An unmanned gantry crane cooperative scheduling method for a smart land port logistics hub

Abstract

The present application relates to the field of railway freight automation, and provides an unmanned gantry crane cooperative scheduling method for a smart land port logistics hub, comprising collecting multi-source data of the unmanned gantry crane; constructing a dynamic priority scoring model to determine the optimal task instruction to be executed; generating a predicted trajectory within a future prediction time length according to the continuous time sequence state vector of each unmanned gantry crane, discretizing it into a multi-dimensional space-time position point sequence, and performing overlap detection within the future time period for any two unmanned gantry cranes to generate potential space-time conflict events in the future; constructing a multi-objective cost function to calculate the value of the generation to generate corresponding avoidance control parameters to generate a temporary exclusive virtual job corridor and issue control instructions. By introducing multi-source data deep fusion based on artificial intelligence, space-time conflict prediction and dynamic virtual job corridor control methods, the unmanned gantry crane realizes high-precision perception, high-reliability scheduling and multi-machine cooperative operation in the complex railway operation scene of the smart land port logistics hub.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of railway freight automation, and in particular to a method for cooperative scheduling of unmanned gantry cranes for smart land port logistics hubs. BACKGROUND

[0002] Under the development background of smart land port logistics hubs, railway container yards bear the core hub role of land freight distribution, and rail-mounted gantry cranes (RMG) as key equipment for yard loading, unloading, container shifting, and truck loading and unloading directly affect the overall throughput capacity of the yard. With the upgrading of smart land port systems to unmanned, intelligent, and collaborative, the number of devices in the yard is increasing, with multiple gantry cranes, multiple trucks, and multiple types of tasks coexisting, and the scheduling system needs to handle a large amount of real-time data, conflict judgment, and multi-device collaboration, significantly increasing system load.

[0003] Existing gantry crane scheduling methods generally use task allocation methods based on fixed rules or simple priority, mainly relying on single or a small number of sensors to obtain device location, bin information, and operation status. In terms of multi-device collaboration, traditional systems mainly use static avoidance rules (such as fixed work zones and interlocking areas) to avoid collisions; conflict judgment is mainly based on the current coordinates of the devices, without considering the future movement trends of the devices to predict potential conflicts. At the same time, existing technologies have limited task splitting capabilities for tasks from different sources such as yards, trains, and trucks, making it difficult to automatically generate instruction chains for cross-device collaboration based on factors such as block type, distance interval, and operation path. In addition, in the face of sensor errors, multi-source data asynchronization, environmental interference, and other factors, existing systems lack stable and reliable data fusion and self-calibration mechanisms. Therefore, existing technologies mainly have the following deficiencies:

[0004] (1) Multi-source perception data lacks a unified space-time reference required for smart land port scenarios, resulting in insufficient accuracy of position estimation and state judgment;

[0005] (2) Static task scheduling cannot adapt to the business needs of multi-operation mode integration in smart land ports, and existing methods cannot dynamically optimize based on hub operation mode, task waiting time, device state, path cost, and other factors;

[0006] (3) Unable to predict potential space-time conflicts in high-density operation scenarios in smart land ports, existing technologies can only judge conflicts based on the current coordinates of the devices, without considering potential space-time conflicts that may occur in the future trajectories of multiple gantry cranes, which can easily lead to path contention or motion interference. SUMMARY

[0007] In view of the technical problems of inconsistency of multi-source perception data, static scheduling rules, difficulty in predicting potential space-time conflicts between multiple gantry cranes, and insufficient coordination of cross-device instructions in the prior art, the present application proposes an unmanned gantry crane collaborative scheduling method for a smart land port logistics hub, which constructs a unified virtual berth coordinate system, realizes multi-modal state data fusion, automatically splits instructions based on task categories and path intervals, and predicts conflicts based on future motion trends, thereby realizing intelligent collaborative scheduling between multiple gantry cranes, internal trucks, and reach stackers.

[0008] To achieve the above-mentioned purpose, the present application realizes the following technical solutions:

[0009] The present application provides an unmanned gantry crane collaborative scheduling method for a smart land port logistics hub, which comprises the following steps:

[0010] Step S1: In response to the to-be-executed job instruction, a unified virtual berth coordinate system is established, and multi-source data of the positions, environments, and device states of all unmanned gantry cranes are collected;

[0011] Step S2: Based on the job mode, task priority, distance between the job target berth and the current position of the unmanned gantry crane, and instruction waiting time, a dynamic priority scoring model is constructed to score and sort the to-be-executed job instruction, and determine the optimal task instruction to be executed for each unmanned gantry crane; based on the block type of the starting position and the target position of the task and the task distance, the optimal task instruction is split into a sub-instruction sequence containing a crane, an internal truck, and a reach stacker;

[0012] Step S3: Under a unified time reference, the continuous time series state vector of each unmanned gantry crane is used to perform rolling trajectory deduction, and the predicted trajectories of the trolley, the car, and the hoist within the future prediction time are generated and discretized into a multi-dimensional space-time position point sequence; based on the multi-dimensional space-time position point sequence, three-dimensional safety envelope overlap detection is performed on any two unmanned gantry cranes within the future time period, and a future potential space-time conflict event is generated when overlap is detected;

[0013] Step S4: In response to the future potential space-time conflict event, a multi-objective cost function is constructed based on the task delay cost, path detour cost, and speed adjustment cost, the preset avoidance actions are calculated in terms of cost value, the avoidance action with the minimum cost value is selected, the corresponding avoidance control parameters are generated, and temporary exclusive virtual job corridors are generated for the unmanned gantry cranes that have priority to pass in the virtual berth coordinate system; under the running constraints of the virtual job corridors, the multiple unmanned gantry cranes are controlled in real time and the final control instructions are issued.

[0014] Further, step S2 comprises:

[0015] Obtaining task data in the to-be-executed job instruction, the task data including task identification, task type, starting position and target position and corresponding position type, task distance, task priority, instruction waiting time, container information and state data of associated equipment;

[0016] Based on the task data and the current position and current operation mode of the unmanned gantry crane, the corresponding normalized distance factor, operation mode factor, priority factor and time urgency factor are calculated;

[0017] According to the current operation mode, the weights of each factor are dynamically allocated, and the dynamic priority score of each task is obtained by weighted summation;

[0018] Based on the dynamic priority score, all to-be-executed instructions are sorted in descending order, and the task ranked first is determined as the optimal task instruction to be executed by the target unmanned gantry crane. At the same time, the optimal task instruction is split and verified. When the starting position and target position of the optimal task instruction have different position types and / or the task distance exceeds a preset threshold, the optimal task instruction is split into a sub-instruction sequence executed by the crane, the inner container truck and the front crane in sequence.

[0019] Further, based on the task data and the current position and current operation mode of the unmanned gantry crane, the corresponding normalized distance factor, operation mode factor, priority factor and time urgency factor are calculated, including:

[0020] Based on the absolute distance between the target position of the task and the current position of the unmanned gantry crane, the distance factor is obtained by inverse proportional normalization processing;

[0021] The basic score of the task type is obtained by querying the preset score table, and the operation mode factor is calculated by combining the dynamic weight of the current operation mode;

[0022] The preset priority score is queried according to the task priority, and the priority factor is calculated by normalization processing and multiplication of the corresponding weight;

[0023] Based on the instruction waiting time, the time urgency factor is calculated by using an incremental growth coefficient in different time intervals through a preset piecewise acceleration function.

[0024] Further, according to the current operation mode, the weights of each factor are dynamically allocated, and the dynamic priority score of each task is obtained by weighted summation, including:

[0025] According to the current operation mode, the weight configuration table is queried, and the corresponding dynamic weight coefficient is allocated to the distance factor, operation mode factor, priority factor and time urgency factor respectively;

[0026] The values of each factor and the corresponding weight coefficient are weighted and summed to obtain the dynamic priority score of each task.

[0027] Further, in step S1, the multi-source data of the positions, environments and equipment states of all unmanned gantry cranes are collected, including:

[0028] The local clocks of the laser radar, industrial camera, positioning module, environmental sensor and equipment control device are calibrated through the IEEE1588 precision clock synchronization protocol, and the multi-source data including the position, environment and equipment running state are collected under a unified time reference;

[0029] According to the pre-calibrated external parameter matrix, coordinate transformation is performed on the multi-source data, and is mapped to the berth coordinate system to realize unified space reference, and data cleaning and completion are performed on the mapped multi-source data, including abnormal point elimination based on a local sliding window and data missing completion using cubic spline interpolation.

[0030] Further, in step S1, the multi-source data of the positions, environments and equipment states of all unmanned gantry cranes are collected, including:

[0031] The dynamic confidence weight of each data source in the multi-source data is calculated based on the inherent reliability factor, environmental adaptability factor and cross consistency factor through linear weighting, and the multi-source data in the same time slice is weighted and fused according to the confidence weight to generate the continuous time sequence state vector of the real-time position, motion speed and working attitude of each unmanned gantry crane, wherein the inherent reliability factor is a static evaluation value obtained in the deployment stage according to the basic measurement performance, noise level and historical operation statistics of the sensor, the environmental adaptability factor is an environmental adaptability parameter dynamically determined according to real-time environmental perception data and combined with a pre-set environmental-performance mapping relationship, and the cross consistency factor is a consistency index calculated according to the weighted Euclidean distance of the current observation value of the data source and the observation value of the same type data source in the same time slice, and the value of the consistency index is the reciprocal of the weighted Euclidean distance.

[0032] Further, in step S3, the continuous time sequence state vector of each unmanned gantry crane is used to perform rolling trajectory deduction to generate the predicted trajectory of the trolley, the car and the hoist within a future prediction time, and to discretize it into a multi-dimensional space-time position point sequence, including:

[0033] Based on the continuous time sequence state vector of each unmanned gantry crane, a space-time trajectory sequence within a future preset prediction time is generated for each unmanned gantry crane;

[0034] The space-time trajectory sequence is discretized according to a unified time slice to obtain a discretized future path point set, and a multi-dimensional space-time position point sequence is constructed based on the future path point set to represent the spatial coordinates and working attitude parameters of each unmanned gantry crane at each time slice within the future prediction time.

[0035] Further, in step S3, based on the multi-dimensional space-time position point sequence, three-dimensional safety envelope overlap detection of any two unmanned gantry cranes in the future period is performed, and a future potential space-time conflict event is generated when overlap is detected, including:

[0036] Based on the multi-dimensional space-time position point sequence, a three-dimensional safety envelope body covering the future prediction length is constructed for each unmanned gantry crane, wherein the three-dimensional safety envelope body is a dynamic safety buffer zone formed according to the large car boundary, the trolley transverse width and the hoist lifting space of the unmanned gantry crane;

[0037] In the future prediction length, the three-dimensional safety envelope bodies corresponding to any two unmanned gantry cranes are subjected to spatial overlap detection at each time slice, wherein the spatial overlap detection includes performing intersection volume calculation based on the three-dimensional boundary box of the envelope body, and when the intersection volume is greater than zero, it is determined that there is envelope overlap in the corresponding time slice;

[0038] When the three-dimensional safety envelope bodies are detected to overlap in any future time slice, the corresponding future potential space-time conflict event is generated, and the time slice index and the on-duty device number of the future potential conflict event are recorded.

[0039] Further, in step S4, in response to the future potential space-time conflict event, a multi-objective cost function is constructed based on the task delay cost, the path detour cost and the speed adjustment cost, the value of each preset avoidance action is calculated, the avoidance action with the minimum value is selected, the corresponding avoidance control parameter is generated, and the temporary exclusive virtual working corridor for the unmanned gantry crane that has priority to pass in the virtual berth coordinate system is generated, including:

[0040] After detecting the future potential conflict event, the predicted trajectory of the unmanned gantry crane involved in the future potential conflict event and the corresponding conflict time slice are extracted;

[0041] For each preset avoidance action, the adjustment amount of the avoidance action under the conflict time slice is calculated based on the predicted trajectory, and the avoidance action includes a speed adjustment action, a path detour action and a first-pass action; wherein,

[0042] The adjustment amount of the speed adjustment action is the minimum speed change amount of the target unmanned gantry crane passing through the conflict space before and after the conflict time slice; the adjustment amount of the path detour action is the minimum path offset amount of the target unmanned gantry crane not entering the conflict envelope at the conflict time slice; and the adjustment amount of the first-pass action is the minimum time delay amount of the target unmanned gantry crane passing through again after the conflict envelope is released;

[0043] The corresponding task delay cost, path detour cost and speed adjustment cost are calculated based on the adjustment amount, a multi-objective cost function is weighted and summed according to preset weight coefficients, a corresponding cost value is obtained, the avoidance action with the minimum cost value is taken as the target avoidance action, and the corresponding avoidance control parameter is obtained based on the target avoidance action;

[0044] A target berth interval corresponding to the avoidance control parameter is determined in the berth coordinate system, and a virtual work corridor is constructed in the target berth interval; wherein the virtual work corridor is a temporary exclusive running area that only allows the target unmanned gantry crane to enter within a preset duration; the virtual work corridor is marked as a temporary exclusive running constraint of the target unmanned gantry crane, and the virtual work corridor and the corresponding temporary exclusive running constraint are automatically released after the target unmanned gantry crane completes the crossing of the target berth interval.

[0045] Further, in step S4, the multiple unmanned gantry cranes are controlled in real time and the final control instructions are issued, including:

[0046] Based on the temporary exclusive running constraint of the virtual work corridor, each unmanned gantry crane is dynamically assigned a corresponding executable work instruction, so as to ensure that the work path and the work target berth do not invade the virtual work corridor and the safety limiting area of any unmanned gantry crane;

[0047] The instruction execution state and the equipment running state of each unmanned gantry crane are monitored in real time, and when a device fault, a work timeout or a new space-time conflict event is detected, an instruction rescheduling and conflict resolution process is triggered immediately;

[0048] According to the dependency relationship of the sub-instruction sequence, the internal spreader and the reach stacker are cooperatively scheduled to ensure that the internal spreader or the reach stacker can arrive at the specified position in time for the connection work after the crane completes the work;

[0049] Based on the real-time cooperative scheduling control, control instructions containing the target position, the running speed and the work posture parameters of the work are issued to the corresponding unmanned gantry crane, the internal spreader and the reach stacker.

[0050] Compared with the prior art, the present application has at least one of the following beneficial effects:

[0051] The present application realizes high-precision perception, high-reliability scheduling and multi-machine collaborative operation of the unmanned gantry crane in the complex railway operation scene of the intelligent land port logistics hub by introducing a multi-source data deep fusion method based on artificial intelligence, a time-space conflict prediction model and a dynamic virtual operation corridor control method. Heterogeneous data such as laser radar, vision, environment and weather, track strain and cargo weight can be fused under a unified time-space reference, solving the problem of insufficient perception accuracy caused by inconsistent sampling frequencies of multi-source data and information islands in traditional systems. At the same time, by establishing a three-dimensional envelope overlap prediction for future time periods, potential operation conflicts are identified in advance and a multi-objective cost function is used to optimize the avoidance strategy, significantly reducing the path interference and operation conflict probability in the multi-machine collaborative process. Using a dynamic priority scoring model and task splitting mechanism, the allocation of operation instructions is more in line with the on-site operation rules, improving the continuity of the operation link and the overall utilization efficiency of the equipment. The time-periodic exclusive constraint realized by the virtual operation corridor can realize fine avoidance and travel coordination without affecting the overall throughput capacity. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 The flowchart of the steps of the unmanned gantry crane collaborative scheduling method of the present application for the intelligent land port logistics hub. DETAILED DESCRIPTION

[0053] To make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] Those skilled in the art can understand that, unless specifically stated otherwise, the singular forms "a", "an" and "the" used herein can also include the plural forms. It should be further understood that the phrase "comprising" used in the specification of the present application means that the features, integers, steps, operations, elements and / or components exist, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0055] First embodiment

[0056] In the prior art, the scheduling of unmanned gantry cranes in railway yards generally relies on fixed rules for task allocation and path avoidance. Common methods include simple instruction sorting based on task type or manually set priority, nearest distance allocation combined with current position of the trolley, and static lock position or fixed waiting strategy to avoid path conflicts between devices. Although such scheduling methods can achieve basic job flow in actual scenarios, in a smart land and port logistics hub with high concurrency, multi-node linkage, and multi-type job subject collaboration, the scheduling decisions often only consider the current device position without considering the evolution of future trajectories. With increasing job density, path competition, invalid waiting, and sudden stop avoidance may occur when multiple devices are running simultaneously, resulting in decreased overall throughput efficiency of the yard. In addition, due to the ineffective fusion of multi-source sensor data, there is a time delay and error accumulation in the spatial information between different devices, which makes it difficult for the scheduling system to accurately determine potential future conflicts. The static lock area of traditional methods often occupies a large space, significantly reducing the available work area and further limiting overall efficiency.

[0057] The inventors found that the above efficiency bottleneck is not only due to the shortcomings of task allocation rules, but also due to the lack of future-oriented prediction capabilities, lack of quantifiable conflict resolution strategies, and lack of dynamic management capabilities for spatial resources in the scheduling system. In particular, in areas with multiple machine intersections, if the future trajectories of the trolleys, trolleys, and hoists of each unmanned gantry crane can be rolled forward and potential future space-time conflict events can be identified in advance, active avoidance can be performed tens of seconds before the conflict occurs. At the same time, by constructing a multi-objective cost function including task delay cost, path detour cost, and speed adjustment cost, the overall impact of different avoidance actions can be quantitatively evaluated, thereby realizing the transition from fixed stop rules to dynamic optimal decisions. In particular, in the multi-machine intersection area of a smart land and port logistics hub, if the future trajectories of the trolleys, trolleys, and hoists of each unmanned gantry crane can be rolled forward and potential future space-time conflict events can be identified in advance, active avoidance can be performed tens of seconds before the conflict occurs, significantly improving the safety and continuity of the system. At the same time, by constructing a multi-objective cost function including task delay cost, path detour cost, and speed adjustment cost, the overall impact of different avoidance actions can be quantitatively evaluated, thereby realizing the fundamental transition from static rules to dynamic optimal decisions. Further, to avoid the rough occupation of space by traditional lock methods, the present application proposes a virtual work corridor as a dynamic exclusive area, allowing devices to only temporarily occupy the corresponding area during necessary time slices, thereby significantly improving the spatial reuse efficiency of the yard.

[0058] Based on the above analysis and technical considerations, this invention proposes a collaborative scheduling method for unmanned gantry cranes in smart land port logistics hubs. Through key steps such as dynamic task priority scoring, rolling future trajectory projection, three-dimensional safety envelope overlap detection, multi-objective cost solving, and virtual operation corridors, it achieves forward-looking collaborative scheduling of multiple unmanned gantry cranes. This scheme can proactively avoid conflicts before they occur, preventing equipment waiting and downtime; it enables refined spatial management, improving bay space utilization; and it makes multi-equipment operation chains more compact, reducing path redundancy and connection delays. Compared with existing technologies, this invention significantly reduces the incidence of spatiotemporal conflicts, improves equipment operating efficiency and overall throughput. Furthermore, although gantry cranes are the primary execution equipment, container handling tasks in railway freight scenarios often involve multiple different areas. To ensure that unmanned gantry cranes can obtain the correct target position throughout the entire process, this invention introduces the participation logic of internal trucks and reach stackers into the scheduling algorithm to handle handling tasks that cross street blocks, cross regulatory zones, or exceed designated areas. It is important to emphasize that the core scheduling object of this invention is the unmanned gantry crane; other equipment serves only as auxiliary operational links in the task chain to ensure the continuity and optimization of gantry crane scheduling decisions. The specific implementation method is as follows:

[0059] like Figure 1 As shown, this invention provides a collaborative scheduling method for unmanned gantry cranes in smart land port logistics hubs, comprising the following steps:

[0060] Step S1: Respond to the work instructions to be executed, establish a unified virtual position coordinate system, and collect multi-source data on the position, environment, and equipment status of all unmanned gantry cranes;

[0061] Step S2: Based on the operation mode, task priority, distance between the target position and the current position of the unmanned gantry crane, and instruction waiting time, construct a dynamic priority scoring model to score and sort the operation instructions to be executed, and determine the optimal task instruction to be executed for each unmanned gantry crane; based on the street type and task distance between the starting position and the target position of the task, the optimal task instruction is broken down into a sub-instruction sequence containing crane, internal truck and reach stacker.

[0062] Step S3: Under a unified time reference, the rolling trajectory is extrapolated based on the continuous temporal state vector of each unmanned gantry crane to generate the predicted trajectory of the trolley, crane and spreader within the future prediction time. The trajectory is discretized into a multi-dimensional spatiotemporal position point sequence. Based on the multi-dimensional spatiotemporal position point sequence, the three-dimensional safety envelope overlap detection is performed on any two unmanned gantry cranes in the future time period. When overlap is detected, a potential future spatiotemporal conflict event is generated.

[0063] Step S4: In response to a future potential space-time conflict event, a multi-objective cost function is constructed based on the task delay cost, path detour cost and speed adjustment cost, the cost value of each preset avoidance action is calculated, the avoidance action with the minimum cost value is selected, the corresponding avoidance control parameter is generated, and the temporary exclusive virtual operation corridor for the unmanned gantry crane that has priority to pass in the virtual berth coordinate system is generated. Under the operation constraints of the virtual operation corridor, the multiple unmanned gantry cranes are controlled in real time and the final control instruction is issued.

[0064] Further, step S2 comprises:

[0065] Obtaining task data in the to-be-executed operation instruction, the task data including task identification, task type, starting position and target position and corresponding position type, task distance, task priority, instruction waiting time, container information and state data of associated equipment;

[0066] Based on the task data and the current position and current operation mode of the unmanned gantry crane, corresponding normalized distance factor, operation mode factor, priority factor and time urgency factor are calculated;

[0067] According to the current operation mode, weights are dynamically assigned to each factor, and a dynamic priority score of each task is obtained by weighted summation;

[0068] Based on the dynamic priority score, all to-be-executed instructions are sorted in descending order, and the task ranked first is determined as the optimal task instruction to be executed by the target unmanned gantry crane. At the same time, the optimal task instruction is split and verified. When the starting position and target position of the optimal task instruction have different position types and / or the task distance exceeds a preset threshold, the optimal task instruction is split into a sub-instruction sequence executed by the crane, the inner container truck and the front-end crane in sequence.

[0069] Specifically, it is determined whether an original task needs to be split into multiple executable sub-instructions;

[0070] It is determined whether the starting position and target position of the task have the same position type, for example, a yard-yard task (LL)

[0071] When the TaskStartingPosition and the TaskTargetPosition do not belong to the same block, the processor splits the task into a three-segment instruction chain of "crane, inner truck and reach stacker", when they belong to the same block, if the area belongs to the customs supervision area, only the reach stacker instruction is generated; if it belongs to the main block, it is judged whether the bay interval it crosses exceeds the threshold SplitIntervalNJK, when it exceeds the threshold or is not in the same beam, it is also split into a three-segment instruction of crane, inner truck and crane, otherwise only the crane instruction is generated.

[0072] Further, based on the task data and the current position and current operation mode of the unmanned gantry crane, corresponding normalized distance factor, operation mode factor, priority factor and time urgency factor are calculated, including:

[0073] Based on the absolute distance between the target position of the task and the current position of the unmanned gantry crane, the distance factor is obtained by normalization processing and multiplication by a preset weight;

[0074] The basic score of the task type is obtained by querying the preset score table, and the operation mode factor is calculated by combining the dynamic weight of the current operation mode;

[0075] The priority score is queried according to the task priority, and the priority factor is calculated by normalization processing and multiplication by the corresponding weight;

[0076] Based on the instruction waiting time, the time urgency factor is calculated by using an incremental growth coefficient in different time intervals through a preset piecewise acceleration function.

[0077] Further, according to the current operation mode, the weights of each factor are dynamically allocated, and the dynamic priority score of each task is obtained by weighted summation, including:

[0078] According to the current operation mode, the weight configuration table is queried, and the corresponding dynamic weight coefficients are allocated to the distance factor, the operation mode factor, the priority factor and the time urgency factor respectively;

[0079] The values of each factor and the corresponding weight coefficients are weighted and summed to obtain the dynamic priority score of each task.

[0080] Specifically, before the task scoring and sorting in step S2 is performed, the scheduling mode is first determined based on the current operation scenario. The scheduling mode is automatically switched by the processor according to the global operation state, task type distribution and real-time request source, including four types of intelligent balanced mode, loading priority mode, unloading priority mode and truck priority mode.

[0081] In the intelligent balancing mode, the processor reads the task type, priority, standby time and target berth of all tasks to be executed from the memory, calls a unified weight configuration table, assigns similar basic weights to different operation types such as loading, unloading, truck and yard unloading, and performs comprehensive ranking according to the scoring model of step S2 to maintain overall balance between different operation types and avoid long-term backlog of a certain type of task.

[0082] In the loading priority mode, the processor significantly increases the weight of the operation mode factor corresponding to the train loading task (for example, S_mode is increased from 0.2 to 0.4-0.5), so that the train loading task is arranged in priority in the comprehensive score; when there is a train loading task, only the priority score and scheduling of the train loading task are performed, and only after all train loading tasks are completed, the scheduling of other types of tasks is resumed.

[0083] In the unloading priority mode, the processor increases the operation mode weight of the train unloading task and reduces the weight of other types of tasks, so that the train unloading task is arranged in priority in the scoring model; when there is a train unloading task, the system only performs priority sorting on the train unloading related tasks, and resumes the normal scheduling logic after the train unloading task is completed.

[0084] In the truck priority mode, the processor increases the priority weight of the yard-truck and truck-yard tasks triggered by the truck, and automatically increases the weight proportion of the operation mode factor S_mode and the time urgency factor S_time in the scoring process, so as to reduce the waiting time of the external truck and the internal truck; when there are multiple truck related tasks, only the priority sorting of the truck task is performed, and after all the truck tasks are completed, other types of tasks are scheduled

[0085] Further, in step S1, the multi-source data of the position, environment and equipment state of all unmanned gantry cranes are collected, including:

[0086] The local clocks of the laser radar, industrial camera, positioning module, environmental sensor and equipment control device are calibrated through the IEEE1588 precision clock synchronization protocol, and the multi-source data including position, environment and equipment running state are collected under the unified time reference;

[0087] According to the pre-calibrated extrinsic matrix, coordinate transformation is performed on the multi-source data, and is mapped to the berth coordinate system to realize the unification of the space reference, and data cleaning and completion are performed on the mapped multi-source data, including abnormal point elimination based on local sliding window and data missing completion by cubic spline interpolation.

[0088] Further, in step S1, the multi-source data of the position, environment and equipment state of all unmanned gantry cranes are collected, including:

[0089] The dynamic confidence weight of each data source in the multi-source data is calculated based on the inherent reliability factor, the environmental adaptability factor and the cross-consistency factor through linear weighting, and the multi-source data in the same time slice is weighted and fused according to the confidence weight, to generate the continuous time sequence state vector of the real-time position, motion speed and working posture of each unmanned gantry crane, wherein the inherent reliability factor is a static evaluation value obtained in the deployment stage according to the basic measurement performance, noise level and historical operation statistics of the sensor, the environmental adaptability factor is an environmental adaptability parameter dynamically determined according to the real-time environmental perception data and the preset environmental-performance mapping relationship, and the cross-consistency factor is a consistency index calculated according to the weighted Euclidean distance between the current observation value of the data source and the observation value of the same type of data source, and the value of the consistency index is the reciprocal of the weighted Euclidean distance.

[0090] Specifically, in step S1, first, the laser radar (such as Velodyne VLP-16), the industrial camera (such as 2 million pixels @ 60fps), the positioning module, the environmental sensor and the local clock of the device control unit deployed on the unmanned gantry crane are high-precision synchronized. The above-mentioned sensing units are aligned with the main clock of the scheduling system through the IEEE1588 (PTP) precise clock synchronization protocol, so that all the data collected by different sensors are provided with unified global time stamps, thereby ensuring the consistency of the subsequent fusion processing in the time dimension.

[0091] After completing the time synchronization, a unified spatial coordinate system with the unmanned gantry crane basic coordinate system as the reference is established based on the pre-calibrated external parameter matrix, and the laser radar point cloud data, the visual detection frame, the device posture and the running parameters are mapped to the coordinate system, forming a spatially aligned data description framework. For the abnormal observation points generated in the mapping process, the system uses Grubbs criterion to remove outliers based on the local sliding window statistical characteristics, and uses cubic spline interpolation method to complete the data vacancy area formed after the removal, to ensure the continuity and stability of the fused data.

[0092] After completing the data regularization, dynamic confidence weighted fusion is performed on the multi-source data. For this purpose, a corresponding data source model is established for the laser radar, the industrial camera and other sensors, and a dynamic confidence weight is calculated for each type of data source. The confidence weight is composed of the inherent reliability factor, the environmental adaptability factor and the cross-consistency factor: the inherent reliability factor is set based on the factory accuracy and historical stability of the sensor; the environmental adaptability factor is dynamically adjusted according to the current light, rain, snow, dust and other environmental states and the preset "environment-performance mapping table"; and the cross-consistency factor is obtained by calculating the weighted Euclidean distance between the current observation value and the observation value of the same type of sensor and taking its reciprocal, to represent the consistency between different observation sources.

[0093] According to the above credibility weight, low-level fusion is performed on directly observable numerical data, such as container center coordinates, device attitude parameters, etc. The credibility weighted average is used to generate an optimized estimation value. For abstract states such as "whether the track is passable" and "whether the operation channel is open", D-S evidence theory is used for high-level fusion. Track strain data, visual detection results, and maintenance records are used as independent evidence sources to calculate basic probability distribution, and finally the confidence distribution of "safe-warning-danger" state is synthesized.

[0094] In addition, in order to improve the stability of long-term operation, a closed-loop calibration mechanism is also constructed in this step. The operation instructions generated based on the above fusion results are compared with the real motion trajectory feedback by the high-precision RTK system after the actual execution of the unmanned gantry crane. When abnormal deviation is detected between the two, the inherent reliability factor of the corresponding sensor is automatically corrected, so that the fusion model maintains precision stability in long-term operation.

[0095] For example, taking three unmanned gantry cranes in a railway container yard as an illustrative object, the collection, regularization and fusion process of multi-source data in step S1 is described. Before the operation starts, the laser radar (such as Velodyne VLP-16), industrial camera (200 million pixels @ 60fps), RTK positioning module, environmental sensor and device controller deployed on each unmanned gantry crane are aligned with the master clock of the scheduling center through IEEE1588 precise clock synchronization protocol, so that the point cloud data, image data, attitude parameters and running state collected by different sensors are all with unified global time stamp. For example, at the same time 08:00:01, the data collected by the laser radar, industrial camera and environmental module of G1 all correspond to the same time slice, avoiding the spatial calculation error caused by millisecond-level time drift in traditional systems.

[0096] After time synchronization, all kinds of perception data are mapped to the unified spatial coordinate system in terms of the external parameter matrix calibrated in the deployment phase. For example, the point cloud of the container top surface detected by the G1 laser radar is (1.2, -0.8, 3.5) m in the radar coordinate system, and after transformation, the corresponding coordinates of the bay are about (B17.2, 12.6 m, 4.9 m). The target center point calculated by the industrial camera detection frame is (B17.3, 12.7 m, 5.0 m) after coordinate conversion. Both are automatically recognized by the system as the same target object. On this basis, the system performs quality check and data completion on the mapped data. When a device is disturbed by wind and snow, resulting in abnormal points in the laser radar point cloud or local time slice missing frames, the system can identify these outliers through a local sliding window statistical model and remove them according to the Grubbs criterion. For the data gaps formed after removal, a continuous point cloud sequence is generated by cubic spline interpolation, so that the input data remains stable and complete.

[0097] After space-time normalization, dynamic confidence weights are assigned to each data source, and fusion processing is performed accordingly. Under certain snow weather conditions, the images collected by the G1 industrial camera have obvious noise, and its environmental adaptability factor decreases with the decrease of light, while the laser radar is not significantly affected, and its environmental factor remains at a high level. Based on the inherent reliability factor, environmental adaptability factor and cross-sensor consistency factor, the confidence weights of laser radar and industrial camera are adjusted in real time, so that the weight of laser radar in fusion under snow conditions is significantly improved, avoiding misjudgment caused by visual noise. Using the above weight, quantifiable information such as container center coordinates and gantry crane attitude is subjected to low-level weighted fusion. For example, in the observation of the B25 region by G2, the laser radar and the industrial camera give the position estimates of B25.1 and B24.9 respectively, and after weighted fusion, the system obtains a more stable fusion coordinate B25.03. For abstract states such as whether there is an obstacle on the track, the D-S evidence theory is used, and the track strain gauge, visual detection and maintenance records are used as independent evidence sources to infer the confidence distribution of "safe, warning, dangerous" states. When the "danger" probability in the fusion confidence exceeds the threshold, the operation restriction of the track being impassable is generated.

[0098] In addition, the embodiment constructs a closed-loop calibration mechanism. When G1 executes a "go to B18" instruction, if it is predicted that it will reach B17.8 after 2 seconds, but the actual RTK returned position is B17.5, it is determined that there is an offset error and the inherent reliability factor of the laser radar is automatically reduced, so that subsequent fusion relies more on visual data or other sensors. After continuous calibration, the three-dimensional position estimation error converges from the original ±9 cm to ±3 cm, thereby continuously ensuring the accuracy and reliability of the unmanned gantry crane state vector.

[0099] In step S2, in order to accurately quantify the priority order of task scheduling, the processor constructs a dynamic priority scoring function , wherein represents the i-th to-be-executed task. The scoring model fuses multiple business factors in a linear weighting manner, and the specific calculation formula is as follows:

[0100] ;

[0101] , wherein , respectively, are dynamic weight coefficients of a distance factor , a job mode factor , a priority factor , and a time urgency factor , satisfying a normalization condition . The system dynamically adjusts the above weights according to the current job mode (such as a loading priority mode or an intelligent balanced mode). The specific calculation logic of each factor is as follows:

[0102] (1) Distance factor : used to represent the proximity of the task. The Manhattan distance algorithm is used to calculate the path distance between the current position of the unmanned gantry crane and the target position of the task , and an inverse proportional normalization processing is adopted:

[0103] ;

[0104] , wherein is the maximum diagonal distance of the yard map, is the minimum job safety distance. The formula ensures that the task with closer distance obtains higher distance factor score.

[0105] (2) Time urgency factor : in order to avoid long-term accumulation of low-priority tasks, a segmented growth function (i.e., the aforementioned segmented acceleration function) is designed. Let be the waiting time length of the instruction entering the queue, be a preset timeout threshold, then

[0106]

[0107] , wherein are both preset positive growth coefficients. The function shows that within the threshold range, the urgency linearly increases with time; once the threshold is exceeded, the urgency rises exponentially, forcing the scheduling order of the task to be promoted.

[0108] (3) Job mode factor and priority factor​ : Obtain the corresponding basic score according to the currently selected scheduling mode (such as loading priority or unloading priority) and perform normalized mapping.

[0109] Through the above model, the value of all tasks in the queue can be calculated in real time, and the optimal task sequence is regenerated accordingly.

[0110] Further, in step S3, the continuous time sequence state vector of each unmanned gantry crane is used to perform rolling trajectory deduction to generate the predicted trajectory of the trolley, the trolley and the spreader in the future prediction time length, and to discretize it into a multi-dimensional space-time position point sequence, including:

[0111] Based on the continuous time sequence state vector of each unmanned gantry crane, a space-time trajectory sequence of each unmanned gantry crane in the future preset prediction time length is generated, the space-time trajectory sequence is discretized according to the same time slice to obtain a discretized future path point set, and a multi-dimensional space-time position point sequence is constructed based on the future path point set to represent the spatial coordinates and operation posture parameters of each unmanned gantry crane at each time slice in the future prediction time length.

[0112] Specifically, the discretized state vector of any one unmanned gantry crane at time is defined as , the three-dimensional space coordinates of the gantry crane, the corresponding speed of the gantry crane at the three-dimensional space coordinates. Based on the uniform variable speed kinematics principle, the predicted position state of the device at the future time slice (i.e. time) is predicted:

[0113] ;

[0114] wherein, is the three-axis acceleration vector estimated according to the current control instruction (acceleration, deceleration or uniform speed), is the system time slice length.

[0115] Further, in step S3, based on the multi-dimensional space-time position point sequence, three-dimensional safety envelope overlap detection is performed for any two unmanned gantry cranes in the future period, and a future potential space-time conflict event is generated when overlap is detected, including:

[0116] Based on the multi-dimensional space-time position point sequence, a three-dimensional safety envelope body covering the future prediction time length is constructed for each unmanned gantry crane, wherein the three-dimensional safety envelope body is a space envelope region formed according to the trolley boundary, the trolley transverse width and the spreader lifting space of the unmanned gantry crane, and is dynamically updated with the time slice. Specifically, for each gantry crane At any prediction time point Constructing a set of three-dimensional safety envelopes The envelope contains not only the rigid size of the device, but also a dynamic safety buffer , is the three-dimensional coordinate of the point to be measured, wherein, is the predicted center coordinate, is the physical half-length and half-width of the device, is the safety buffer distance associated with the speed, is the maximum reachable working height of the spreader, , is the safety expansion coefficient related to the speed, is the speed modulus of the unmanned gantry crane, is the minimum safety buffer coefficient, that is, the faster the speed, the larger the reserved braking buffer area.

[0117] In the future prediction time, the three-dimensional safety envelopes corresponding to any two unmanned gantry cranes are detected for spatial overlap in each time slice, wherein the spatial overlap detection includes performing intersection volume calculation based on the three-dimensional bounding box of the envelope, and when the intersection volume is greater than zero, it is determined that there is envelope overlap in the corresponding time slice. When the three-dimensional safety envelope is detected to overlap in any future time slice, a corresponding future potential space-time conflict event is generated, and the time slice index and on-duty device number at which the future potential conflict event occurs are recorded. Specifically, the system performs Boolean operation detection on any two adjacent gantry cranes and in the future time period. If there is any time slice satisfying the following volume intersection condition in the future prediction time window

[0118] ;

[0119] that is, it is determined that there is a "future potential space-time conflict event" in the corresponding time slice, and the system records the start time slice and the end time slice as well as the device numbers and as input parameters for subsequent avoidance algorithms.

[0120] Further, in step S4, in response to the future potential space-time conflict event, a multi-objective cost function is constructed based on the task delay cost, the path detour cost, and the speed adjustment cost, the cost value of each preset avoidance action is calculated, the avoidance action with the minimum cost value is selected, the corresponding avoidance control parameter is generated, and the avoidance control parameter generates a temporary exclusive virtual working corridor for the unmanned gantry crane that has priority to pass in the virtual berth coordinate system, including:

[0121] ​After detecting the future potential conflict event, a predicted trajectory of the unmanned gantry crane involved in the future potential conflict event and a corresponding conflict time slice are extracted;

[0122] For each preset avoidance action, an adjustment amount of the avoidance action under the conflict time slice is calculated based on the predicted trajectory, the avoidance action including a speed adjustment action, a path detour action, and a first-passing action; wherein,

[0123] The adjustment amount of the speed adjustment action is a minimum speed change amount of the target unmanned gantry crane passing through the conflict space before and after the conflict time slice; the adjustment amount of the path detour action is a minimum path offset amount of the target unmanned gantry crane not entering the conflict envelope at the conflict time slice; and the adjustment amount of the first-passing action is a minimum time delay amount of the target unmanned gantry crane passing through again after the conflict envelope is released.

[0124] Based on the adjustment amounts, a corresponding task delay cost, path detour cost, and speed adjustment cost are calculated, a multi-objective cost function is weighted and summed according to preset weight coefficients, a corresponding cost value is obtained, the avoidance action with the minimum cost value is taken as a target avoidance action, and corresponding avoidance control parameters are obtained based on the target avoidance action.

[0125] When the avoidance strategy is generated, instead of simply performing an emergency stop, a multi-objective cost function is constructed , various candidate avoidance parameter vectors such as speed adjustment, path detour, and first-passing are evaluated , aiming to find the optimal solution with the minimum total cost value :

[0126] ;

[0127] wherein, is a weight coefficient of each sub-cost, and the physical meaning and calculation method of each sub-cost function are as follows:

[0128] 1. Task delay cost reflects the influence of the avoidance action on the work efficiency, wherein is a task sequence before and after avoidance , and the length of time when the task is forced to stop or slow down before entering the conflict area.

[0129] 2. Path detour cost reflects the occupation cost of space resources, that is, the deviation integral of the avoidance trajectory relative to the theoretically optimal straight-line trajectory .

[0130] 3. Energy consumption cost , reflecting the additional energy consumption and mechanical wear caused by acceleration and deceleration operation, which punishes frequent acceleration and deceleration operation by calculating the sum of squares of speed change.

[0131] 4. Smoothness of operation cost , used to ensure the stability of the anti-roll system, which punishes longitudinal or lateral acceleration peaks exceeding the anti-roll threshold.

[0132] 5. The system calculates the above values by traversing the preset avoidance action library (such as: micro-deceleration 10%, left side detour 0.5 meters, waiting 5 seconds, etc.), and finally selects the control instruction of the action with the smallest value, and establishes a virtual operation corridor in the corresponding space-time region to achieve optimal avoidance.

[0133] Determine the target berth interval corresponding to the avoidance control parameter in the berth coordinate system, and construct a virtual operation corridor in the target berth interval; wherein the virtual operation corridor is a temporary exclusive operation area that only allows the target unmanned gantry crane to enter within a preset duration; mark the virtual operation corridor as a temporary exclusive operation constraint of the target unmanned gantry crane, and automatically release the virtual operation corridor and the corresponding temporary exclusive operation constraint after the target unmanned gantry crane completes the crossing of the target berth interval.

[0134] Further, in step S4, the multiple unmanned gantry cranes are controlled in real time and the final control instruction is issued, including:

[0135] Based on the temporary exclusive operation constraint of the virtual operation corridor, dynamically allocate corresponding executable operation instructions to each unmanned gantry crane to ensure that the operation path and operation target berth do not invade the virtual operation corridor and safety limit area of any unmanned gantry crane;

[0136] Real-time monitoring of the instruction execution state and equipment operation state of each unmanned gantry crane, when detecting equipment failure, operation timeout or new space-time conflict event, immediately trigger instruction rescheduling and conflict resolution process;

[0137] According to the dependency relationship of the sub-instruction sequence, the internal spreader and the front crane are cooperatively scheduled to ensure that the internal spreader or the front crane can arrive at the designated position on time for connection operation after the crane completes the operation.

[0138] Based on real-time cooperative scheduling control, the control instruction containing the target position, operation speed and operation posture parameters of the operation is issued to the corresponding unmanned gantry crane, internal spreader and front crane.

[0139] Having described above several embodiments of the disclosure, any modifications and variations that fall within the scope of the described embodiments are also intended to be within the scope of the disclosure. As will be apparent to those skilled in the art, some modifications and variations to the embodiments described above can be practiced while staying within the scope and spirit of the described embodiments. The foregoing description of the described embodiments has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the described embodiments to the precise form disclosed. Many modifications and variations are possible in light of the above teachings. It is intended that the disclosed embodiments be limited only by the claims.

Claims

1. A method for collaborative scheduling of unmanned gantry cranes in a smart land port logistics hub, characterized by the steps of Comprise: Step S1: in response to the to-be-executed job instruction, a unified virtual berth coordinate system is established, and multi-source data of the positions, environments and equipment states of all unmanned gantry cranes are collected; Step S2: based on the job mode, the task priority, the distance between the target berth of the job and the current position of the unmanned gantry crane, and the instruction waiting time, a dynamic priority scoring model is constructed, the to-be-executed job instruction is scored and sorted, and the optimal task instruction to be executed for each unmanned gantry crane is determined; based on the block types of the starting position and the target position of the task and the task distance, the optimal task instruction is split into a sub-instruction sequence including a crane, an inner container truck and a front loader; Step S3: under a unified time reference, the continuous time series state vectors of each unmanned gantry crane are used for rolling trajectory deduction, the predicted trajectories of the trolley, the trolley and the spreader within a future prediction time period are generated, and are discretized into a multi-dimensional space-time position point sequence; based on the multi-dimensional space-time position point sequence, three-dimensional safety envelope overlap detection of any two unmanned gantry cranes within a future period is performed, and a future potential space-time conflict event is generated when overlap is detected; Step S4: in response to the future potential space-time conflict event, a multi-objective cost function is constructed based on the task delay cost, the path detour cost and the speed adjustment cost, the cost values of preset avoidance actions are calculated, the avoidance action with the minimum cost value is selected, corresponding avoidance control parameters are generated, and the avoidance control parameters generate a temporary exclusive virtual job corridor for the unmanned gantry crane that has priority to pass in the berth coordinate system; under the running constraints of the virtual job corridor, real-time collaborative scheduling control is performed on multiple unmanned gantry cranes, and the final control instruction is issued.

2. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 1, characterized in that, Step S2 comprises: Obtaining task data in the to-be-executed job instruction, the task data including task identification, task type, starting position and target position and corresponding position type, task distance, task priority, instruction waiting time, container information and state data of associated equipment; Based on the task data and the current position and current job mode of the unmanned gantry crane, corresponding normalized distance factor, job mode factor, priority factor and time urgency factor are calculated; According to the current job mode, weights are dynamically assigned to each factor, and the dynamic priority score of each task is obtained by weighted summation; Based on the dynamic priority score, all the to-be-executed job instructions are sorted in descending order, and the task ranked first is determined as the optimal task instruction to be executed by the target unmanned gantry crane; meanwhile, the optimal task instruction is split and verified, when the starting position and the target position of the optimal task instruction have different position types and / or the task distance exceeds a preset threshold, the optimal task instruction is split into a sub-instruction sequence executed by the crane, the inner container truck and the front loader in sequence.

3. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 2, characterized in that, Based on the task data and the current position and current operation mode of the unmanned gantry crane, a corresponding normalized distance factor, operation mode factor, priority factor and time urgency factor are calculated, including: Based on the absolute distance between the target position of the task and the current position of the unmanned gantry crane, the distance factor is obtained by inverse proportional normalization processing; The basic score of the task type is obtained by querying a preset score table, and the operation mode factor is calculated by combining the dynamic weight of the current operation mode; The priority factor is calculated by querying a preset priority score, normalizing processing and multiplying the corresponding weight. Based on the instruction waiting time, the time urgency factor is calculated by using an incremental growth coefficient in different time intervals through a preset segmented acceleration function.

4. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 2, characterized in that, According to the current operation mode, the dynamic priority score of each task is obtained by dynamically allocating weights to each factor and performing weighted summation, including: According to the current operation mode, the distance factor, the operation mode factor, the priority factor and the time urgency factor are respectively allocated corresponding dynamic weight coefficients by querying a predefined weight configuration table; The values of each factor are weighted and summed with the corresponding weight coefficients to obtain the dynamic priority score of each task.

5. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 4, characterized in that, In step S1, the multi-source data of the position, environment and equipment state of all unmanned gantry cranes are collected, including: The local clocks of the laser radar, industrial camera, positioning module, environmental sensor and equipment control device are calibrated through the IEEE1588 precision clock synchronization protocol, and the multi-source data including position, environment and equipment running state are collected under the unified time reference; According to the pre-marked extrinsic parameter matrix, coordinate transformation is performed on the multi-source data, and is mapped to the berth coordinate system to realize space reference unification, and data cleaning and completion are performed on the mapped multi-source data, including abnormal point rejection based on local sliding window and data missing completion using cubic spline interpolation.

6. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 5, characterized in that, In step S1, the multi-source data of the position, environment and equipment state of all unmanned gantry cranes are collected, including: The dynamic confidence weights of each type of data source in the multi-source data are calculated based on an inherent reliability factor, an environmental adaptability factor, and a cross-consistency factor through linear weighting, and the multi-source data in the same time slice is weighted and fused according to the confidence weights, to generate the continuous time sequence state vectors of the real-time position, motion speed, and working posture of each unmanned gantry crane, wherein the inherent reliability factor is a static evaluation value obtained in the deployment stage according to the basic measurement performance, noise level, and historical operation statistics of the sensor, the environmental adaptability factor is an environmental adaptability parameter dynamically determined according to real-time environmental perception data and in combination with a preset environmental-performance mapping relationship, and the cross-consistency factor is a consistency index calculated according to the weighted Euclidean distance of the current observation value of the data source in the same time slice and the observation value of the same type of data source, and the value of the consistency index is the reciprocal of the weighted Euclidean distance.

7. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 6, characterized in that, In step S3, the continuous time sequence state vectors of each unmanned gantry crane are used to perform rolling trajectory deduction, to generate the predicted trajectories of the trolley, the trolley, and the spreader in the future prediction time length, and to discretize them into a multi-dimensional space-time position point sequence, including: Based on the continuous time sequence state vectors of each unmanned gantry crane, a space-time trajectory sequence in a future preset prediction time length is generated for each unmanned gantry crane; The space-time trajectory sequence is discretized according to a unified time slice to obtain a discretized future path point set, and the multi-dimensional space-time position point sequence is constructed based on the future path point set to represent the spatial coordinates and working posture parameters of each unmanned gantry crane in each time slice in the future prediction time length.

8. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 6, characterized in that, In step S3, based on the multi-dimensional space-time position point sequence, three-dimensional safety envelope overlap detection is performed on any two unmanned gantry cranes in the future time period, and a future potential space-time conflict event is generated when overlap is detected, including: Based on the multi-dimensional space-time position point sequence, a three-dimensional safety envelope body covering the future prediction time length is constructed for each unmanned gantry crane, wherein the three-dimensional safety envelope body is a dynamic safety buffer zone formed according to the trolley boundary, the trolley transverse width, and the spreader lifting space of the unmanned gantry crane; In the future prediction time length, the three-dimensional safety envelope bodies corresponding to any two unmanned gantry cranes are subjected to spatial overlap detection in each time slice, wherein the spatial overlap detection includes performing intersection volume calculation based on the three-dimensional bounding box of the envelope body, and when the intersection volume is greater than zero, it is determined that there is envelope overlap in the corresponding time slice; When the three-dimensional safety envelope bodies overlap in any future time slice, the corresponding future potential space-time conflict event is generated, and the time slice index and the on-duty equipment number at which the future potential conflict event occurs are recorded. 9.The method of claim 7, wherein, In step S4, in response to the future potential spatio-temporal conflict event, a multi-objective cost function is constructed based on task delay cost, path detour cost and speed adjustment cost, the cost value of each preset avoidance action is calculated, the avoidance action with the minimum cost value is selected, the corresponding avoidance control parameter is generated, and the avoidance control parameter generates a temporary exclusive virtual working corridor for the unmanned gantry crane in the virtual berth coordinate system, including: After detecting the future potential conflict event, the predicted trajectory of the unmanned gantry crane involved in the future potential conflict event and the corresponding conflict time slice are extracted; For each preset avoidance action, the adjustment amount of the avoidance action in the conflict time slice is calculated based on the predicted trajectory, and the avoidance action includes a speed adjustment action, a path detour action and a first-pass action; wherein, The adjustment amount of the speed adjustment action is the minimum speed change amount of the target unmanned gantry crane passing through the conflict space before and after the conflict time slice; the adjustment amount of the path detour action is the minimum path offset amount of the target unmanned gantry crane not entering the conflict envelope in the conflict time slice; and the adjustment amount of the first-pass action is the minimum time delay amount of the target unmanned gantry crane passing through again after the conflict envelope is released; The corresponding task delay cost, path detour cost and speed adjustment cost are calculated based on the adjustment amount, the multi-objective cost function is weighted and summed according to the preset weight coefficient, the corresponding cost value is obtained, the avoidance action with the minimum cost value is selected as the target avoidance action, and the corresponding avoidance control parameter is obtained based on the target avoidance action; The target berth interval corresponding to the avoidance control parameter is determined in the berth coordinate system, and the virtual working corridor is constructed in the target berth interval; wherein, the virtual working corridor is a temporary exclusive running area that only allows the target unmanned gantry crane to enter within a preset duration; the virtual working corridor is marked as a temporary exclusive running constraint of the target unmanned gantry crane, and the virtual working corridor and the corresponding temporary exclusive running constraint are automatically released after the target unmanned gantry crane crosses the target berth interval.

10. The unmanned gantry crane coordination scheduling method for the smart land port logistics hub according to claim 9, characterized in that, In step S4, a plurality of unmanned gantry cranes are controlled in real time and the final control instruction is issued, including: Based on the temporary exclusive running constraint of the virtual working corridor, each unmanned gantry crane is dynamically assigned a corresponding executable working instruction to ensure that the working path and the working target berth do not invade the virtual working corridor and the safety limiting area of any unmanned gantry crane; The instruction execution state and the equipment running state of each unmanned gantry crane are monitored in real time, and when a device fault, a working timeout or a new spatio-temporal conflict event is detected, an instruction rescheduling and conflict resolution process is triggered immediately; According to the dependency relationship of the sub-instruction sequence, the inner container truck and the reach stacker are cooperatively scheduled to ensure that the inner container truck or the reach stacker can arrive at the designated position in time for the connection operation after the completion of the operation of the crane; Based on the real-time cooperative scheduling control, the control instructions including the target position, the operation speed and the operation posture parameters of the work are issued to the corresponding unmanned gantry crane, the inner container truck and the reach stacker.

Images