ASC collaborative scheduling method based on linear regression and mixed integer programming model

Abstract

The present invention discloses an ASC collaborative scheduling method based on linear regression and mixed integer programming models. The method comprises the following steps: the operating time of the ASC is estimated according to a multivariate linear regression algorithm, and the instructions to be scheduled are screened out in combination with a set time window. Then, a mixed integer programming model is adopted to allocate the screened instructions to the seaside / landside ASC based on the minimum penalty in combination with the operating range and empty load distance of the ASC. The instructions allocated to the seaside ASC are pre-sorted, and re-entry and re-exit tasks are inserted in the spare time of the loading instructions, or instructions of other operation types are inserted according to the priority of the instructions. For the landside ASC, the instructions are pre-sorted according to the card swiping time of the instructions, and queue-jumping processing is performed when the landside re-entry and re-exit instructions are successfully matched. The method realizes a rail crane production scheduling system which reduces the idle operation of the ASC, reduces the interaction delay, reduces the occupancy of the interaction area, avoids interference and conflict between ASCs, and improves the utilization rate of idle ASCs.

Description

Technical Field

[0001] The present invention belongs to the technical field of container terminals, and in particular relates to an ASC collaborative scheduling method based on linear regression and mixed integer programming models. Background Art

[0002] The current yard layout of automated terminals is mostly configured with a dual ASC (railway crane) yard mode: one yard is equipped with a sea-side ASC and a land-side ASC operating synchronously.

[0003] The yard is laid out perpendicular to the berth line, with its ends corresponding to the port's seaside (ship berthing side) and landside (onshore operations side), respectively. Transfer zones (TZs) are located at either end of the yard to store containers awaiting loading and unloading. ASCs are installed on rails and operate within a single area; landside and seaside ASCs are prohibited from operating across the yard.

[0004] Compared to single-ASC operations, dual-ASC dual-end operations improve space utilization and parallel operation speed. However, synchronization and interference between two ASCs operating in the same container area restrict overall ASC loading and unloading efficiency, making ASC scheduling more complex. Summary of the Invention

[0005] The present invention proposes an ASC collaborative scheduling method based on linear regression and mixed integer programming models. Through real-time scheduling of different types of operations based on priorities, coordination penalties, and other parameters, it realizes functions such as rail hoisting in and out, real-time adjustment of interactive positions according to horizontal transportation equipment, temporary placement of containers nearby to release equipment resources as quickly as possible, and real-time adjustment of positions according to changes in target positions, thereby improving yard operation efficiency.

[0006] The present invention is achieved by adopting the following technical solutions:

[0007] A collaborative scheduling method for ASCs based on linear regression and mixed integer programming is proposed, which includes:

[0008] 1) Obtain candidate job instructions from the job queue and use the multivariate linear regression method to estimate the estimated ASC job time, including:

[0009] Calculate the historical time consumed by each container during ASC operation based on historical data, denoted as M;

[0010] Get the historical location data corresponding to each container, denoted as X;

[0011] When the efficiency of the remote operator of the bridge is relatively stable, M and X have a linear relationship. A multiple linear regression model is used to calculate the parameters based on historical data samples and predict the estimated operation time based on the current container location information. The expression of the best fitting line is:

[0012]

[0013] Where y represents the estimated operation time of ASC, M is the historical time consumed by each container during ASC operation, X is the location of the corresponding container, and h is the difference between the starting position and the target position of the instruction;

[0014] 2) Filter out the instructions to be scheduled, including:

[0015] Select instructions starting from the current time + within the time window, and instructions with a time window earlier than the current time;

[0016] 3) Allocate the filtered dispatch instructions to the seaside ASC and landside ASC based on a mixed integer programming model, including:

[0017] Building a mixed integer programming model

[0018] , , , assign the instruction with the smallest instruction penalty to the seaside ASC and the landside ASC;

[0019] in, represents the penalty score of the i-th rail crane (i value is 1, 2, representing the landside ASC and seaside ASC, respectively) performing task j; Indicates whether to execute, with a value of 1 indicating execution and 0 indicating non-execution. The optimization goal is to minimize the total penalty score and find the optimal solution for the order of work instructions. The constraints are that each task can only be executed once and each rail crane can only match one task at a time. The penalty score is set based on the ASC's operating range and empty distance. The operating range is set based on the actual terminal operations, and the empty distance is calculated based on the instruction's starting and target positions.

[0020] 4) For sea-side ASCs, all instructions are pre-sorted based on the set operation priority, and sea-side re-entry and re-exit instructions are inserted into the idle time within the instruction's estimated operation time period. For land-side ASCs, all instructions are pre-sorted based on the instruction's card swipe time, and then a land-side re-entry and re-exit instruction is matched. If a match is successful, the instruction is queued and processed.

[0021] 5) Form the scheduling results and generate the job instruction queue that ASC can execute.

[0022] In some embodiments of the present invention, step 4 further includes:

[0023] Inserting pre-sorted instructions with the same priority level into the spare time within the estimated operation time period of the instructions; wherein sorting the instructions with the same priority level includes:

[0024] Obtain the function parameters of the pre-sorted instructions with the same priority and calculate their function penalties;

[0025] Put the one with the smallest penalty points first;

[0026] Among them, the functional parameters include the operating data of the rail crane, the time window of the instruction, the re-entry and re-exit on the sea side, the relay bay position range, the nearest container placement on the sea side / land side, and the exchange bay position range in the interactive area.

[0027] Compared with the prior art, the advantages and positive effects of the present invention are as follows: in the ASC collaborative scheduling method based on linear regression and mixed integer programming models provided by the present invention, the operating time of the ASC is estimated according to the multivariate linear regression algorithm, and the instructions to be called are screened out in combination with the set time window. Then, a mixed integer programming model is used to allocate the screened instructions to the seaside ASC and the landside ASC based on the minimum penalty in combination with the operating range and empty distance of the ASC. Then, the instructions allocated to the seaside ASC are pre-sorted according to the set instruction priority, and based on the estimated operation time of the loading instruction, it is determined whether there is free time within this time period for performing re-entry and re-exit tasks or inserting instructions of other operation types according to the instruction priority. For the landside ASC, the instructions are pre-sorted according to the instruction swiping time, and the landside re-entry and re-exit instructions are matched. When the match is successful, the queue is interrupted, and finally an optimized instruction operation sequence is formed; thus, a rail crane production scheduling system is realized that reduces ASC idle operation, reduces interaction delay, reduces interaction area occupancy, avoids interference and conflict between ASCs, and improves the utilization rate of idle ASCs.

[0028] Other features and advantages of the present invention will become more apparent after reading the detailed description of the embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The figure shows the execution steps of the ASC collaborative scheduling method based on linear regression and mixed integer programming model proposed in the present invention. DETAILED DESCRIPTION

[0030] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0031] Before describing the rail crane scheduling method and system proposed in the present invention, the following preliminary task-related parameter definitions and inventive concept preparations are made:

[0032] 1. The yard operation types are divided into 7 types, namely:

[0033] LOAD (loading), DSCH (unloading), RECV (receiving container), DLVR (lifting container), REHANDLE (turning over), SHFT (relay), YARD (moving container in yard).

[0034] 2. There are 10 priorities for setting different job instructions, namely:

[0035] On the rail crane (UNDERWAY = -1), need to return to the yard (LADEN_REDIRECT = 0), land side timeout task (LS_OVER_DUE = 1), loading task (WS_LOAD = 2), unloading task (WS_DSCH = 3), land side box receiving and sending (LS_RECV_DLVR = 4), moving boxes between yards (WS_MOVE_OUT = 5), moving boxes within the yard (WS_MOVE_IN = 6), yard overturning (REHANDLE = 7), yard relay (SHIFT = 8).

[0036] 3. Each job instruction needs to be prioritized according to the defined job instruction priority.

[0037] 4. Consider the time window (time range) for the work tasks that the rail crane needs to perform.

[0038] This time window is configurable on the front end. Under the EMT (estimated time) corresponding to each job task, the scheduling of the rail crane has a definable time window. If the current time falls exactly within the time window, the job task is configured as medium priority. If the current time is later than the time window, the job task is configured as high priority. If the current time is earlier than the time window, the job task is configured as low priority.

[0039] 5. The no-load distance and energy consumption of the rail crane need to be considered.

[0040] If only the priority of the work task is considered when dispatching tasks, the local optimal solution of the minimum empty-load distance of the rail crane will be ignored. If only the minimum empty-load distance is considered, it cannot accurately reflect the influence of priority on the scheduling of the rail crane, which will ultimately affect the operation of the rail crane and the improvement of operation efficiency.

[0041] 6. Define rail-mounted crane loading and unloading operations (i.e., container collection operations are performed sequentially after container collection operations), real-time adjustment of the interaction zone position based on horizontal transport equipment (i.e., real-time adjustment of the interaction zone position based on route optimization and operation changes of horizontal transport equipment), Revision operations that temporarily place containers nearby to quickly free up equipment resources (i.e., when yard operations are busy, the ASC places containers nearby and quickly performs the next operation), and Refine operations that adjust the position in real time based on changes in the target location (i.e., real-time adjustment of the operation position as the planned target location for the container changes).

[0042] The above operation design is designed according to the operation requirements and operation parameters, and is not limited by the present invention.

[0043] The scheduling method provided by the present invention needs to take into account the above-mentioned considerations to ensure that the impact between rail cranes is reduced as time changes and the number of site instructions changes, so that the scheduling of operation instructions can meet the optimal effect in the short term.

[0044] In combination with the above, the ASC collaborative scheduling method based on linear regression and mixed integer programming model proposed in the present invention is described in detail. Figure 1 Shown, including:

[0045] (1) Preliminary data processing.

[0046] Includes: 1. Loading job instructions;

[0047] 2. Initialize the yard data;

[0048] 3. Check the starting position of the instruction;

[0049] 4. Based on the command status, position change and other characteristics, combined with the above description 2, mark the command priority;

[0050] 5. Distinguish whether it is a sea-side rail crane operation or a land-side rail crane operation.

[0051] (2) Rail crane operation scheduling

[0052] 1. Obtain the alternative operation instructions in the operation queue and estimate the estimated operation time of the ASC (yard rail crane) based on the operation efficiency of the ASC.

[0053] The estimated operation time of ASC was calculated using a multiple linear regression method, including:

[0054] a) Calculate the historical time consumed by each container during ASC operation based on historical data, denoted as M;

[0055] b) Obtain the historical location data corresponding to each container, denoted as X;

[0056] c) When the efficiency of the remote operator of the bridge is relatively stable, M and X have a linear relationship. A multiple linear regression model is used to calculate the parameters based on historical data samples and predict the estimated operation time based on the current container location information. The expression of the best fitting line is:

[0057]

[0058] Where y represents the estimated operation time of ASC, M is the historical time consumed by each container during ASC operation, X is the location of the corresponding container, and h is the difference between the starting position and the target position of the instruction.

[0059] 2. Filter out the instructions to be scheduled:

[0060] Selects instructions starting from the current time + within the time window, and instructions with a time window earlier than the current time.

[0061] 3. Allocate the instructions to be dispatched filtered out in 2 to the seaside ASC and the landside ASC.

[0062] Establish a mixed integer programming model in the following form:

[0063]

[0064]

[0065]

[0066] in, represents the penalty score of the i-th rail crane (i value is 1, 2, representing the landside ASC and seaside ASC, respectively) performing task j; Indicates whether to execute. The value is 1 for execution and 0 for non-execution.

[0067] The optimization goal is to minimize the total penalty points and find the optimal solution for the job instruction sequence; the constraints are that each task can only be executed once and each rail crane can only match one task at a time.

[0068] The penalty points here are set based on the ASC's operating range and empty distance. The operating range is configurable at the system front end and can be set according to the actual operation of the terminal. The empty distance is calculated based on the starting position and target position of the instruction.

[0069] This step comprehensively considers the ASC's operating range, the command's starting and target positions, and the ASC's no-load distance for executing the command, and assigns the task with the smallest command penalty to the seaside ASC or the landside ASC.

[0070] 4. The seaside / landside ASC will pre-sort all assigned instructions.

[0071] During the ASC sorting on the sea side, the instructions for the cargo being hoisted on the rail are considered first, followed by the instructions for whether to return to the yard, and finally the loading instructions.

[0072] The seaside ASC also determines whether there is spare time within the time period for re-entry and re-exit tasks based on the estimated operation time of the loading instruction, or inserts operation instructions of other operation types according to the instruction priority (such as inter-yard container transfer, intra-yard container transfer, yard overturning, etc., which are detailed in steps 5 and 6).

[0073] The land-side ASC dispatcher pre-sorts the instructions according to the card swiping time of the instructions, and matches the instructions that enter and exit the land side repeatedly. If the instruction is matched successfully, the instruction will be queued.

[0074] 5. Obtain the function parameters of the pre-sorted instructions with the same priority and calculate their function penalties.

[0075] The following table shows an example of functional parameters and corresponding functional penalty points, including the speed and acceleration of the rail crane, the time window for operation, re-entry and re-exit on the land and sea sides, the relay bay range, the nearest container placement on the land and sea sides, and the exchange bay range in the interactive area. In each operation instruction, the attached functional parameters are different, and the calculated penalty points are also different.

[0076] The penalty points corresponding to the function parameters shown in the table can be set through the front-end. Table 1 is only an implementation example. In actual applications, the function parameters and penalty points can be adjusted, modified, increased or decreased.

[0077] Table 1

[0078]

[0079] 6. Based on the penalty scores obtained in step 5, the instruction with the smallest priority penalty is placed at the front of the instructions of the same type.

[0080] 7. Adjust the pre-sorting results to form the final instruction job sequence, form the scheduling result, and generate the job instruction queue that ASC can execute.

[0081] In the ASC collaborative scheduling method based on linear regression and mixed integer programming model provided by the present invention, the operating time of the ASC is estimated according to the multivariate linear regression algorithm, and the instructions to be scheduled are screened out in combination with the set time window. Then, the mixed integer programming model is used to allocate the screened instructions to the seaside ASC and the landside ASC based on the minimum penalty in combination with the operating range and empty distance of the ASC. Then, the instructions allocated to the seaside ASC are pre-sorted according to the set instruction priority, and according to the estimated operating time of the loading instruction, it is determined whether there is free time within this time period for performing re-entry and re-exit tasks or inserting instructions of other operation types according to the instruction priority. For the landside ASC, the instructions are pre-sorted according to the card swiping time of the instruction, and the landside re-entry and re-exit instructions are matched, and queue insertion is performed when the match is successful. Finally, the optimized instruction operation sequence is formed, realizing a rail crane production scheduling system that reduces ASC empty-load operation, reduces interaction delay, reduces interaction area occupancy, avoids interference and conflict between ASCs, and improves the utilization rate of idle ASCs.

[0082] It should be pointed out that the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. ASC collaborative scheduling method based on linear regression and mixed integer programming model, characterized by: include: 1) Obtain candidate job instructions from the job queue and use the multivariate linear regression method to estimate the estimated ASC job time, including: Calculate the historical time consumed by each container during ASC operation based on historical data, denoted as M; Get the historical location data corresponding to each container, denoted as X; When the efficiency of the remote operator of the bridge is relatively stable, M and X have a linear relationship. A multiple linear regression model is used to calculate the parameters based on historical data samples and predict the estimated operation time based on the current container location information. The expression of the best fitting line is: ； Where y represents the estimated operation time of ASC, M is the historical time consumed by each container during ASC operation, X is the location of the corresponding container, h is the difference between the starting location and the target location of the instruction; n is the number of containers; 2) Filter out the instructions to be scheduled, including: Select instructions starting from the current time + within the time window, and instructions with a time window earlier than the current time; 3) Allocate the filtered dispatch instructions to the seaside ASC and landside ASC based on a mixed integer programming model, including: Building a mixed integer programming model , , , assign the instruction with the smallest instruction penalty to the seaside ASC and the landside ASC; in, represents the penalty points of the i-th rail crane performing task j, where i values are 1 and 2, representing landside ASC and seaside ASC, respectively; Indicates whether to execute, with a value of 1 indicating execution and 0 indicating non-execution. The optimization goal is to minimize the total penalty score and find the optimal solution for the order of work instructions. The constraints are that each task can only be executed once and each rail crane can only match one task at a time. The penalty score is set based on the ASC's operating range and empty distance. The operating range is set based on the actual terminal operations, and the empty distance is calculated based on the instruction's starting and target positions. 4) For sea-side ASCs, all instructions are pre-sorted based on the set operation priority, and sea-side re-entry and re-exit instructions are inserted into the idle time within the instruction's estimated operation time period. For land-side ASCs, all instructions are pre-sorted based on the instruction's card swipe time, and then a land-side re-entry and re-exit instruction is matched. If a match is successful, the instruction is queued and processed. 5) Form the scheduling results and generate the job instruction queue that ASC can execute.

2. The ASC collaborative scheduling method based on linear regression and mixed integer programming model according to claim 1 is characterized in that: Step 4 also includes: Inserting pre-sorted instructions with the same priority level into the spare time within the estimated operation time period of the instructions; wherein sorting the instructions with the same priority level includes: Obtain the function parameters of the pre-sorted instructions with the same priority and calculate their function penalties; Put the one with the smallest penalty points first; Among them, the functional parameters include the operating data of the rail crane, the time window of the instruction, the re-entry and re-exit on the sea side, the relay bay position range, the nearest container placement on the sea side / land side, and the exchange bay position range in the interactive area.

Images