An inner-span type tire crane instruction scheduling method based on multi-dimensional dynamic priority and predictive scheduling

Abstract

The application discloses an inner-cross type tire crane instruction scheduling method based on multi-dimensional dynamic priority and predictive scheduling, which comprises the following steps: step S1, candidate instruction set generation and hard constraint filtering; step S2, opportunity overtaking window evaluation; step S3, nonlinear dynamic sorting of regular instructions; and step S4, instruction issuing and rescheduling. The application quantifies and decides whether to execute the overtaking instruction which can bring higher net benefit through the opportunity overtaking window evaluation, so as to improve the work efficiency. The application introduces the instruction waiting time factor and the maximum tolerance threshold with exponential growth characteristics, so as to ensure that the priority of the instruction waiting for a long time can be systematically improved, and the scheduling fairness can be ensured. In addition, the local congestion factor is used to reduce the priority of the instruction in the congestion area, so as to relieve the traffic bottleneck.

Description

Technical Field

[0001] This invention relates to the field of automated port technology, and in particular to a method for scheduling instructions for inboard rubber-tired gantry cranes based on multidimensional dynamic priority and predictive scheduling. Background Technology

[0002] In the complex operation of automated container terminals, the scheduling of straddle-type rubber-tired gantry cranes is a core element determining overall throughput efficiency. However, the operating area of ​​these gantry cranes has only one lane, and it is located inside the yard crane, making overtaking conditions quite challenging. Current scheduling methods are often rudimentary, such as strictly adhering to a "first-come, first-served" mechanized principle or making decisions based solely on a single optimization objective like minimizing travel distance. These traditional methods reveal their inherent limitations when dealing with rapidly changing real-world operational scenarios.

[0003] Rigid scheduling logic often leads to low operational efficiency. A system might order a rubber-tired gantry crane to travel a long distance unloaded to serve a truck that is first in the queue but physically far away. This not only significantly increases the total travel and energy consumption of the equipment but also forces other nearby vehicles to wait for extended periods, directly reducing the overall throughput of the terminal. On the other hand, simply pursuing a locally optimal strategy (such as always serving the nearest vehicle) sacrifices scheduling fairness, potentially causing indefinite delays of work orders in remote or less busy areas, creating a "work order starvation" phenomenon that severely impacts service quality. Traditional systems typically lack the ability to dynamically perceive and predict local traffic congestion within the yard, failing to proactively avoid or alleviate it, which can easily exacerbate operational bottlenecks.

[0004] Therefore, how to design an integrated scheduling method that can intelligently balance efficiency, fairness and congestion control, and dynamically adapt to changes in the field, is a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To overcome the shortcomings of the existing technology, the present invention provides a method for scheduling instructions for straddle-type tire cranes based on multi-dimensional dynamic priority and predictive scheduling, characterized by the following steps:

[0006] Step S1: Generation of candidate instruction set and hard constraint filtering: Collect all waiting operation instructions and filter them according to the preset "order of instructions in the same vehicle" hard constraint rule to generate a set of currently executable candidate instructions.

[0007] Step S2: Opportunity overtaking window evaluation: Evaluate whether there is an overtaking opportunity that meets the preset conditions in the candidate instruction set. If it does, calculate the net benefit of the overtaking operation through the quantitative analysis module. When the net benefit is greater than the preset opportunity threshold, determine the instruction with the closer target position in the queue as the optimal instruction and proceed directly to step S4; otherwise, proceed to step S3.

[0008] Step S3: Nonlinear dynamic sorting of regular instructions: If the optimal instruction is not determined in step S2, then for all instructions in the candidate instruction set, calculate the comprehensive priority score for each instruction through a comprehensive priority function, and select the instruction with the highest score as the optimal instruction.

[0009] Step S4: Instruction Issuance and Rescheduling: The optimal instruction determined in Step S2 or Step S3 is issued to the tire crane for execution; and after the instruction is completed, the system status is updated, and the process returns to Step S1 to start a new round of scheduling decision-making.

[0010] Furthermore, the process of calculating the net benefit of the overtaking operation in step S2 includes: calculating the generalized total cost under the sequential execution strategy and the overtaking execution strategy respectively, and taking the difference between the two; wherein, the generalized total cost is a quantitative cost value that combines the total travel distance required for the tire crane to complete the work sequence and the total waiting time incurred by the vehicle as a result.

[0011] Furthermore, the comprehensive priority function in step S3 has evaluation dimensions including at least: the travel distance of the instruction, the waiting time of the instruction, and the local congestion situation of the instruction target area.

[0012] Furthermore, the comprehensive priority score is calculated as follows: based on the evaluation dimensions, corresponding walking distance factors, waiting time factors, and local congestion factors are generated for each candidate instruction, and then the scores of each factor are weighted and summed.

[0013] Furthermore, the calculation of the travel distance factor is inversely proportional to the distance from the current position of the tire crane to the target position of the command.

[0014] Furthermore, the calculation of the instruction waiting time factor is related to the cumulative waiting time of the instruction, and when the cumulative waiting time of the instruction exceeds its average waiting time, its factor score increases exponentially.

[0015] Furthermore, when the cumulative waiting time of any instruction reaches the preset "maximum tolerance threshold", its priority is directly set to the highest.

[0016] Furthermore, the calculation of the local congestion factor is inversely proportional to the number of vehicles already in the queue under the instruction target position.

[0017] Beneficial effects:

[0018] This invention, through opportunistic overtaking window evaluation, can identify specific scenarios where "serving nearby vehicles first" is more efficient than "serving distant vehicles sequentially." It quantifies and compares the total costs of the two strategies and executes an overtaking operation when the net benefit exceeds a preset threshold. This directly reduces the idle travel of the tire crane and the waiting time of vehicles.

[0019] In the nonlinear dynamic sorting of conventional instructions, this invention introduces an instruction wait time factor that causes instructions with excessively long wait times to have their priority scores increase exponentially. Simultaneously, a set "maximum tolerance threshold" forces any instruction whose wait time reaches the upper limit to be placed with the highest priority, thus systematically avoiding the problem of some instructions being left unscheduled for extended periods due to location or other reasons.

[0020] This invention incorporates a local congestion factor into the decision-making model, taking into account the number of vehicles already queued in the target area. Instructions with more queued vehicles in the target area receive a lower priority score, thus guiding the tire-mounted crane to prioritize serving areas with better traffic flow, proactively preventing and alleviating traffic bottlenecks.

[0021] In summary, this invention integrates multiple dimensions such as walking distance, waiting time, and local congestion into a comprehensive priority score, achieving a quantitative trade-off between multiple objectives. Furthermore, after each instruction is completed, the system immediately restarts the complete closed-loop rescheduling process based on the updated on-site status, ensuring that decisions are always based on the latest real-time data and possessing a high degree of dynamic adaptability. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below with reference to a specific embodiment.

[0024] Example 1:

[0025] This invention provides a method for scheduling in-span rubber-tired gantry cranes based on multi-dimensional dynamic prioritization and predictive scheduling, aiming to optimize the operational efficiency of in-span rubber-tired gantry cranes in automated terminals. This method is particularly suitable for "6+1" configurations, i.e., operating areas with six container stacking areas and one one-way lane, where physical overtaking is typically limited, but there is still room for efficiency optimization under specific conditions.

[0026] The execution entity of this method is a scheduling and control system deployed on the central server of the terminal. This system communicates in real time with the PLCs of the rubber-tired gantry cranes and the vehicle identification system at the yard entrance. The system collects the data required for the scheduling method and executes the issuance of scheduling instructions. The specific steps are as follows:

[0027] Step S1: Generation of candidate instruction sets and hard constraint filtering

[0028] At the start of a decision cycle, such as at the end of each work cycle or when a new vehicle enters the yard, the scheduling system first gathers all waiting vehicle operation instructions to form an initial instruction pool.

[0029] The system then rigorously filters the initial instruction pool according to preset hard constraints to generate a logically executable "candidate instruction set." This step ensures that all subsequent decisions are based on safety and compliance. The hard constraint that must be followed is the "sequentiality of instructions within the same vehicle," that is:

[0030] For multiple operational instructions for the same container truck, the execution order of the bay numbers must be strictly followed from smallest to largest. Any instruction that is not qualified to be executed if its preceding smaller bay instruction has not been completed will be temporarily removed from the candidate instruction set.

[0031] Step S2: Assessment of Opportunity Overtaking Window

[0032] After generating the candidate instruction set, the system will enter a special evaluation phase designed to identify and utilize "opportunistic overtaking" windows that can significantly improve operational efficiency. This evaluation is initiated when the following scenarios are met:

[0033] (1) There is a vehicle in the queue that is ranked high by arrival time, but whose target operation position is very far from the current position of the tire crane. This vehicle is referred to as the "far vehicle".

[0034] (2) There is a vehicle in the queue that is later in the order of arrival time, but whose target operation position is very close to the current position of the tire crane. This vehicle is referred to as the "near vehicle".

[0035] (3) The system confirms through site status data that there are no other vehicles waiting or working on the physical path from the current position of the tire crane through the "near vehicle" position to the "far vehicle" position, ensuring that overtaking operations will not cause conflicts or new congestion.

[0036] Once the above conditions are triggered, the system will activate the quantitative analysis module. This module makes a decision by calculating the generalized total cost of the two strategies, "execute overtaking" and "execute in sequence," and comparing their "net benefits."

[0037] Specifically, the generalized total cost C is defined as:

[0038]

[0039] Wherein, Cost_distance represents the total travel distance required for the tire crane to complete the relevant instruction sequence; Cost_time represents the total waiting time of the relevant vehicles under this strategy; W_d and W_t are the weighting coefficients of distance and time costs, respectively, which can be adjusted by the operator according to the operation objectives of different time periods.

[0040] Net benefit NB is defined as:

[0041]

[0042] If the NB value is greater than a pre-set "opportunity threshold," it indicates that the overtaking maneuver will bring significant benefit improvement. The system will directly select the "approach vehicle" command as the optimal command and issue it, ending this round of decision-making. Conversely, if there is no overtaking opportunity or the net benefit is insufficient, the decision-making process will proceed to the next step without interruption.

[0043] Step S3: Non-linear dynamic sorting of regular instructions

[0044] If step S2 does not directly produce a decision, the system will perform a comprehensive non-linear dynamic ranking of all instructions in the candidate instruction set to determine the current optimal routine operation instruction. This step aims to balance multiple objectives such as efficiency, fairness, and site congestion.

[0045] A comprehensive priority function is constructed, and the system calculates a comprehensive priority score P_i for each candidate instruction i. This score is generated by a multi-dimensional nonlinear function:

[0046]

[0047] D, T, and Q represent the three core decision factors: walking distance, instruction waiting time, and local congestion, respectively.

[0048] Specifically, regarding the core decision factors mentioned above:

[0049] The walking distance factor D measures the immediate efficiency of the operation and can be expressed as:

[0050]

[0051] Where Bay_crane is the current bay position of the tire crane, and Bay_target_i is the target bay position of instruction i. This formula ensures that the shorter the travel distance, the higher the factor score.

[0052] The instruction waiting time factor T is used to ensure fairness in operations and avoid long vehicle waiting times, and can be expressed as:

[0053]

[0054] Where t_wait_i is the cumulative waiting time of instruction i since its generation (this time is recorded based on the time the vehicle enters the yard), t_avg is the average waiting time of all pending instructions in the current yard, and k (k>1) is the time sensitivity coefficient. This formula makes the priority of instructions with excessively long waiting times increase exponentially. When the waiting time t_wait_i of any instruction reaches the preset "maximum tolerance threshold", its priority will be directly set to the top.

[0055] The local congestion factor Q is used to prevent and alleviate traffic bottlenecks in specific areas, and its form is as follows:

[0056]

[0057] Where N_queue_i is the number of vehicles already queued at target bay i. This formula ensures that the more vacant the target bay, the higher the factor score. When the queue size N_queue_i exceeds the "congestion warning line," the system can apply additional penalty weights to this factor.

[0058] The system calculates the final priority score of each candidate instruction using a weighted summation method:

[0059]

[0060] The instruction with the highest score is selected as the optimal result for this round of decision-making.

[0061] Step S4: Instruction Issuance and Closed-Loop Rescheduling

[0062] The system issues the optimal command determined in step S2 or S3 to the tire crane's control system for execution. The instant the command is completed, the system immediately updates the global state, including the tire crane's new position, removing completed commands from the queue, and synchronizing any newly arriving vehicle commands. Subsequently, the system immediately returns to step S1, using the latest field status as input, to initiate a completely new and comprehensive scheduling decision-making process, thereby achieving uninterrupted, adaptive closed-loop control in response to dynamic changes in the field.

[0063] Example 2:

[0064] This embodiment aims to illustrate the complete decision-making process of the dynamic intelligent scheduling method described in Embodiment 1 using a specific dock operation scenario. This embodiment will use the quantization formula defined in Embodiment 1 and set specific parameters for the weighting coefficients and thresholds for calculation. The specific scenario parameter settings in this embodiment are as follows:

[0065] Weighting coefficients: walking distance weight W_d=0.5; waiting time weight W_t=0.4; congestion situation weight W_q=0.1.

[0066] Waiting time factor parameter: time sensitivity coefficient k=2.

[0067] Thresholds: Maximum tolerable waiting time = 600 seconds; congestion warning line = 2 vehicles.

[0068] Assume the scenario is set up as follows at initial time N:

[0069] Initial position of the tire crane: Bay_crane = 31 Bays.

[0070] Initial settings for the pending instruction queue:

[0071] Instruction A (Car 1): Target 0.7 B; Instruction B (Car 1): Target 15 B; Instruction C (Car 2): Target 0.9 B; Instruction D (Car 3): Target 27 B; Instruction E (Car 4): Target 31 B; Instruction F (Car 5): Target 33 B; Instruction G (Car 6): Target 47 B.

[0072] The initial waiting time (t_wait) for each instruction is assumed to be: Instruction A: 300 seconds; Instruction C: 240 seconds; Instruction D: 180 seconds; Instruction E: 120 seconds; Instruction F: 90 seconds; Instruction G: 60 seconds.

[0073] Initial setting of the number of vehicles queuing at each bay (N_queue): It is assumed that there are no other vehicles queuing at any target bay at this moment, that is, N_queue is 0.

[0074] Based on the above scenario, this embodiment provides an instruction scheduling method based on Embodiment 1, with the following specific steps:

[0075] Step S1: Generation of candidate instruction sets and hard constraint filtering

[0076] The system filters all pending instructions. According to the "sequential order of instructions within the same vehicle" rule, instruction B (vehicle 1, 15 B) does not meet the execution conditions because its preceding task instruction A (vehicle 1, 07 B) has not yet been completed, and is therefore temporarily filtered. Thus, after this filtering step, the candidate instruction set is: {Instruction A (07 B), Instruction C (09 B), Instruction D (27 B), Instruction E (31 B), Instruction F (33 B), Instruction G (47 B)}.

[0077] Step S2: Assessment of Opportunity Overtaking Window

[0078] The system evaluates whether there are instruction pairs that satisfy the overtaking model.

[0079] (1) Scenario analysis: The tire crane is located at 31B. The close-range commands are E (31B), F (33B), and D (27B); the long-range commands are A (07B), C (09B), and G (47B).

[0080] (2) Condition judgment: Although there are near and far instructions, the instructions are relatively densely distributed and do not meet the key site openness condition of "no other vehicles waiting to work on the path".

[0081] (3) Decision-making: If no overtaking opportunity that meets the preset triggering conditions is found, the system does not make a decision in this step, and the decision-making process automatically proceeds to the next step.

[0082] Step S3: Non-linear dynamic sorting of regular instructions

[0083] The system calculates the comprehensive priority score P for each instruction in the candidate instruction set according to the formula defined in Example 1.

[0084] (1) High-priority condition judgment:

[0085] Waiting time: The system checks the waiting time of each instruction, and none of the instructions (maximum 300 seconds) have reached the "maximum tolerance threshold" (600 seconds).

[0086] Congestion situation: The number of vehicles queuing for all target bays (all 0) has not reached the "congestion warning line" (2 vehicles).

[0087] No high-priority events were triggered.

[0088] (2) Calculation of scores for each factor:

[0089] The average waiting time t_avg = (300 + 240 + 180 + 120 + 90 + 60) / 6 = 165 seconds.

[0090] The calculation results of each instruction factor are shown in the table below:

[0091] The system calculates the comprehensive priority score P for each instruction in the candidate instruction set according to the formula defined in Example 1.

[0092] Table 1 Calculation Results of Each Instruction Factor

[0093] instruction Target Bay Walking distance D t_wait T Q A 7 24 0.04 300s 7.94 1 C 9 22 0.043 240s 6.03 1 D 27 4 0.2 180s 4.38 1 E 31 0 1 120s 3.02 1 F 33 2 0.333 90s 2.42 1 G 47 16 0.059 60s 1.86 1

[0094] The overall score for each instruction can be calculated based on the table above.

[0095] Step S4: Instruction Issuance and Status Update

[0096] Comparing the overall priority scores of all instructions, instruction A (07B) has the highest score, P_A = 3.32. Therefore, the system selects instruction A as the optimal instruction and issues it. This result shows that at time N, although instruction A travels a very long distance, its longest waiting time, under the established weighting system (with a higher time weight W_t), causes its accumulated priority due to the long waiting time to surpass the distance advantage of other instructions.

[0097] After receiving the instruction, the tire crane will move to bay 07 to complete the operation on vehicle 1. After the operation is completed, the system will update the tire crane's position to bay 07, remove instruction A, and add instruction B (vehicle 1, bay 15) to the candidate instruction set for the next round, and then return to step S1 to start a new round of decision-making.

[0098] This embodiment clearly demonstrates how the scheduling method proposed in this invention weighs multiple objectives based on a quantified priority model and makes scheduling decisions continuously and automatically based on dynamically changing field data.

[0099] Example 3:

[0100] This embodiment aims to specifically illustrate the application of step S2: opportunistic overtaking window evaluation in the scheduling method described in Embodiment 1. This embodiment will construct a special scenario that meets the overtaking conditions and demonstrate how the system makes an overtaking judgment through a quantitative decision model.

[0101] This embodiment uses the same preset parameters as Embodiment 2, where the "opportunity threshold" is set to 50, meaning that the system will only execute overtaking when the quantified net benefit exceeds this value.

[0102] Suppose that at a certain time T, the yard presents a special operational situation: the rubber-tired gantry crane is currently docked at 59B, and among all the pending instructions, there are two critical vehicles. One is vehicle number 8 (hereinafter referred to as the "distant vehicle"), which is the first vehicle in the order of entering the yard, and its target is 05B, which is far away; the other is vehicle number 9 (hereinafter referred to as the "near vehicle"), which is the second vehicle in the order, and its target is 55B, which is very close.

[0103] The key point of this scenario is that, through real-time site monitoring data, the system confirms that, starting from the current position of the tire crane (59 B), passing through the "near vehicle" target position (55 B), and then to the "far vehicle" target position (05 B), there are no other vehicles waiting or operating along the entire route. The site condition fully meets the ideal condition of "relatively empty site" required by this method to allow overtaking operations. Under this special scenario, the scheduling method is initiated. After completing the compliance filtering of the instructions in step S1, the system proceeds to step S2 to evaluate the opportunistic overtaking window.

[0104] The system identified that the current scenario fully triggered the "overtaking scenario model": there were "far-flying vehicles" (instruction X) that were ranked higher but had a distant target, and "near-flying vehicles" (instruction Y) that were ranked lower but had a closer target, and the physical path was open. Therefore, the system did not proceed to step S3 as in the conventional process, but instead directly activated the net benefit quantification decision module.

[0105] The core task of this module is to perform a precise cost-benefit analysis of overtaking behavior. It first identifies the "far vehicle" instruction (0.5 B) and the "near vehicle" instruction (5.5 B) to be decided. Then, following the definitions and formulas disclosed in Example 1, the module calculates the "generalized total cost (C)" for both strategies in parallel. For the "sequential execution" strategy (serving the far vehicle first, then the near vehicle), the generalized total cost C_sequential is calculated to be 95.2; for the "overtaking execution" strategy (serving the near vehicle first, then the far vehicle), the generalized total cost C_overtake is only 30.2.

[0106] After obtaining the generalized total cost of the two strategies, the module calculates the net benefit (NB) of the overtake by measuring the difference between them, i.e., NB = C_sequential - C_overtake = 65. This net benefit value directly quantifies the efficiency gain that can be achieved by performing the overtake. Finally, the module compares this net benefit value (65) with the system's preset "opportunity threshold" (50). Since the net benefit significantly exceeds the threshold, it indicates that the overtake can bring a sufficiently large benefit, and the system therefore ultimately decides to perform the overtake.

[0107] Based on this clear and quantitative analysis, the system made the final decision directly in step S2: to issue instruction Y (vehicle number 9, 55 BE) as the optimal instruction to the tire crane for execution. This round of scheduling decision-making then ended, thus successfully seizing a valuable overtaking opportunity that could significantly improve operational efficiency.

[0108] In addition to the above embodiments, this application also includes other implementation methods. All technical solutions formed by equivalent transformation or equivalent substitution should fall within the protection scope of the claims of this invention.

Claims

1. A method for scheduling instructions for in-span tire cranes based on multi-dimensional dynamic priority and predictive scheduling, characterized in that, Includes the following steps: Step S1: Generation of candidate instruction sets and hard constraint filtering; Collect all waiting operation instructions and filter them according to the preset hard constraint rules of the sequential order of instructions in the same vehicle to generate a set of candidate instructions that can be executed at present. Step S2: Assessment of the opportunistic overtaking window; The candidate instruction set is evaluated to determine whether there is an overtaking opportunity that meets the preset conditions. If it does, the net benefit of the overtaking operation is calculated by quantitative analysis. When the net benefit is greater than the preset opportunity threshold, the instruction with the closer target position in the queue is determined as the optimal instruction and the process proceeds directly to step S4. Otherwise, the process proceeds to step S3. Step S3: Non-linear dynamic sorting of regular instructions; If the optimal instruction is not determined in step S2, then for all instructions in the candidate instruction set, a comprehensive priority score is calculated for each instruction using a comprehensive priority function, and the instruction with the highest score is selected as the optimal instruction. The comprehensive priority function evaluation dimensions include at least: the distance the instruction travels, the waiting time of the instruction, and the local congestion situation in the instruction's target area; The comprehensive priority score is calculated as follows: based on the evaluation dimensions, corresponding walking distance factors, waiting time factors, and local congestion factors are generated for each candidate instruction, and then the scores of each factor are weighted and summed. The calculation of the instruction waiting time factor is related to the cumulative waiting time of the instruction, and when the cumulative waiting time of the instruction exceeds its average waiting time, its factor score increases exponentially. The calculation of the local congestion factor is inversely proportional to the number of vehicles already in the queue under the command target position; Step S4: Instruction issuance and rescheduling; The optimal instruction determined in step S2 or step S3 is issued to the tire crane for execution; and after the operation is completed, the system status is updated, and the process returns to step S1 to start a new round of scheduling decision-making.

2. The instruction scheduling method according to claim 1, characterized in that, The process of calculating the net benefit of the overtaking operation in step S2 includes: calculating the generalized total cost under the sequential execution strategy and the overtaking execution strategy respectively, and taking the difference between the two; wherein, the generalized total cost is a quantitative cost value that combines the total travel distance required for the tire crane to complete the work sequence and the total waiting time incurred by the vehicle as a result.

3. The instruction scheduling method according to claim 1, characterized in that, The calculation of the travel distance factor is inversely proportional to the distance from the current position of the tire crane to the target position of the command.

4. The instruction scheduling method according to claim 1, characterized in that, When the cumulative waiting time of any instruction reaches the preset maximum tolerance threshold, its priority is directly set to the highest.

Images