the path selection information;

[0077] S400, obtaining ship storage location prediction information according to the objective function of the yard status prediction sequence;

[0078] S500: Complete the ship segment logistics scheduling according to the ship storage location prediction information.

[0079] The embodiment of the present invention discloses step S100 of acquiring yard status information, including the following steps:

[0080] S110, obtaining ship location information;

[0081] S120, obtaining obstacle location information;

[0082] S130, obtaining the timetable information of the ships in the yard;

[0083] S140: The ship position information, obstacle position information, and schedule information of the ships in the yard are used as the yard status information.

[0084] The embodiment of the present invention discloses that step S200 obtains path selection information according to the storage yard status information, including the following steps:

[0085] S210, obtaining a new ship starting position;

[0086] S220, obtaining the current position of the new ship;

[0087] S230, obtaining a new candidate target position of a ship;

[0088] S240, calculating the path loss information according to the starting position of the new ship, the current position of the new ship and the candidate target position of the new ship;

[0089] S250, using the WNGO algorithm to calculate a storage position sequence according to the path loss information;

[0090] S260: Obtain path selection information according to the storage location sequence and the yard status information.

[0091] As an optional embodiment, the present invention obtains the starting position of the new ship in S210, that is, the position of the ship before entering the yard. This position is usually the starting point of the ship outside the yard, which can be the initial position of the dock or the transport vehicle.

[0092] In step S220, the current position of the new ship in the storage yard is obtained. This position is the coordinates after the ship enters the storage yard and moves to the current point.

[0093] In step S230, candidate target locations of the new ship in the storage yard are obtained. These locations are where the ship can be stored, and are determined according to the layout of the storage yard and the status of other ships. Implementation methods include:

[0094] Generate candidate locations: Based on the yard layout and storage rules, all possible target storage locations are calculated. This usually requires taking into account the locations of existing vessels and free cells.

[0095] Evaluate candidate locations: Use the yard's status information (e.g., storage density, available space) to conduct a preliminary evaluation of candidate locations to ensure their suitability.

[0096] Select candidate locations: Provide a list of qualified candidate locations to subsequent algorithms for further optimization.

[0097] In step S240, the path loss from the current position of the new ship to each candidate target position is calculated. The path loss refers to the total cost or time from the current position to the target position, including the actual moving time and the additional loss caused by obstacles. Implementation methods include:

[0098] Define the path loss function: including the distance loss from the current location to the target location and the additional loss caused by obstacles.

[0099] Path search: Use the A* algorithm or other path search algorithms to calculate the loss values ​​of different paths.

[0100] Calculate loss information: According to the calculation result of the path loss function, obtain the loss information of each path.

[0101] In step S250, the WNGO algorithm (combining the Whale Optimization Algorithm and the Northern Goshawk Optimization Algorithm) is used to calculate the best storage location sequence. This sequence is the optimal storage location configuration, minimizing the overall path loss and other constraints. Implementation methods include:

[0102] Initialize the WNGO algorithm: set the algorithm parameters, including the weights of the whale optimization algorithm and the northern goshawk optimization algorithm.

[0103] Apply WNGO: Use the WNGO algorithm to optimize the path loss information and generate a storage location sequence.

[0104] Update model: Based on the results of the WNGO algorithm, update the storage location sequence and optimize the yard layout.

[0105] In step S260, the optimal path selection for the new ship is determined based on the optimized storage location sequence and the current status information of the yard. This information is used to guide the specific storage operation of the ship in the yard. The implementation method includes:

[0106] Apply storage location sequence: Apply the optimized storage location sequence to the current yard status and calculate the final path selection.

[0107] Generate path information: Generate the final path selection information, including the moving route and storage points, based on the storage location sequence.

[0108] Update yard status: Apply the routing information to the yard layout and update the yard status in real time.

[0109] The embodiment of the present invention discloses that step S240 calculates the path loss information according to the starting position of the new ship, the current position of the new ship and the candidate target position of the new ship, and the formula used includes:

[0110]

[0111] in, is the total path loss in the yard state at time t, is the actual path loss from the starting position of the new ship to the current position of the new ship. The actual path loss includes the loss of moving the obstacle block. is the path loss from the current position of the new ship to the candidate target position of the new ship.

[0112] The embodiment of the present invention discloses that step S250 uses the WNGO algorithm to calculate the storage location sequence according to the path loss information, including the following steps:

[0113] S251, the whale optimization algorithm and the neighborhood gradient optimization algorithm are integrated to obtain the WNGO algorithm;

[0114] S252, obtaining state transfer hysteresis and predictability optimization information of the storage yard according to the path loss information and the WNGO algorithm;

[0115] S253. Optimize information based on the state transfer hysteresis and predictability of the storage yard to obtain a storage location sequence.

[0116] The embodiment of the present invention discloses that step S300 obtains the objective function of the yard state prediction sequence according to the path selection information, and the formula used includes:

[0117]

[0118]

[0119] Among them, R t The reward for selecting the storage location of the yard state at time t is W = [ω1, ω2, …, ω N ] is the weight of the predicted state of the yard; MaxF(R t ,W) is the objective function of the yard status prediction sequence;

[0120] is the path loss of the new block entering the yard, is the average path loss of all rearranged blocks transported out of the yard in the next rearrangement task, is the average path loss for emptying the yard;

[0121] The number of times new blocks are moved into the yard, is the average number of times all rearranged blocks are moved in the yard, is the average number of moves to empty the yard.

[0122] The embodiment of the present invention discloses that step S400 obtains ship storage location prediction information according to the objective function of the yard status prediction sequence, including the following steps:

[0123] S410, obtaining state fitness evaluation information according to the objective function of the yard state prediction sequence;

[0124] S420: According to the state fitness evaluation information, duplicate states are eliminated to obtain ship storage location prediction information.

[0125] The present invention is directed to the following problems of the related art:

[0126] 1. Plane storage location allocation problem during ship segment processing

[0127] The stockyards of large shipyards can be divided into normal stockyards and common stockyards. Normal stockyards belong to specific workshops, while common stockyards store blocks from multiple workshops. If a ship block has completed the previous process and the next process cannot start immediately, the ship block will be stored in the normal stockyard near the subsequent workshop. If there is no remaining space in the normal stockyard, the ship block will be stored in the common stockyards. The normal stockyard is relatively small and can store a small number of ship blocks. Therefore, there is little temporary relocation work in the normal stockyard when the ship blocks are transported in and out of the stockyard. However, the common stockyards are large in scale and can store a large number of ship blocks, and there will be many obstacles when entering and leaving the stockyard. In this case, it is also important to reduce the use of relocation transporters, because the resource loss of transporters during temporary relocation cannot be ignored.

[0128] The task of moving blocks in common stockyards to different addresses is defined as rearrangement, which is usually performed to remove obstacles that hinder the movement of target blocks. If the transporter has to perform rearrangement work every time it transports blocks, it will not only affect the transportation efficiency of the yard, but also greatly increase the transportation cost. Therefore, in large shipyards, rearrangement work is usually carried out during the break time after normal working hours. The rearrangement task transports the blocks that need to be transported to the workshop for processing in the next working period to a specific location in the common stockyards close to the main road (see Figure 2 The green part in the common stockyards is the exit location of the common stockyards. In the daily operation of the yard, rearrangement is one of the tasks that must be minimized in the logistics of ship blocks. When the ship blocks arrive at the yard, fully considering the storage location of other ship blocks and the time of leaving the yard to optimize the yard layout is one of the effective ways to reduce the rearrangement tasks. Reference for the shipyard yard layout and rearrangement task process Figure 2 .

[0129] 2. Decision-making issues in the process of shipyard yard layout optimization

[0130] In the process of shipyard yard layout optimization, it is necessary to consider the storage location of new blocks when they enter the yard, the path of blocks entering and leaving the yard storage location, and the selection of the storage location of the obstructed blocks. Taking the process of new blocks entering the yard as an example, Figure 3As shown in the figure, when a new block arrives, a target location for storage needs to be selected. Then, according to the actual situation of the yard, the entrance and the path to the target location are selected, and the fitness of the target location is calculated along the selected path. The above steps are repeated until the optimal storage location is determined. After the optimal storage location and path are determined, if there is an obstacle block in the path, the obstacle block needs to be moved to another storage location.

[0131] As an optional embodiment, the technology of the present invention includes:

[0132] 1. Large-scale storage yard layout optimization model based on deep reinforcement learning

[0133] Modeling the optimization of the yard layout is a complex process, especially when new blocks are blocked from entering the target storage location in the yard. As the obstructing blocks are moved, the state of the yard also changes, so it is difficult for the model to obtain real rewards after executing actions. In order to overcome this problem, this work fully considers the stacking location and schedule of other blocks in the yard when designing the reward function, further weakening the impact of yard state changes on the selection of storage locations. According to the hysteresis and predictability of yard state transfer, this paper proposes a similarity coding whale-goshawk optimization algorithm (WNGO) for predicting the optimal storage location of the yard.

[0134] like Figure 4 As shown in the left figure "Stockyard state prediction and layout model pre-training", the top-k storage location sequences with the highest fitness are selected and converted into corresponding stockyard states, which are then used for pre-training of large-scale stockyard layout models. Since the reward for selecting a storage location is affected by the path into the stockyard, different paths will also have different rewards. This paper uses the A* algorithm for path search to obtain the optimal path. The selection of storage locations includes the selection of storage locations after new blocks arrive at common stockyards and the selection of locations where obstacle blocks re-enter common stockyards.

[0135] 2. Yard status prediction based on WNGO

[0136] (1) Yard status prediction

[0137] like Figure 5As shown in the figure, the system obtains the actual state of the yard at time t0 as the initial state, and then uses the WNGO algorithm to calculate the storage location with the highest fitness, and converts it into the predicted state of the yard at time t1. Similarly, the predicted state of the yard at time t2 is obtained by using the predicted state of the yard at time t1 as the initial state prediction, and so on to finally obtain a state sequence consisting of n predicted states. Since the prediction of the yard state has a high uncertainty, it is necessary to repeat the above operation to obtain more predicted state sequences, and then obtain the Top-k predicted state sequences with the highest fitness, and remove the same states in the sequence as the pre-trained state set of DRL.

[0138] In order to reduce the computational complexity, except for the change in the yard state caused by the storage of the ship block at the current moment, the changes in the yard state caused by other forms are not considered. That is, the fitness calculation of the storage location and the selection of the transportation path at time t1 only refer to the initial state of the yard at time t1, and the calculation at time t2 uses the yard state after the operation at time t1. The objective function of the yard state prediction sequence is defined as shown in formula (1).

[0139]

[0140]

[0141] Among them, R t The reward for selecting the storage location of the yard state at time t is W = [ω1, ω2, …, ω N ] is the weight of the predicted status of the yard. is the path loss of the new block entering the yard, is the average path loss of all rearranged blocks transported out of the yard in the next rearrangement task, is the average path loss of emptying the yard. Similarly, The number of times new blocks are moved into the yard, is the average number of times all rearranged blocks are moved in the yard, is the average number of moves to empty the yard.

[0142] (2) WNGO optimization algorithm

[0143] WOA randomly updates their positions based on each other's positions rather than the position of the current best whale. Although the global search capability in WOA is improved in the early stage, the local search capability is insufficient in the later stage. NGO is designed to simulate the hunting behavior of the northern goshawk, including prey identification and attack, pursuit and escape. Although it also randomly updates their positions based on the positions of each other's prey rather than the position of the current best goshawk, the efficiency of global search is not as good as the WOA algorithm. The development of the second stage enhances the local search capability of the NGO algorithm, allowing it to obtain a better combination than WOA within a feasible time. The present invention integrates the WOA and NGO algorithms to design the WNGO algorithm, uses the WOA algorithm to optimize the positions of some goshawk prey, guides the NGO algorithm to search in a better space, and avoids falling into the local optimum.

[0144] Phase 1: Prey Identification (Global Search)

[0145] In the first stage of hunting, the northern goshawk randomly selects a prey and then quickly attacks it. Due to the random selection of prey in the search space, this stage enhances the global exploration ability of NGO, with the goal of determining the optimal area, but the exploration efficiency is very low during random search. The WNGO algorithm divides the prey position X into two parts: one is the prey position generated by the WOA algorithm; the other is to avoid the algorithm from falling into the local optimal prey position. The guiding factor is to quickly determine the high-quality search area, and the random factor is to avoid the algorithm from falling into the local optimal area.

[0146] P i ~X i , i=1,2,…,N, j=1,2,…,Z, (3)

[0147]

[0148] Among them, P i is the prey position of the i-th northern goshawk, is its objective function value. i is the position of the northern goshawk, F i is its objective function value. is to generate a new recommended location at time t, represents the position of the prey at time t-1. In formula (5), t is the iteration counter and T is the maximum number of iterations. Formula (6) expresses the updating process of the prey position, represents the j-th position variable of the prey at time t, It is the objective function value of the position variable of this dimension at time t-1. for The j-th dimension position variable, is its objective function value. In formula (7) yes The j-th dimension position variable, Rd() is the rounding function, r is a random number in the interval [0,1], and I is a random number with a value of 1 or 2. The parameters r and I are random numbers used to generate randomness in the search and update behavior. In formula (8) It is a new service package in the first phase. is its objective function value.

[0149] Phase 2: Chase and Escape Operations (Local Search)

[0150] After the Northern Wilderness Goshawk attacks its prey, it will try to escape. Therefore, during the process of following and chasing, the Northern Wilderness Goshawk continues to chase its prey. Due to the Northern Wilderness Goshawk's high speed, they can chase their prey in almost any situation and eventually catch it. The simulation of this behavior increases the algorithm's ability to exploit local search in the search space.

[0151]

[0152]

[0153] For new solutions, is the objective function value, yes The j-th dimension position variable.

[0154] 3. Path selection optimization based on A* algorithm

[0155] The A* algorithm is an efficient heuristic search algorithm with good path optimization effect in two-dimensional grid maps. Therefore, the A* algorithm is used as the path search algorithm for storage locations in the yard. The definition of the A* algorithm is shown in formula (16).

[0156]

[0157] in, is the total path loss in the yard state at time t, is the actual path loss from the starting position to the current search position (including the loss of moving the obstacle block), is the path loss from the current search position to the end position.

[0158] The A* algorithm estimates the cost evaluation function value of the path node and uses it as the comprehensive priority of the node. When selecting the next node to be traversed, it selects the node with the highest comprehensive priority and gradually finds the optimal path. Therefore, the A* algorithm can quickly find the path with the smallest cost and the shortest distance. The specific process is as follows:

[0159] (1) Obtain the yard status at time t, and obtain the starting node s and the target node list D = [d1, d2, ..., d L ], initialize open and close lists;

[0160] (2)s is the starting node, d i As the target node, take s as a square to be checked and save it to the open list;

[0161] (3) Find the grid nodes that can be reached around the starting node s, put them into the open list, and set s as the parent node;

[0162] (4) Delete the starting node s from the open list and save it to the close list;

[0163] (5) Calculate the neighboring squares value;

[0164] (6) Select from the open list The square M1 with the smallest value is deleted from the open list and saved to the close list;

[0165] (7) Check all adjacent squares of M1:

[0166] ① The squares in the close list are no longer considered;

[0167] ② If the square is not in the open list, add it to the open list and calculate value, and set M1 as the parent node;

[0168] ③ If an adjacent square M2 is already in the open list, calculate a new path from s to square M2 and determine whether the parent node and Value: If the new node M2 If the value is lower, set M1 as the parent node and recalculate value, H value remains unchanged, if the new node M2 If the value is high, it means that the new path cost is higher. The value remains unchanged.

[0169] (8) Repeat (6) to (7) until the target node d appears in the open list. i , get the starting node s to the target node d i The optimal path of

[0170] (9) Repeat (2) to (8) to obtain the optimal path from the starting node s to all target nodes in the list D;

[0171] (10)Select The path with the smallest value is regarded as the global optimal path.

[0172] 4. Yard block storage optimization based on deep reinforcement learning algorithm

[0173] (1) Analysis of the optimization problem of ship block storage in the yard

[0174] In order to reduce the total resource consumption of new block storage and rearrangement tasks in large common stockyards, this study trained a stockyard storage optimization agent (SOA). The states, actions and rewards of SOA are defined as shown in Table 1. The reward of SOA is proportional to the number of relocations and the length of the transportation path in each stage. Assume that the stockyard is a rectangular grid with entrances on any side, with n storage units in the horizontal direction and m storage units in the vertical direction. Due to the problem of errors in the block schedule caused by uncertain factors, the present invention divides the storage unit status into 4 levels, such as Figure 6 As shown. Level 0 means that the storage unit is empty. Level 1 means that the block in the storage unit will be used in the next working period. Level 2 means that the block will be used in the near future. Level 3 means that the block will not be used in the near future or the schedule is unclear. Level conversion can only be strictly reduced level by level according to the constraints. Therefore, the yard state at time t is l = [0, 1, 2, 3]. Since the reward for selecting storage needs to consider the impact on the reordering task, and the blocks that need to be reordered in the next working period in the yard are known, SOA still uses formula (2) as the reward function.

[0175] Table 1

[0176]

[0177] SOA can be used both in the new block arrival phase and in the rearrangement phase. When a new block arrives at the periphery of the yard, SOA first confirms the storage location, and then uses the A* algorithm to calculate the entrance coordinates and the route with the least loss. The new block and the obstacle block are temporarily placed at the yard exit position, which cannot be on the route with the least loss. When the obstacle block is moved, the transport vehicle transports the new block to the designated storage location. Then observe the block schedule. If the block level at the exit is 1, it will be left at the exit, otherwise it will be treated as a new block and a storage location will be reselected in the yard. When there are multiple obstacle blocks that need to be relocated at the exit, they are relocated in the reverse order of the movement order. This rule is to prevent the block that enters the yard first from becoming a new obstacle block. The difference in the rearrangement phase is that the order of rearranging blocks in the yard needs to be determined first, and then the exit and route are determined in sequence using the A* algorithm. Finally, the transport vehicle is used to achieve the rearrangement of blocks and the placement of obstacle blocks.

[0178] (2) Reinforcement learning algorithm for solving the ship block layout problem

[0179] In the actual application environment, the transfer of the yard state is predictable, and there is a small probability that a large-scale state transfer will occur. Therefore, in order to maintain the stability of the model, the present invention uses the PPO algorithm to train SOA. PPO is a policy gradient algorithm for reinforcement learning, and its design goal is to limit the amplitude of each step of the policy update by introducing a regularization term, thereby ensuring the stability of the training process.

[0180] There are two mainstream methods for updating policy gradients in the PPO algorithm, namely KL penalty and Clip. Their main function is to limit the amplitude of policy gradient updates, thereby deriving different neural network parameter update methods. Compared with KL penalty, Clip has the advantages of easy-to-understand update rules, no need to set additional hyperparameters, and more explicit constraints on policy distribution updates. Therefore, the present invention uses Clip as a method for updating policy gradients. PPO uses the Clip strategy to constrain the amplitude of changes between the new and old strategies, and calculates the surrogate loss to optimize the strategy. At a given time step t, for action a t , the action probability ratio r under the new and old strategies t The calculation of (θ) is shown in formula (15). θ (a t ∣s t ) means in state s t Take action a t New strategy π θ The probability of Indicates that in state s t Take action a t The old strategy probability.

[0181]

[0182] The Clip loss function is used as part of the total loss function, and its definition is shown in formula (18). represents the expected estimate for time step t, is the estimated advantage value at time step t. clip() is a clamping function that takes the input r t (θ) is restricted to the range (1-∈, 1+∈), where ∈ is a hyperparameter.

[0183]

[0184] use is the total loss function. is the value function loss, c2E r (π θ) is the regularization term of the policy distribution entropy, and c1 and c2 are hyperparameters.

[0185]

[0186] The beneficial effects brought by the present invention are as follows:

[0187] (1) In the absence of actual yard layout data and corresponding public data sets, effective training of large-scale yard layout optimization models can still be achieved, thus achieving yard layout optimization while protecting shipyard data.

[0188] (2) Since the distribution of generated data is more uniform than that of actual human experience operations, our model will have higher robustness and accuracy when dealing with extreme situations and real-time storage decisions.

[0189] On the other hand, an embodiment of the present invention further provides a ship segmented logistics scheduling system with cross-system business data fusion, the system comprising:

[0190] The first module is used to obtain the yard status information;

[0191] The second module is used to obtain path selection information according to the yard status information;

[0192] The third module is used to obtain the objective function of the yard status prediction sequence according to the path selection information;

[0193] The fourth module is used to obtain the prediction information of the ship storage location according to the objective function of the yard status prediction sequence;

[0194] The fifth module is used to complete the ship segment logistics scheduling based on the ship storage location prediction information.

[0195] On the other hand, an embodiment of the present invention also provides a ship segmented logistics scheduling device with cross-system business data integration, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method for ship segmented logistics scheduling with cross-system business data integration as mentioned above.

[0196] The cross-system business data fusion ship segmented logistics scheduling device of the embodiment of the present invention includes a memory and a processor.

[0197] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0198] The memory may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage. Among them, ROM can store static data or instructions required by the processor or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at run time. In addition, the memory may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (e.g., DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0199] The memory stores executable codes, and when the executable codes are processed by the processor, the processor can execute part or all of the methods described above.

[0200] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the above method.

[0201] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0202] The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims

1. A ship segmented logistics scheduling method based on cross-system business data fusion, characterized in that: The method comprises the following steps: Get the yard status information; Acquiring path selection information according to the yard status information; According to the path selection information, an objective function of a yard state prediction sequence is obtained; According to the objective function of the yard state prediction sequence, obtaining ship storage location prediction information; According to the predicted information of the storage location of the ship, the logistics scheduling of the ship is completed in sections.

2. The method according to claim 1, characterized in that The method of obtaining the storage yard status information comprises the following steps: Obtain ship position information; Obtain obstacle location information; Get the schedule information of the ships in the yard; The ship position information, the obstacle position information, and the schedule information of the ships in the yard are used as the yard status information.

3. The method according to claim 1, characterized in that The step of obtaining the path selection information according to the storage yard status information comprises the following steps: Get the new ship's starting position; Get the current position of the new ship; Get the candidate target position of new ship; Calculating path loss information according to the starting position of the new ship, the current position of the new ship and the candidate target position of the new ship; According to the path loss information, a storage position sequence is calculated using a WNGO algorithm; Path selection information is obtained according to the storage location sequence and the storage yard status information.

4. The method according to claim 3, characterized in that The path loss information is calculated according to the starting position of the new ship, the current position of the new ship and the candidate target position of the new ship, and the formula used includes: in, is the total path loss in the yard state at time t, is the actual path loss from the starting position of the new ship to the current position of the new ship, wherein the actual path loss includes the loss of moving the obstacle block. is the path loss from the current position of the new ship to the candidate target position of the new ship.

5. The method according to claim 3, characterized in that: The method of calculating the storage position sequence using the WNGO algorithm according to the path loss information comprises the following steps: The Whale Optimization Algorithm and the Neighborhood Gradient Optimization Algorithm are integrated to obtain the WNGO algorithm. According to the path loss information and the WNGO algorithm, the state transfer hysteresis and predictability optimization information of the storage yard is obtained; A storage location sequence is obtained based on the state transfer hysteresis and predictability optimization information of the storage yard.

6. The method according to claim 1, characterized in that The objective function of the yard state prediction sequence is obtained according to the path selection information, and the formula used includes: Among them, R t The reward for selecting the storage location of the yard state at time t is W = [ω1, ω2, …, ω N ] is the weight of the predicted state of the yard; Max F(R t ,W) is the objective function of the yard status prediction sequence; is the path loss of the new block entering the yard, is the average path loss of all rearranged blocks transported out of the yard in the next rearrangement task, is the average path loss for emptying the yard; The number of times new blocks are moved into the yard, is the average number of times all rearranged blocks are moved in the yard, is the average number of moves to empty the yard.

7. The method according to claim 1, characterized in that The method of obtaining ship storage location prediction information according to the objective function of the yard status prediction sequence comprises the following steps: According to the objective function of the yard state prediction sequence, state fitness evaluation information is obtained; According to the state fitness evaluation information, repeated states are eliminated to obtain ship storage location prediction information.

8. A ship segmented logistics dispatching system with cross-system business data integration, characterized in that: The system comprises: The first module is used to obtain the yard status information; The second module is used to obtain path selection information according to the yard status information; The third module is used to obtain the objective function of the yard state prediction sequence according to the path selection information; The fourth module is used to obtain the ship storage location prediction information according to the objective function of the yard status prediction sequence; The fifth module is used to complete the ship segment logistics scheduling according to the ship storage location prediction information.

9. A device for ship segmented logistics scheduling with cross-system business data integration, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for ship segmented logistics scheduling with cross-system business data integration as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images