CN-122022688-A - Modularized management method, system, equipment and product based on intelligent warehouse in-out warehouse-in warehouse

CN122022688ACN 122022688 ACN122022688 ACN 122022688ACN-122022688-A

Abstract

The invention discloses a modularized management method, system, equipment and product based on intelligent warehouse entry and exit, and relates to the technical field of intelligent warehouse entry and emergency logistics. According to the method, disaster text is analyzed through a natural language processing model to be a disaster situation vector, the vector is mapped to at least one fire-fighting function module identifier based on a mapping rule, an initial fire-fighting equipment set is determined according to the identifier, an equipment association diagram taking fire-fighting equipment as a node and a weighted value of functional association degree and goods position proximity as side weight is combined, a complete fire-fighting equipment set with optimal internal functions and space synthesis is generated through a community finding algorithm, then a simulated library task is generated in a storage digital twin model based on equipment real-time goods positions, a multi-agent reinforcement learning algorithm is utilized to plan a global conflict-free collaborative goods taking instruction set for a plurality of AGVs, and finally simulation verification is performed by issuing instructions to an AGV controller, so that fire-fighting emergency response efficiency can be greatly improved.

Inventors

Qian Wentian
SHEN YUCHEN
WANG JIE

Assignees

安徽中科数智信息科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. The modularized management method based on the intelligent warehouse in-out warehouse is characterized in that the modularized management method is executed by a management server of each automatic guided vehicle AGV in the intelligent warehouse in-out warehouse in a communication connection mode, and comprises the following steps: acquiring disaster text information of an emergency fire command; invoking a pre-trained natural language processing model to analyze the disaster text information into a structured disaster situation vector, wherein the disaster situation vector comprises at least two of a disaster type, a height of a position where the disaster is located, a personnel state of the position where the disaster is located and a dangerous object type of the position where the disaster is located; Mapping the disaster situation vector into a unique identifier of at least one fire-fighting function module based on a pre-constructed situation and function mapping rule, wherein the situation and function mapping rule is used for defining the corresponding relation between different disaster situation vectors and a plurality of preset fire-fighting function modules, and each fire-fighting function module is associated with at least one necessary fire-fighting equipment and at least one selectable fire-fighting equipment; Determining an initial set of fire fighting equipment for responding to the emergency fire command based on the unique identification of the at least one fire fighting function module; Generating a complete fire-fighting equipment set for responding to the emergency fire-fighting command through a community discovery algorithm based on the initial fire-fighting equipment set and a pre-constructed fire-fighting equipment association diagram, wherein the fire-fighting equipment association diagram takes all fire-fighting equipment in a warehouse as nodes, and takes a weighted value of the functional co-occurrence association degree and the physical cargo space proximity degree between any two fire-fighting equipment as a weight for connecting edges of two nodes corresponding to the any two fire-fighting equipment one by one, and the optimization objective of the community discovery algorithm is to comprehensively optimize the functional association degree and the storage position proximity degree inside the generated equipment set; Triggering and generating a simulated ex-warehouse task comprising a plurality of goods taking subtasks in a warehouse-in and warehouse-out storehouse digital twin model based on real-time goods positions of all fire-fighting equipment in the complete fire-fighting equipment set; planning a global conflict-free collaborative delivery instruction set for a plurality of AGVs responsible for executing the simulated delivery task based on a multi-agent reinforcement learning algorithm, wherein the collaborative delivery instruction set comprises motion control instructions independently generated for each AGV in the plurality of AGVs; And carrying out simulation verification on the planned collaborative goods taking instruction set, and after verification is passed, issuing each motion control instruction in the collaborative goods taking instruction set to a corresponding AGV controller for execution.
2. The modular management method of claim 1, wherein determining an initial set of fire equipment for responding to the emergency fire command based on the unique identification of the at least one fire function module comprises: searching all the necessary fire-fighting equipment respectively associated with the at least one fire-fighting function module in a warehouse according to the unique identification of the at least one fire-fighting function module, and taking the searching result as an initial fire-fighting equipment set for responding to the emergency fire-fighting command; For each fire-fighting function module in the at least one fire-fighting function module, if all the selectable fire-fighting equipment associated with the corresponding module can be found in the warehouse according to the corresponding unique identifier, the model and the number of the specific equipment are selected from all the selectable fire-fighting equipment according to at least one specific quantization parameter in the disaster situation vector and a predefined equipment adaptation rule, and the selected result is added into the initial fire-fighting equipment set, wherein the specific quantization parameter comprises at least one of a height value, estimated trapped people and dangerous chemical types at the position of the disaster, and the equipment adaptation rule is used for defining the mapping relation between different quantization parameter ranges and the model and the number of the selectable fire-fighting equipment.
3. The modular management method of claim 2, wherein generating, by a community discovery algorithm, a complete set of fire fighting equipment for responding to the emergency fire command based on the initial set of fire fighting equipment and a pre-constructed fire fighting equipment association graph, comprises: Running a community finding algorithm on a pre-constructed fire-fighting equipment association graph by taking all equipment in the initial fire-fighting equipment set as target nodes respectively to find a connected subgraph which comprises all the target nodes and maximizes the ratio of the sum of weights of edges in the subgraph to the sum of weights of all possible edges in the subgraph, wherein the fire-fighting equipment association graph takes all fire-fighting equipment in a warehouse as nodes, and takes the weighted value of the functional co-occurrence association degree and the physical cargo space proximity between any two fire-fighting equipment as the weight for connecting the edges of two nodes corresponding to the any two fire-fighting equipment one by one, and the optimization target of the community finding algorithm is to comprehensively optimize the functional association and the storage proximity in the generated equipment set; And adding the fire-fighting equipment corresponding to the non-target node contained in the searched connected subgraph into the initial fire-fighting equipment set to finally form a complete fire-fighting equipment set for responding to the emergency fire-fighting command.
4. The modular management method of claim 1, wherein planning a global collision-free collaborative pick instruction set for a plurality of AGVs responsible for performing the simulated delivery tasks based on a multi-agent reinforcement learning algorithm comprises: Taking the storage digital twin model as a simulation environment, and respectively modeling each AGV in the intelligent storage warehouse-in and warehouse-out as an intelligent body; constructing a multi-agent deep reinforcement learning model, wherein the multi-agent deep reinforcement learning model comprises a centralized value commentator network and a distributed strategy executor network which corresponds to each agent one by one; The multi-agent deep reinforcement learning model is trained offline and fine-tuned online by a composite rewarding function for guiding agent cooperation, using an actor-comment family training framework, in which, during training, the centralized value comment family network is used for evaluating the value of the global state according to the joint state and global map information of all the agents, and guiding each distributed strategy executor network to optimize each executing strategy so as to maximize the expectation of long-term accumulated rewards, and the composite rewarding function is obtained by weighting and summing the following formulas: In the formula, A serial number representing the agent in question, The time step is indicated as such, Represent the first The individual agents step in time Is a composite prize of (1), Represents the first The individual agents are at the time step Is a task rewarding item of (1), and if the first The individual agents are at the time step Successfully reaches the corresponding target goods position and completes the virtual picking action, the task rewarding item obtains a forward fixed value rewarding, otherwise, the task rewarding item is zero, Represents the first The individual agents are at the time step And at said time step If the first is Overlapping the bounding box of each agent with the bounding box of any other agent or static obstacle in the simulation scene, the collision penalty term is awarded a negative fixed value, otherwise the collision penalty term is zero, Represents the first The individual agents are at the time step Is a synergistic efficiency bonus item of (c), and at the time step Counting the passing behaviors of all the agents at the shared resource point, if the first agent The intelligent agent forms alternating passing or following actions without passing conflict with other intelligent agents in a preset competition area, so that the cooperative efficiency rewarding item obtains a forward and non-fixed value rewarding which is dynamically calculated according to the smoothness degree of the local traffic flow, otherwise, the cooperative efficiency rewarding item is zero, 、 And Respectively represent preset positive weight coefficients and have ; Aiming at each AGV, the corresponding distributed strategy executor network independently generates a corresponding motion control instruction according to the real-time local observation information of the corresponding AGV after training; and integrating the motion control instructions of the AGVs to form a global conflict-free collaborative goods taking instruction set planned by the AGVs for executing the simulated delivery task based on the multi-agent reinforcement learning algorithm.
5. The modular management method of claim 1, wherein performing simulation verification of the planned collaborative fetch instruction set comprises: In the storage digital twin model, carrying out acceleration simulation of a full task flow according to the planned collaborative goods taking instruction set and a preset AGV motion dynamics model; if deadlock, collision or task timeout is predicted in the simulation process, judging that the verification is not passed, otherwise judging that the verification is passed.
6. The modular management method of claim 1, wherein after issuing each motion control command to a respective AGV controller for execution, the method further comprises: In the process of executing a goods taking task by an entity AGV, verifying whether the electric quantity, the pressure value or the last maintenance time of the target firefighting equipment is in an available threshold range or not by reading an RFID tag on the picked target firefighting equipment or an Internet of things sensor carried on the picked target firefighting equipment in communication connection; If the verification is not passed, other fire-fighting equipment is selected nearby according to the equipment association diagram to replace the target fire-fighting equipment, and the goods taking path of the affected AGV is re-planned in real time.
7. The modular management system based on the intelligent warehouse in-out warehouse is characterized by being suitable for being arranged in a management server of each AGV in the intelligent warehouse in-out warehouse in a communication connection mode, and comprising a disaster text acquisition unit, a text analysis processing unit, a functional module mapping unit, an equipment set initial unit, an equipment set final unit, a simulation task generation unit, a collaborative goods taking planning unit and an instruction verification issuing unit which are sequentially in communication connection; the disaster text acquisition unit is used for acquiring disaster text information of the emergency fire control command; The text analysis processing unit is used for calling a pre-trained natural language processing model and analyzing the disaster text information into a structured disaster situation vector, wherein the disaster situation vector comprises at least two of a disaster type, a disaster position height, a disaster position personnel state and a disaster position dangerous object type; The function module mapping unit is used for mapping the disaster situation vector into a unique identifier of at least one fire-fighting function module based on a pre-constructed situation and function mapping rule, wherein the situation and function mapping rule is used for defining the corresponding relation between different disaster situation vectors and a plurality of preset fire-fighting function modules, and each fire-fighting function module is associated with at least one necessary fire-fighting equipment and at least one selectable fire-fighting equipment; the equipment set initial setting unit is used for determining an initial fire-fighting equipment set for responding to the emergency fire-fighting command according to the unique identification of the at least one fire-fighting functional module; The equipment set finalizing unit is used for generating a complete fire-fighting equipment set for responding to the emergency fire-fighting command through a community finding algorithm based on the initial fire-fighting equipment set and a pre-constructed fire-fighting equipment association diagram, wherein the fire-fighting equipment association diagram takes all fire-fighting equipment in a warehouse as nodes, and takes a weighted value of the functional co-occurrence association degree and the physical goods space proximity between any two fire-fighting equipment as a weight for connecting edges of two nodes corresponding to the any two fire-fighting equipment one by one, and the optimization objective of the community finding algorithm is to comprehensively optimize the functional association and the storage proximity inside the generated equipment set; The simulation task generating unit is used for triggering and generating a simulation ex-warehouse task comprising a plurality of goods taking sub-tasks in a warehouse digital twin model of the intelligent warehouse ex-warehouse based on real-time goods positions of all fire-fighting equipment in the complete fire-fighting equipment set; The collaborative delivery planning unit is configured to plan a global collision-free collaborative delivery instruction set for a plurality of AGVs responsible for executing the simulated delivery task based on a multi-agent reinforcement learning algorithm, where the collaborative delivery instruction set includes motion control instructions that are independently generated for each of the plurality of AGVs; The command verification issuing unit is used for carrying out simulation verification on the planned collaborative goods taking command set, and issuing each motion control command in the collaborative goods taking command set to a corresponding AGV controller for execution after verification is passed.
8. A computer device, comprising a storage module, a processing module and a transceiver module, which are connected in turn in communication, wherein the storage module is used for storing a computer program, the transceiver module is used for receiving and transmitting a message, and the processing module is used for reading the computer program and executing the modularized management method according to any one of claims 1-6.
9. A computer readable storage product having instructions stored thereon which, when executed on a computer, perform the modular management method of any of claims 1-6.
10. A computer program product comprising a computer program or instructions which, when executed by a computer, implement the modular management method of any of claims 1 to 6.

Description

Modularized management method, system, equipment and product based on intelligent warehouse in-out warehouse-in warehouse Technical Field The invention belongs to the technical field of intelligent storage and emergency logistics, and particularly relates to a modularized management method, system, equipment and product based on intelligent storage in-out storage. Background The success of emergency rescue of disaster accidents such as fire is highly dependent on the speed of initial response and the scientificity of resource scheduling. Fire-fighting equipment warehouse is used as a physical hub of emergency resources, and the intelligent level of the management mode directly determines the starting efficiency of rescue force. At present, fire equipment warehouse management mainly goes through the following development stages: The manual management mode is that the equipment management and allocation are carried out in early stage by completely relying on paper account and manual memory, namely, the equipment is required to be searched and checked one by one in site when the equipment is delivered out of the warehouse, the efficiency is low, the key equipment is easy to make mistakes or miss in emergency, and the real-time state (such as electric quantity and pressure) of the equipment cannot be effectively monitored; The information management mode is characterized in that along with the application of technologies such as bar codes, RFID (Radio Frequency Identification ) and the like, the digital input and tracking of equipment information are realized, and the inventory and the position of the equipment can be quickly checked through a warehouse management system (Warehouse MANAGEMENT SYSTEM, WMS); Automated warehouse mode in recent years, some advanced fire-fighting warehouses begin to introduce hardware devices such as automated stereoscopic warehouse and automated guided vehicles (Automated Guided Vehicle, AGVs), for example, public data displays, the fire-fighting equipment warehouse for realizing automatic picking from goods to people by AGVs has been built by the mart, related technical schemes (such as CN 121329287A) also propose warehouse management methods combining path planning and AR display, which significantly improve the automation level of material handling, but do not solve the core intelligent decision problem of emergency dispatch, the operation logic is usually to execute a preset and fixed warehouse-out instruction, or perform simple partition picking according to the types, so the core limitation is that the system cannot understand why the warehouse-out is not possible, and thus the optimal warehouse-out scheme cannot be actively generated. Specifically, the prior art scheme has the following bottlenecks to be solved: (1) The method has the advantages that response is stiff, the disaster situation cannot be dynamically adapted, namely, the minimum unit of system scheduling is a single device or a pre-packaged fixed material box, when the disaster situation with complex and changeable conditions is faced, equipment bags with complete functions and accurate configuration cannot be dynamically combined according to specific characteristics of the disaster (such as building height, whether dangerous chemicals exist and/or the scale of trapped personnel and the like); (2) The scheduling and path planning are disjoint, namely, the equipment scheduling scheme (what to take) and the AGV execution scheme (how to take) are two cutting links, the existing path planning mostly aims at single AGV efficiency or simple task queues, and the capability of carrying out global collaborative goods taking path planning for a dynamically generated and unprecedented equipment combination under an emergency multi-task concurrency scene is lacking, so that traffic conflict and bottleneck are easily generated in a warehouse, and the optimization of the overall delivery efficiency of the system cannot be realized; (3) The system lacks closed loop verification capability, namely the capability of carrying out quick simulation and conflict previewing on the whole scheme (particularly a plurality of AGV cooperative paths) in a virtual environment before the scheme is executed, and in addition, if a certain equipment fault is found in the executing process, the system is difficult to quickly and automatically start a standby scheme, so that the fault tolerance is poor. Therefore, an integrated solution capable of understanding disaster conditions, intelligent assembly, collaborative planning and closed-loop verification is urgently needed in the field of fire-fighting intelligent storage so as to break through the bottleneck from 'automatic execution' to 'intelligent decision'. Disclosure of Invention The invention aims to provide a modularized management method, a modularized management system, computer equipment, computer readable storage products and computer program products based on intelligent warehouse e