An intelligent sorting scheduling method and system for multimodal transport scenarios

Abstract

The application discloses a kind of intelligent sorting scheduling method and system for multimodal transport scene, it is related to warehousing technical field, the method includes: the transport mode identifier of real-time acquisition multiple transport vehicles, the physical attribute information of corresponding goods and the running state information of stereoscopic sorting system;Multi-modal transport rigid rule information is parsed, and the dynamic boundary condition of multi-modal transport is generated;Transport mode identifier, physical attribute information, running state information and dynamic boundary condition are imported into deep reinforcement learning model;Output according to the goods gravity center parameter dynamic allocation warehousing level goods shelf allocation strategy, and the equipment task sequence of control AGV handling robot motion;Automatic generation sorting execution instruction, to correspond driving stereoscopic sorting system is sorted goods.The application forms "safe storage-time limit coupling-space coordination" technical closed loop, significantly reduces the comprehensive logistics cost of multimodal transport whole link.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of warehousing, in particular to an intelligent sorting and scheduling method and system for a multimodal transport scenario. BACKGROUND

[0002] With the acceleration of global supply chain integration, multimodal transport (railway, highway, and water transport coordination) has become the core mode of bulk logistics.

[0003] According to relevant statistics, medium and large goods face severe challenges in the sorting link due to their special and high weight characteristics. Traditional flat sorting schemes cannot meet the time efficiency coupling requirements of the composite transport scenario, such as the hard connection between railway trains and port tidal windows, resulting in low operational efficiency of hub nodes and rising logistics costs.

[0004] Therefore, the prior art also has the problem of low logistics transfer efficiency in the transfer scenario of the multimodal transport scenario, resulting in high logistics costs. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application aims to provide an intelligent sorting and scheduling method and system for a multimodal transport scenario, which aims to solve the above-mentioned problems described in the prior art.

[0006] The first aspect of the present application provides an intelligent sorting and scheduling method for a multimodal transport scenario, the method comprising:

[0007] real-time collection of transportation mode identifiers of multiple transport vehicles, physical attribute information of corresponding goods, and running state information of a three-dimensional sorting system, wherein the physical attribute information at least includes the gravity center parameter of the goods;

[0008] obtaining pre-set multimodal transport hard rule information, analyzing the multimodal transport hard rule information, and generating dynamic boundary conditions for multimodal transport;

[0009] importing the transportation mode identifier, the physical attribute information, the running state information, and the dynamic boundary condition into a pre-trained deep reinforcement learning model;

[0010] data processing by the deep reinforcement learning model, outputting a goods shelf allocation strategy for dynamically allocating a warehouse level according to the gravity center parameter of the goods, and a device task sequence for controlling the movement of an AGV handling robot;

[0011] generating sorting execution instructions automatically according to the goods shelf allocation strategy and the device task sequence to drive the three-dimensional sorting system to sort goods.

[0012] According to an aspect of the above technical solution, the pre-set multimodal transport hard rule information is obtained, the multimodal transport hard rule information is parsed, and the step of generating the dynamic boundary condition of multimodal transport includes:

[0013] The pre-constructed rule database is called to extract the pre-set multimodal transport hard rule information.

[0014] The multimodal transport hard rule is parsed according to semantics to generate at least one dynamic boundary condition for multimodal transport.

[0015] According to an aspect of the above technical solution, in the step of parsing the multimodal transport hard rule according to semantics to generate at least one dynamic boundary condition for multimodal transport, the dynamic boundary condition includes:

[0016] The arrival and departure time limit constraint boundary of the railway train arrival and departure time;

[0017] And the time limit constraint boundary of the port tidal cycle corresponding to the port operation.

[0018] According to an aspect of the above technical solution, the step of processing data through the deep reinforcement learning model, outputting a storage level rack allocation strategy dynamically allocated according to the center of gravity parameters of the goods, and controlling the equipment task sequence of the AGV carrying robot moving includes:

[0019] The deep reinforcement learning model is used for data processing, and the target storage level corresponding to the target goods is determined according to the physical attribute information of the target goods.

[0020] According to the target storage level corresponding to the target goods, the equipment task sequence of the AGV carrying robot moving corresponding to the target storage level is determined.

[0021] According to an aspect of the above technical solution, the step of processing data through the deep reinforcement learning model, outputting a storage level rack allocation strategy dynamically allocated according to the center of gravity parameters of the goods, and controlling the equipment task sequence of the AGV carrying robot moving includes:

[0022] According to the physical attribute information of the target goods, the weight and center of gravity parameters corresponding to the target goods are extracted.

[0023] According to the center of gravity parameters corresponding to the target goods, the space vertical component and horizontal eccentricity rate in the center of gravity parameters are extracted.

[0024] According to the rack structure configuration information of the rack, the space vertical component and horizontal eccentricity rate corresponding to the target goods, a plurality of storage levels corresponding to the target goods are determined and scored to determine a target storage level.

[0025] According to an aspect of the above technical solution, the step of determining the equipment task sequence for controlling the AGV carrying robot to perform corresponding movement according to the target storage level corresponding to the target goods comprises:

[0026] According to the target storage level corresponding to the target goods, a four-dimensional task sequence is generated in the three-dimensional space topology of the stereoscopic sorting system.

[0027] A safety time window is added to each sorting task in the four-dimensional task sequence to determine the equipment task sequence for controlling the AGV carrying robot to perform corresponding movement.

[0028] According to an aspect of the above technical solution, the plurality of transportation carriers at least includes a truck, a railway freight train, and a shipping freighter.

[0029] The second aspect of the present application provides an intelligent sorting scheduling system for a multimodal transport scenario, which is applied to the method in the above technical solution, and the system comprises:

[0030] A data acquisition module is configured to collect the transportation mode identifier of the plurality of transportation carriers, the physical attribute information of the corresponding goods, and the operation state information of the stereoscopic sorting system in real time, wherein the physical attribute information at least includes the gravity center parameter of the goods.

[0031] A rule analysis module is configured to obtain pre-set multimodal transport hard rule information, analyze the multimodal transport hard rule information, and generate dynamic boundary conditions for multimodal transport.

[0032] A data import module is configured to import the transportation mode identifier, the physical attribute information, the operation state information, and the dynamic boundary conditions into a pre-trained deep reinforcement learning model.

[0033] A reinforcement learning module is configured to perform data processing through the deep reinforcement learning model, output a goods shelf allocation strategy for dynamically allocating storage levels according to the gravity center parameter of the goods, and output an equipment task sequence for controlling the AGV carrying robot to perform movement.

[0034] An instruction generation module is configured to automatically generate sorting execution instructions according to the goods shelf allocation strategy and the equipment task sequence, so as to drive the stereoscopic sorting system to sort goods.

[0035] The third aspect of the present application provides a readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method in the above technical solution.

[0036] The fourth aspect of the present application provides an electronic device, comprising a storage, a processor and a computer program stored in the storage and executable on the processor, wherein the processor implements the method described in the above technical solution when executing the computer program.

[0037] Compared with the prior art, the intelligent sorting and scheduling method and system for the multimodal transport scene have the beneficial effects that:

[0038] The present application acquires the preset multimodal transport hard rule information, analyzes the multimodal transport hard rule information, and generates dynamic boundary conditions of multimodal transport; the transport mode identifier, the physical attribute information, the running state information and the dynamic boundary conditions are imported into a pre-trained deep reinforcement learning model; the deep reinforcement learning model is used for data processing, and a rack allocation strategy for dynamically allocating storage levels according to the gravity center parameters of the goods and a device task sequence for controlling the movement of the AGV carrying robot are output; according to the rack allocation strategy and the device task sequence, a sorting execution instruction is automatically generated to drive the stereoscopic sorting system to sort goods. The present application forms a "safe storage-time coupling-space coordination" technical closed loop, which not only eliminates the inherent defects of manual scheduling and rigid automation systems, but also reconstructs the logistics hub operation mode with high flexibility and intelligent decision-making, and finally significantly reduces the comprehensive logistics cost of the multimodal transport whole link. BRIEF DESCRIPTION OF DRAWINGS

[0039] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0040] Figure 1 A flowchart of the intelligent sorting and scheduling method for the multimodal transport scene in an embodiment of the present application;

[0041] Figure 2 A structural block diagram of the intelligent sorting and scheduling system for the multimodal transport scene in an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the objectives, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. However, the present application can be realized in many different forms, and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0043] It is to be understood that where an element such as a layer, region or substrate is described as being "on" another element, it can be directly on the other element or intervening elements can also be present. Where an element is described as being "connected" or "coupled" to another element, it can be directly connected or coupled or intervening elements can be present. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0045] Embodiment one

[0046] Please refer to Figure 1 The first embodiment of the present application provides an intelligent sorting and scheduling method for a multimodal transport scenario, the method comprising steps S10-S50:

[0047] Step S10, real-time collection of transportation mode identifiers of various transport vehicles, physical attribute information corresponding to the goods, and running state information of the stereoscopic sorting system, wherein the physical attribute information at least includes the gravity center parameter of the goods.

[0048] First of all, it needs to be pointed out that the purpose of the method shown in this embodiment is to intelligently sort the goods involved in the multimodal transport scenario, which involves sorting and scheduling the goods, thereby improving the efficiency and quality of the goods transfer and storage.

[0049] In this embodiment, when intelligently sorting and scheduling, it is necessary to first collect the transportation mode identifiers of various transport vehicles, the physical attribute information corresponding to the goods, and the current running state information of the stereoscopic sorting system, wherein the transportation mode identifier is used to represent the transportation channel of the goods, such as the transportation channel or mode of the truck, the train and the ship, the physical attribute information is used to represent the quantity, volume, gravity center and other parameters of the goods corresponding to a transport vehicle, and the running state information is used to represent the use of the stereoscopic sorting system at the current time node, so as to effectively cope with the complex connection, time limit and requirement and package size difference of the truck, train and ship in the multimodal transport mode at the storage end, especially for large goods. This embodiment realizes seamless time limit coupling of truck, train and ship by deconstructing the hard rules such as railway train time window and port tidal cycle and converting them into algorithm boundary conditions, effectively breaking through the multimodal transport connection bottleneck.

[0050] In step S20, the pre-set hard rule information of multimodal transport is acquired, the hard rule information of multimodal transport is parsed, and the dynamic boundary condition of multimodal transport is generated.

[0051] In the embodiment, after the transport mode identifier of the plurality of transport vehicles, the physical attribute information of the goods, and the running state information of the stereoscopic sorting system are collected in real time, the pre-set hard rule information of multimodal transport is acquired, the hard rule information of multimodal transport is parsed, and the dynamic boundary condition corresponding to the multimodal transport is generated.

[0052] In the embodiment, the step of acquiring the pre-set hard rule information of multimodal transport, parsing the hard rule information of multimodal transport, and generating the dynamic boundary condition of multimodal transport includes:

[0053] The pre-constructed rule database is called to extract the pre-set hard rule information of multimodal transport.

[0054] The hard rule of multimodal transport is parsed according to semantics, and at least one dynamic boundary condition for multimodal transport is generated.

[0055] In the step of parsing the hard rule of multimodal transport according to semantics and generating at least one dynamic boundary condition for multimodal transport, the dynamic boundary condition includes:

[0056] The arrival and departure time limit constraint boundary of the railway train;

[0057] And the time limit constraint boundary of the port tidal cycle corresponding to the wharf operation.

[0058] Specifically, first, the pre-constructed rule database in which a plurality of hard rules of multimodal transport are stored is called, and then the corresponding hard rule data of multimodal transport is extracted, each hard rule information of multimodal transport is parsed according to semantics, and at least one dynamic boundary condition for multimodal transport is generated. The dynamic boundary condition at least includes the arrival and departure time limit constraint boundary of the railway train in the form of land transportation, and the time limit constraint boundary of the port tidal cycle corresponding to the wharf operation, and of course can also have the constraint boundary of the highway gate capacity. In the embodiment, the physical coupling rules of the railway train time window, the port tidal cycle, and the highway gate capacity are parsed to construct the dynamic constraint boundary condition, so as to parse the hard time lock of the railway train time window, extract the wharf operation buffer parameter of the port tidal cycle, and quantify the dynamic throughput threshold of the highway gate capacity.

[0059] In step S30, the transport mode identifier, the physical attribute information, the running state information, and the dynamic boundary condition are imported into the pre-trained deep reinforcement learning model.

[0060] In the embodiment, after the transportation mode identifier, the physical attribute information of the goods, the operation state information of the stereoscopic sorting system, and the dynamic boundary condition are collected and generated, the transportation mode identifier, the physical attribute information, the operation state information, and the dynamic boundary condition are input into a deep reinforcement learning model as input parameters of the model, and then data processing is performed through preset rules of the deep reinforcement learning model, and finally corresponding data output is performed.

[0061] In step S40, data processing is performed through the deep reinforcement learning model, and a goods shelf allocation strategy dynamically allocated according to a goods gravity center parameter is output, and a device task sequence for controlling the movement of an AGV handling robot is output.

[0062] First of all, it needs to be pointed out that in the embodiment, the deep reinforcement learning model includes a first attention branch and a second attention branch, and the first attention branch and the second attention branch are fused through a graph attention fusion layer. The first attention branch is used for analyzing the physical attributes of the goods, and a feature vector is extracted through a 3D convolutional neural network, and the second attention branch is used for analyzing the dynamic boundary conditions of multimodal transport, such as railway time window and tidal operation cycle, to generate time-sensitive constraint coding.

[0063] Specifically, when outputting the shelf allocation strategy, the storage level of the goods is determined according to the gravity center parameter of the goods, for example, the lower the gravity center of the goods, the lower the storage level required, and vice versa. A stable inclination angle for placing the goods on the shelf is generated in advance to ensure storage safety, and the volume and weight of the goods can also be used to determine the inclination angle. When generating the device task sequence of the AGV handling robot, the storage level of the goods needs to be determined, and the type of transfer equipment needs to be planned, and the movement path from the carrier to the stereoscopic warehouse needs to be planned. The embodiment dynamically optimizes the storage level and the shelf posture based on the gravity center parameter of the goods through the deep reinforcement learning model, and essentially solves the safety risk and storage efficiency problem of traditional schemes in the sorting of large and irregular goods.

[0064] In step S50, a sorting execution instruction is automatically generated according to the shelf allocation strategy and the device task sequence, so as to drive the stereoscopic sorting system to sort the goods.

[0065] In the embodiment, after the shelf allocation strategy and the corresponding device task sequence are output by the deep reinforcement learning model, a sorting execution instruction for the stereoscopic sorting system is generated according to the shelf allocation strategy and the device task sequence, so as to drive the stereoscopic sorting system to sort the goods, thereby improving the sorting efficiency of the goods in the multimodal transport scenario and the transfer efficiency of the goods, thereby effectively reducing the logistics cost.

[0066] More specifically, the embodiment utilizes four-dimensional space-time path planning combined with a dynamic safety buffer mechanism to construct a high-precision collaborative network in a three-dimensional operation space, thereby ensuring the safety of equipment operation in complex environments and minimizing unnecessary energy consumption.

[0067] Compared with the prior art, the intelligent sorting and scheduling method for the multimodal transport scene has the beneficial effects that:

[0068] The embodiment collects the transportation mode identifier of various transport vehicles, the physical attribute information of the goods, and the running state information of the three-dimensional sorting system in real time, wherein the physical attribute information at least includes the gravity center parameter of the goods; obtains pre-set multimodal transport hard rule information, analyzes the multimodal transport hard rule information, and generates dynamic boundary conditions for multimodal transport; imports the transportation mode identifier, the physical attribute information, the running state information, and the dynamic boundary conditions into a pre-trained deep reinforcement learning model; performs data processing through the deep reinforcement learning model, outputs a goods shelf allocation strategy for dynamically allocating a warehouse level according to the gravity center parameter of the goods, and a device task sequence for controlling the movement of the AGV handling robot; and generates sorting execution instructions automatically according to the goods shelf allocation strategy and the device task sequence to drive the three-dimensional sorting system to sort goods. The embodiment forms a "safe storage-time coupling-space collaboration" technical closed loop, which not only eliminates the inherent defects of manual scheduling and rigid automation systems, but also reconstructs the logistics hub operation mode with high flexibility and intelligent decision-making, thereby significantly reducing the comprehensive logistics cost of the multimodal transport whole link.

[0069] Embodiment Two

[0070] The second embodiment of the application also provides an intelligent sorting and scheduling method for a multimodal transport scene. The method shown in the embodiment is basically similar to the method shown in the first embodiment, and the difference lies in that:

[0071] In the embodiment, the step of performing data processing through the deep reinforcement learning model to output a goods shelf allocation strategy for dynamically allocating a warehouse level according to the gravity center parameter of the goods and a device task sequence for controlling the movement of the AGV handling robot includes:

[0072] According to the physical attribute information of the target goods, determining the target warehouse level corresponding to the target goods through the deep reinforcement learning model;

[0073] According to the target warehouse level corresponding to the target goods, determining the device task sequence for controlling the corresponding movement of the AGV handling robot.

[0074] Specifically, in the data processing by the deep reinforcement learning model, first, according to the physical attribute information of the target goods, such as the quantity, volume, weight and center of gravity parameters of the target goods, the target storage level corresponding to the target goods in the stereoscopic warehouse is determined, which is specifically that a plurality of storage levels corresponding to the target goods in the stereoscopic warehouse are determined in advance, then the plurality of storage levels, i.e. storage points, are evaluated to determine a target storage level, and then according to the target storage level corresponding to the target goods, a device task sequence for controlling the AGV handling robot to move is determined to create a corresponding handling task and its corresponding movement path.

[0075] For example, when the goods have a high transfer priority, the corresponding AGV handling robot is first determined according to the type of the goods, and then the movement path of the AGV handling robot is generated correspondingly, so that it has a high priority for passing, which can effectively guarantee the transfer efficiency of special goods.

[0076] In the data processing by the deep reinforcement learning model, according to the physical attribute information of the target goods, the target storage level corresponding to the target goods is determined, which includes:

[0077] According to the physical attribute information of the target goods, the weight and center of gravity parameters of the target goods are extracted;

[0078] According to the center of gravity parameters corresponding to the target goods, the spatial vertical component and horizontal eccentricity rate in the center of gravity parameters are extracted;

[0079] According to the shelf structure configuration information of the shelf and the spatial vertical component and horizontal eccentricity rate corresponding to the target goods, a plurality of storage levels corresponding to the target goods are determined and scored to determine a target storage level.

[0080] Specifically, in determining the target storage level corresponding to the target goods, the weight and center of gravity parameters corresponding to the target goods are extracted according to the physical attribute information of the target goods, then the spatial vertical component and horizontal eccentricity rate corresponding to the center of gravity parameters are calculated according to the target center of gravity parameters, and finally according to the shelf structure configuration information of the shelf, the spatial vertical component and the horizontal eccentricity rate, a plurality of storage levels meeting the requirements are queried in the stereoscopic warehouse, the plurality of storage levels are scored, and a target storage level most suitable for placing the target goods is determined.

[0081] More specifically, the point cloud data corresponding to the target goods is acquired by a three-dimensional laser scanner, the weight distribution is measured in combination with a pressure sensor array, the spatial coordinates of the center of gravity are calculated, the spatial vertical component and the horizontal eccentricity rate are calculated according to the spatial coordinates of the center of gravity, and then a hierarchical scoring model is established according to the rack structure configuration information, which specifically includes the rated load of each layer of the rack, the spatial height margin, that is, the adjustable height range of each layer of the rack, and the support stiffness, that is, the tolerance coefficient of each layer of the rack to eccentric load. Through the above hierarchical scoring model, all levels of the three-dimensional warehouse are filtered and quantitatively scored according to the above rack configuration information, to determine a candidate set of levels preliminarily corresponding to the target goods. All candidate levels in the candidate set are sorted according to the score to determine the target storage level corresponding to the target goods.

[0082] The step of determining the device task sequence for controlling the AGV handling robot to move according to the target storage level corresponding to the target goods, comprises:

[0083] Generating a four-dimensional task sequence in the three-dimensional spatial topology of the three-dimensional sorting system according to the target storage level corresponding to the target goods;

[0084] Adding a safety time window to each sorting task in the four-dimensional task sequence to determine the device task sequence for controlling the AGV handling robot to move.

[0085] Specifically, in the embodiment, after determining the target storage level corresponding to the target goods, a four-dimensional task sequence is first generated in the three-dimensional spatial topology of the three-dimensional sorting system, that is, a three-dimensional spatial model, and then a corresponding safety time window is added to each sorting task in the four-dimensional task sequence to determine the device task sequence for controlling the AGV handling robot to move, thereby ensuring the transfer efficiency of the AGV handling robot when transferring the target goods.

[0086] More specifically, in the three-dimensional spatial model of the three-dimensional sorting system, a basic path point set is first generated, and a space-time state search algorithm is used to calculate the optimal physical path from the starting point of the goods to the target rack according to the position coordinates of the target storage level. The algorithm considers three-dimensional constraints such as rack height, channel width, and elevator position, and outputs an initial trajectory containing N path points. Subsequently, a time dimension binding operation is performed, and a time stamp accurate to milliseconds is added to each path point according to the transportation mode time requirement. In the specific calculation, the task start time t0 is taken as the reference, and the travel time and goods transfer time of each path segment are accumulated, for example, the AGV uniform speed in the railway scene is set to 2.2 m / s, and the shipping scene is set to 1.5 m / s.

[0087] Then, based on the real-time speed vector difference of adjacent AGVs, the maximum braking acceleration and the sensing delay, a dynamic safety threshold is calculated. By introducing a hierarchical safety coefficient to adjust the sensitivity, more buffer space is reserved in high-risk areas, ensuring the safety and efficiency of transfer.

[0088] Embodiment three

[0089] Referring to Figure 2 The third embodiment of the present application provides an intelligent sorting and scheduling system for a multimodal transport scenario, which is applied to the method of any of the above embodiments. The system comprises:

[0090] A data acquisition module 10 is configured to collect the transport mode identifiers of various transport vehicles, the physical attribute information of the goods, and the operating state information of the three-dimensional sorting system in real time, wherein the physical attribute information at least includes the gravity center parameter of the goods.

[0091] A rule analysis module 20 is configured to obtain pre-set multimodal transport hard rule information, analyze the multimodal transport hard rule information, and generate dynamic boundary conditions for multimodal transport.

[0092] A data import module 30 is configured to import the transport mode identifiers, the physical attribute information, the operating state information, and the dynamic boundary conditions into a pre-trained deep reinforcement learning model.

[0093] A reinforcement learning module 40 is configured to process data through the deep reinforcement learning model, output a goods shelf allocation strategy for dynamically allocating storage levels according to the gravity center parameter of the goods, and control the device task sequence of the AGV handling robot for movement.

[0094] An instruction generation module 50 is configured to automatically generate sorting execution instructions according to the goods shelf allocation strategy and the device task sequence, so as to drive the three-dimensional sorting system to sort goods.

[0095] Compared with the prior art, the intelligent sorting and scheduling system for a multimodal transport scenario has the following advantages:

[0096] The embodiment collects the transportation mode identifier of various transportation carriers, the physical attribute information of the goods, and the running state information of the stereoscopic sorting system in real time, wherein the physical attribute information at least includes the gravity center parameter of the goods; obtains the pre-set hard rule information of multimodal transport, analyzes the hard rule information of multimodal transport, and generates the dynamic boundary condition of multimodal transport; imports the transportation mode identifier, the physical attribute information, the running state information, and the dynamic boundary condition into the pre-trained deep reinforcement learning model; performs data processing through the deep reinforcement learning model, outputs the goods shelf allocation strategy of dynamically allocating the storage level according to the gravity center parameter of the goods, and controls the equipment task sequence of the movement of the AGV carrying robot; according to the goods shelf allocation strategy and the equipment task sequence, automatically generates sorting execution instructions to drive the stereoscopic sorting system to sort goods. The embodiment forms a "safe storage-time coupling-space cooperation" technical closed loop, not only eliminates the inherent defects of manual scheduling and rigid automatic system, but also reconstructs the logistics hub operation mode with high flexibility and intelligence, and finally significantly reduces the comprehensive logistics cost of the multimodal transport whole link.

[0097] Embodiment four

[0098] The fourth embodiment of the present application provides a readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the method described in any of the above embodiments.

[0099] Embodiment five

[0100] The fifth embodiment of the present application provides an electronic device, which includes a storage, a processor, and a computer program stored in the storage and executable on the processor, and the processor executes the computer program to realize the method described in any of the above embodiments.

[0101] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0102] The above embodiments only express several implementation manners of the present application, which are described in a more specific and detailed manner, but cannot be understood as a limitation on the patent scope of the present application. It should be noted that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. An intelligent sorting and dispatching method for a multimodal transport scenario, characterized in that, The method comprises: real-time collection of transportation mode identifiers of various transportation vehicles, physical attribute information of corresponding goods, and operation state information of a stereoscopic sorting system, wherein the physical attribute information at least includes a gravity center parameter of the goods; acquisition of pre-set hard rules of multimodal transport, analysis of the hard rules of multimodal transport, and generation of dynamic boundary conditions of multimodal transport; import of the transportation mode identifiers, the physical attribute information, the operation state information, and the dynamic boundary conditions into a pre-trained deep reinforcement learning model; data processing by the deep reinforcement learning model, output of a goods shelf allocation strategy for dynamic allocation of a warehouse level according to a gravity center parameter of goods, and determination of a device task sequence for movement of an AGV handling robot; automatic generation of sorting execution instructions according to the goods shelf allocation strategy and the device task sequence, so as to drive the stereoscopic sorting system to sort goods; wherein the step of acquiring pre-set hard rules of multimodal transport, analyzing the hard rules of multimodal transport, and generating dynamic boundary conditions of multimodal transport comprises: calling a pre-constructed rule database to extract pre-set hard rules of multimodal transport; parsing the hard rules of multimodal transport according to semantics to generate at least one dynamic boundary condition for multimodal transport, wherein the dynamic boundary condition includes a time limit constraint boundary of a railway train arrival and departure time and a time limit constraint boundary of a port tide cycle corresponding to a port operation; wherein the step of data processing by the deep reinforcement learning model, output of a goods shelf allocation strategy for dynamic allocation of a warehouse level according to a gravity center parameter of goods, and determination of a device task sequence for movement of an AGV handling robot comprises: determination of a target warehouse level corresponding to a target goods according to physical attribute information of the target goods through data processing by the deep reinforcement learning model; determination of a device task sequence for corresponding movement of an AGV handling robot according to the target warehouse level corresponding to the target goods.

2. The intelligent sorting dispatching method for a multimodal transport scenario according to claim 1, characterized in that, The step of determining a target warehouse level corresponding to a target goods according to physical attribute information of the target goods through data processing by the deep reinforcement learning model comprises: extraction of a goods weight and a gravity center parameter corresponding to the target goods according to the physical attribute information of the target goods; extraction of a spatial vertical component and a horizontal eccentricity rate in the gravity center parameter corresponding to the target goods according to the gravity center parameter corresponding to the target goods; determination of a plurality of warehouse levels corresponding to the target goods according to shelf structure configuration information of a shelf and the spatial vertical component and the horizontal eccentricity rate corresponding to the target goods, and scoring to determine a target warehouse level.

3. The intelligent sorting dispatching method for a multimodal transport scenario according to claim 2, characterized in that, The step of determining a device task sequence for corresponding movement of an AGV handling robot according to the target warehouse level corresponding to the target goods comprises: generation of a four-dimensional task sequence in a three-dimensional space topology of the stereoscopic sorting system according to the target warehouse level corresponding to the target goods; addition of a safety time window to each sorting task in the four-dimensional task sequence to determine a device task sequence for corresponding movement of an AGV handling robot.

4. The intelligent sorting and scheduling method for intermodal transportation scenarios according to any one of claims 1-3, characterized in that, The plurality of transportation carriers at least includes a truck, a railway freight train, and a shipping freighter.

5. An intelligent sorting dispatch system for a multimodal transport scenario, characterized by, The system is applied to the method of any one of claims 1-4, and the system comprises: a data acquisition module, configured to collect in real time a transportation mode identifier of the plurality of transportation carriers, physical attribute information of the goods, and operation state information of the stereoscopic sorting system, wherein the physical attribute information at least includes a gravity center parameter of the goods; a rule analysis module, configured to acquire pre-set hard multimodal transport rule information, analyze the hard multimodal transport rule information, and generate dynamic boundary conditions of multimodal transport; a data import module, configured to import the transportation mode identifier, the physical attribute information, the operation state information, and the dynamic boundary conditions into a pre-trained deep reinforcement learning model; a reinforcement learning module, configured to perform data processing through the deep reinforcement learning model, output a goods shelf allocation strategy for dynamically allocating a storage level according to a gravity center parameter of the goods, and control a device task sequence for movement of an AGV handling robot; an instruction generation module, configured to automatically generate sorting execution instructions according to the goods shelf allocation strategy and the device task sequence, so as to drive the stereoscopic sorting system to sort the goods.

6. A readable storage medium, having stored thereon a computer program, characterized in that, The computer program is executed by a processor to implement the method of any one of claims 1-4.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1-4.

Images