Multi-agent construction equipment scheduling method and system, and construction equipment

The multi-agent construction equipment scheduling method improves scheduling decisions by integrating central and distributed agents using large language models and reinforcement learning, addressing inefficiencies in existing methods.

CN119067398APending Publication Date: 2024-12-03SANY HEAVY IND CO LTD (CN) +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411204328.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing construction equipment scheduling methods suffer from poor scheduling effect, practicality, and real-time performance, particularly in construction scenarios, with simple logical judgment rules lacking effectiveness and multi-objective optimization algorithms failing to respond to emergencies.

Method used

A multi-agent construction equipment scheduling method involving a central integrated agent that gathers overall status data, task information, and actual equipment data to generate a target information group, which is used by target equipment agents to make scheduling decisions based on large language models and reinforcement learning algorithms.

Benefits of technology

Enhances scheduling effectiveness, practicality, and real-time performance by enabling better decision-making through centralized and distributed architectures, allowing for adaptive responses to complex construction environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067398A_ABST
    Figure CN119067398A_ABST
Patent Text Reader

Abstract

The invention provides a multi-agent construction equipment scheduling method and system, and construction equipment, and is applied to the technical field of engineering machinery, and the construction equipment scheduling method comprises the steps that a central integrated agent obtains overall state data and target task information of a target construction scene, the actual state data is acquired when the target equipment intelligent body performs construction in the target construction scene; the central integrated agent generates a target information group according to the overall state data, the target task information and the actual state data and sends the target information group to the target equipment agent, and the target equipment agent generates a scheduling decision of the target equipment agent according to the target information group and the actual state data. Therefore, through the target information group generated by the central integrated agent and the actual state data of the target equipment agent, the target equipment agent can make a task decision with a better scheduling effect and better practicability and real-time performance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of engineering machinery, and in particular to a multi-agent construction equipment scheduling method, system, and construction equipment. Background Art

[0002] In construction scenarios such as construction sites and mines, the scheduling of construction equipment is a key link that affects the progress of the project. Therefore, in order to complete the task objectives within the predetermined budget and construction period, it is crucial to develop a suitable construction equipment scheduling method.

[0003] There are two main types of existing construction equipment scheduling methods. One type of method uses simple logical judgment rules to make decisions based on the monitoring of current working conditions or the statistical results of historical working conditions. The scheduling effect and practicality of this type of method are poor. The other type of method is based on a multi-objective optimization algorithm to optimize the global situation. This type of method has poor real-time performance and cannot guarantee the ability to respond to emergencies. Summary of the invention

[0004] In view of this, the present application is dedicated to providing a multi-agent construction equipment scheduling method, system, and construction equipment. In this solution, the central integrated agent obtains the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment agent during construction in the target construction scene, generates a target information group, and sends the target information group to the target equipment agent. The target equipment agent generates a scheduling decision for the target equipment agent based on the target information group and the actual status data. This effectively solves the problems of poor scheduling effect, practicality, and real-time performance in the existing construction equipment scheduling methods.

[0005] In a first aspect, an embodiment of the present application provides a multi-agent construction equipment scheduling method, comprising:

[0006] The central integration agent obtains the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment agent when constructing in the target construction scene, wherein the actual status data includes the scene data and operation data of the target equipment agent;

[0007] The central integration agent generates a target information group according to the overall state data, the target task information and the actual state data, and sends the target information group to the target device agent, wherein the target information group includes target state information, target action information and target reward information, one target state information corresponds to at least one target action information, and one target action information corresponds to one target reward information;

[0008] The target device agent generates a scheduling decision for the target device agent according to the target information group and the actual status data.

[0009] Optionally, the central integrated intelligent agent generates a target information group according to the overall status data, the target task information and the actual status data, including:

[0010] The central integrated intelligent agent generates a plurality of initial information groups according to the overall state data, the target task information and the actual state data, each initial information group including state information, action information and reward information;

[0011] The central integrated intelligent agent calculates the state information, action information and reward information in each group of initial information according to the target task information to obtain first predicted completion information of each group of initial information for the target task information;

[0012] The central integration intelligent agent selects, based on the first preset completion condition, an initial information group corresponding to the first predicted completion information matching the first preset completion condition from among a plurality of the first predicted completion information as the target information group.

[0013] Optionally, the central integrated intelligent agent generates multiple groups of initial information according to the overall state data, the target task information and the actual state data, including:

[0014] The central integrated agent generates the current state information of the target device agent according to the overall state data, the target task information and the actual state data;

[0015] Generate multiple predicted action information of the target device agent according to the current state information of the target device agent, wherein the current state information includes the current actual state data, the overall state data and the completion status of the target task information;

[0016] The central integrated intelligent agent generates multiple predicted reward information of the target device intelligent agent based on the multiple predicted action information and reward mechanism, and the predicted reward information is used to characterize the pros and cons of executing a certain predicted action information.

[0017] Optionally, the generation process of the reward mechanism includes:

[0018] Constructing a reward design model in a central integrated intelligent agent based on the central integrated intelligent agent;

[0019] The reward design model generates a reward mechanism according to preset safety requirements, target requirements, the target task information, and the progress of completion of the target task information.

[0020] Optionally, the target device agent generates a scheduling decision of the target device agent according to the target information group and the actual status data, including:

[0021] The target device agent generates a plurality of candidate scheduling decisions according to the target information group and the actual status data;

[0022] The target equipment agent calculates each candidate decision according to the construction efficiency, cost, equipment loss, construction period and profit level according to the target task information to obtain second predicted completion information of each candidate decision;

[0023] The target device agent selects a candidate decision corresponding to the second predicted completion information matching the second preset completion condition from a plurality of the second predicted completion information as a scheduling decision of the target device agent based on the second preset completion condition.

[0024] Optionally, the multi-agent construction equipment scheduling method provided in the first aspect of the present application further includes:

[0025] The target device agent sends the scheduling decision to the central integration agent.

[0026] Optionally, after the central integration agent acquires the overall status data and target task information of the target construction scene, and the actual status data acquired by the target device agent when constructing in the target construction scene, and before the central integration agent generates a target information group based on the overall status data, the target task information and the actual status data, and sends the target information group to the target device agent, the method further includes:

[0027] The central integrated intelligent agent constructs a state understanding model in the central integrated intelligent agent;

[0028] The state understanding model converts the overall state data, the target task information and the actual state data into data in a target format;

[0029] The central integration agent generates a target information group based on the overall status data, target task information and actual status data converted into a target format, and sends the target information group to the target device agent.

[0030] Optionally, sending the target information group to the target device agent comprises:

[0031] The central integrated intelligent agent constructs a strategy explanation model in the central integrated intelligent agent;

[0032] The policy interpretation model parses the target information group into target information groups corresponding to each target device agent, and sends the parsed target information groups to the corresponding target device agents.

[0033] In a second aspect, the present application embodiment provides a multi-agent construction equipment scheduling system, including: a central integration agent and a target equipment agent, wherein

[0034] The central integrated agent is used to obtain the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment agent when constructing in the target construction scene, wherein the actual status data includes the scene data and operation data of the target equipment agent;

[0035] The central integration agent generates a target information group according to the overall state data, the target task information and the actual state data, and sends the target information group to the target device agent, wherein the target information group includes target state information, target action information and target reward information;

[0036] The target device agent is used to generate a scheduling decision for the target device agent according to the target information group and the actual status data of the target device agent.

[0037] In a third aspect, an embodiment of the present application provides a construction equipment, including:

[0038] The multi-agent construction equipment scheduling system described in the second aspect.

[0039] Based on the above content, the present application provides a multi-agent construction equipment scheduling method, in which the central integrated intelligent agent obtains the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment intelligent agent when constructing in the target construction scene, generates a target information group, and sends the target information group to the target equipment intelligent agent, and the target equipment intelligent agent generates a scheduling decision for the target equipment intelligent agent based on the target information group and the actual status data. In this way, through the target information group generated by the central integrated intelligent agent and the actual status data of the target equipment intelligent agent itself, the target equipment intelligent agent makes a task decision with better scheduling effect, better practicality and better real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 A flowchart of a multi-agent construction equipment scheduling method provided in an embodiment of the present application;

[0042] Figure 2 A flowchart of a multi-agent construction equipment scheduling method provided in an embodiment of the present application;

[0043] Figure 3 A flowchart of a multi-agent construction equipment scheduling method provided in an embodiment of the present application;

[0044] Figure 4 A flowchart of a multi-agent construction equipment scheduling method provided in an embodiment of the present application;

[0045] Figure 5 A flowchart of a multi-agent construction equipment scheduling method provided in an embodiment of the present application;

[0046] Figure 6 A structural block diagram of a multi-agent construction equipment scheduling system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] Unless otherwise defined, the technical terms or scientific terms used in the embodiments of this specification shall have the common meanings understood by persons with ordinary skills in the field to which this specification belongs. The words "first", "second" and similar words used in the embodiments of this specification do not indicate any order, quantity or importance, but are only used to avoid confusion of constituent elements.

[0048] Unless the context requires otherwise, throughout the specification, "plurality" means "at least two", and "including" is interpreted as an open, inclusive meaning, that is, "including, but not limited to". In the description of the specification, the terms "one embodiment", "some embodiments", "exemplary embodiments", "examples", "specific examples" or "some examples" are intended to indicate that a particular feature, structure, material or characteristic associated with the embodiment or example is included in at least one embodiment or example of the specification. The schematic representation of the above terms does not necessarily refer to the same embodiment or example.

[0049] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.

[0050] Application Overview

[0051] Agent: In the field of artificial intelligence, an agent refers to a program or machine that can simulate human intelligent behavior and has interactive capabilities. They can perceive environmental information, understand language and semantics, reason and make decisions, and perform various tasks.

[0052] Large Language Model (LLM) refers to a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as language prediction, text classification, question answering, summary generation, text generation, machine translation, etc., and are an important path to artificial intelligence.

[0053] Reinforcement learning is a type of strategy learning in artificial intelligence and an important machine learning method, also known as reinforcement learning or evaluation learning. Reinforcement learning refers to the learning of mapping from the state in the environment to the action, so as to maximize the accumulated reward value obtained by the action from the environment, and discover the optimal behavior strategy through trial and error.

[0054] As described in the background technology, in the prior art, two main construction equipment scheduling methods are used to schedule the construction process of the project. One is to adopt simple logical judgment rules to make decisions based on the monitoring of the current working conditions or the statistical results of the historical working conditions. The scheduling effect and practicality of this method are poor; the other method is based on a multi-objective optimization algorithm to optimize the global situation. This method has poor real-time performance and cannot guarantee the ability to respond to emergencies.

[0055] In order to solve the above technical problems, the present application provides a multi-agent construction equipment scheduling method, system, and construction equipment.

[0056] The multi-agent construction equipment scheduling method provided in the implementation mode of this specification will be exemplarily described below with reference to the accompanying drawings.

[0057] Exemplary Methods

[0058] like Figure 1 As shown, in this exemplary embodiment, a multi-agent construction equipment scheduling method is provided, which may include the following steps:

[0059] S100: The central integrated intelligent agent obtains the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment intelligent agent when constructing in the target construction scene.

[0060] The target construction scene can be a construction site, a mine or other construction scenes. This application takes the target construction scene as an example. The central integrated intelligent agent obtains the overall status data of the target construction scene, that is, the central integrated intelligent agent obtains the overall status data of the mine. The overall status data of the mine generally includes: the location of construction equipment, the number of construction equipment, the status of construction equipment, the location of the work point, the number of personnel, the location of personnel, the status of ore, mine road conditions, weather information and other information.

[0061] The main ways for the central integrated intelligent body to obtain the overall status data of the mine are: preset information obtained through manual configuration or mine management system, such as mine map; information collected by mine monitoring and sensor equipment; mine monitoring can be high-level monitoring or other monitoring equipment that can provide an overview of the overall environment of the mine, and sensor equipment can be temperature sensors, etc.

[0062] In some embodiments, the target task information can be set by a mine administrator or obtained from a mine management system. Generally, the target task information remains unchanged before the target task is completed. However, when encountering tasks that require extra expedited work or encountering bad weather or other special circumstances, the target task information needs to be adaptively modified. The mine management system is a prior art and will not be explained in detail in this application.

[0063] The central integrated intelligent agent is connected to multiple target device intelligent agents in communication. The central integrated intelligent agent obtains the actual status data of multiple target device intelligent agents when they are constructing in the target construction scene. The actual status data includes: the scene data and operation data of the target device intelligent agent. The scene data of the target device intelligent agent includes: the surrounding environment information of the target device intelligent agent, the number, location and status of other target device intelligent agents around the target device intelligent agent, etc. Through the scene data of the target device intelligent agent, multiple target device intelligent agents can avoid each other during operation to prevent collisions. The operation data of the target device intelligent agent includes: the operation status, health status, location information, etc. of the target device intelligent agent itself.

[0064] In some embodiments, each construction equipment is an equipment agent, and the equipment agents may include energy replenishment agents, transportation agents, mining agents, and other construction agents. The energy replenishment agents include multiple battery swap station agents, the transportation agents include multiple dump truck agents, the mining agents include multiple excavator agents, and other construction agents include multiple crusher agents, multiple drainage pump agents, and other agents. The target equipment agent is at least one of the multiple equipment agents such as the battery swap station agent, dump truck agent, excavator agent, crusher agent, or drainage pump agent.

[0065] In some embodiments, each target device intelligent body is provided with hardware devices related to environmental perception, data transmission, data processing, etc. For example, the hardware devices for environmental perception may include lidar, camera, GPS, various sensors, etc., and the various sensors may include temperature, pressure, angle and other sensors. Among them, lidar, camera, GPS can be used to obtain the location information of the target device intelligent body, the location information and status information of other surrounding target device intelligent bodies, etc., and the various sensors can be used to obtain the target device intelligent body's own operating status, health status and other information; the hardware devices for data transmission may include wired or wireless transmission units such as Bluetooth, 4G / 5G, etc., and the hardware devices for data transmission are used for communication between the central integrated intelligent body and multiple target device intelligent bodies, and between the target device intelligent body and other connected target device intelligent bodies; the hardware devices for data processing may include CPU, GPU, and the hardware devices for data processing are used to process the acquired overall mine status data, target task information, and actual status data acquired by the target device intelligent body when constructing in the target construction scene.

[0066] In some implementations, each target device agent is communicatively connected to each other, and one target device agent can obtain scene data and operation data of other target device agents communicatively connected to it.

[0067] Through information exchange between various target equipment agents, each target equipment agent knows the role it plays in the entire mine construction task and the current overall construction progress, so as to make better decisions that are beneficial to the overall situation and achieve global optimization.

[0068] S200: The central integration agent generates a target information group according to the overall status data, target task information and actual status data, and sends the target information group to the target device agent.

[0069] The central integrated intelligent agent processes all information during the mine construction process. The information is usually multimodal and contains various categories such as tabular data, text, pictures, sounds, etc. This information includes the overall status data of the mine (including but not limited to: the location of construction equipment, the number of construction equipment, the status of construction equipment, the location of the work point, the number of personnel, the location of personnel, the status of ore, the mine road conditions, weather information (such as temperature, wind level, etc.)), target task information (such as completing 1,000 trucks of ore in three days) and the actual status data obtained by the target equipment agent during construction in the mine (including: the surrounding environment information of the target equipment agent, the number, location and status of other target equipment agents around the target equipment agent, the target equipment agent's own operating status, health status, location information, etc.).

[0070] The decision-making of intelligent agents is usually based on pre-defined rules, algorithms and models, which are designed and trained based on human knowledge and experience. Intelligent agents generate decisions and guide their behavior based on these rules, algorithms and models by receiving and processing input information.

[0071] The central integrated intelligent agent needs to be trained before making a decision (generating a target information group). This application constructs an initial central integrated intelligent agent decision model based on a large language model. The large language model can be any one of ChatGPT, GPT, Llama, Falcon, ChatGLM, Wenxin Yiyan, iFlytek Spark, Tongyi Qianwen, Baichuan, etc. The initial central integrated intelligent agent decision model is trained based on the overall status data of the mine construction scene to obtain a trained central integrated intelligent agent decision model.

[0072] The central integrated agent generates a target information group based on the trained decision model, the overall state data, the target task information and the actual state data, and sends the target information group to the target device agent. The target information group includes target state information, target action information and target reward information. There is a corresponding relationship between the target state information, target action information and target reward information. One target state information corresponds to at least one target action information, and one target action information corresponds to one target reward information.

[0073] The main function of the decision made by the central integrated agent decision model (generating target information groups) is to provide auxiliary basis for scheduling decisions to the target device agent, so that the target device agent can choose scheduling decisions that are more beneficial to completing the target task.

[0074] S300: The target device agent generates a scheduling decision for the target device agent based on the target information group and the actual status data.

[0075] Before each target device agent generates a scheduling decision, it is necessary to build and train a device agent decision model.

[0076] The device agent decision model is trained mainly based on the following steps:

[0077] An initial equipment agent decision model is constructed based on the large language model, and the initial equipment agent decision model is trained based on the overall status data of the mine construction scene to obtain a trained equipment agent decision model.

[0078] Each target device agent generates a scheduling decision based on its own trained decision model. Specifically, the target device agent receives the target information group sent by the central integration agent, and based on the target information group and the actual status data obtained by the target device agent during construction in the mine construction scenario, the target device agent generates a scheduling decision based on its own decision model.

[0079] Different scheduling decisions are generated according to the different target equipment agents. Specifically, if the target equipment agent is an excavator agent, the scheduling decisions generated are the excavation position, excavation quantity, and excavation speed of the excavator agent; if the target equipment agent is a dump truck agent, the scheduling decisions generated are the loading and unloading position, loading and unloading quantity, driving speed, driving path, whether to recharge, and the recharging location of the dump truck agent; if the target equipment agent is a battery swap station agent, the scheduling decisions generated are which battery the battery swap station agent replaces for the equipment that needs to be recharged, the charging time, and the charging amount; if the target equipment agent is a drainage pump or crusher agent, the scheduling decisions generated are the construction location, whether to construct, and the construction speed of the drainage pump or crusher agent.

[0080] This application first obtains the overall status data and target task information of the target construction scene through the central integrated intelligent agent, as well as the actual status data obtained by the target device intelligent agent when constructing in the target construction scene. Then the central integrated intelligent agent generates a target information group based on the overall status data, target task information and actual status data. The target information group includes target status information, target action information and target reward information, and sends the target information group to the target device intelligent agent. The target device intelligent agent generates a scheduling decision for the target device intelligent agent based on the target information group and the actual status data. The central integrated intelligent agent and the target device intelligent agent adopt a combination of centralized and distributed architectures, so that the intelligent agent can cope with more complex environments and tasks, and make task decisions with better scheduling effects, better practicality and real-time performance.

[0081] In some embodiments of the present application, Figure 2 As shown, in S200, the central integrated intelligent agent generates a target information group according to the overall state data, the target task information and the actual state data. The specific steps may include:

[0082] S211: The central integrated intelligent agent generates multiple groups of initial information according to the overall state data, target task information and actual state data, each of which includes state information, action information and reward information.

[0083] The central integrated agent decision model generates candidate decisions for each device agent, i.e., multiple groups of initial information groups, based on the candidate instructions, the overall state data of the mine construction scene, the target task information, the actual state data obtained by each device agent during construction in the mine construction scene, and the input prompt text. Among them, the input prompt text can be generated for each device agent based on the currently received mine state information and reward information, and multiple groups of feasible state information, action information, and reward information, i.e., multiple groups of initial information groups, are required to be generated, so that each device agent obtains accurate state information, action information, and reasonable reward information, and the number of initial information groups is at least 2.

[0084] For example, the first initial information group includes status information: the current inventory of loading point 1 is 50 tons of ore, the current inventory of loading point 2 is 10 tons of ore, the current demand of unloading point 2 is 20 tons of ore, there is a dump truck currently swapping batteries at swap station 3, and the swapping is expected to be completed in 2 minutes, there is a queue for swapping batteries at swap station 4, action information A: dump truck 1 goes to loading point 1 to load, reward information A: reward value 2, action information B: dump truck 1 goes to swap station 3 to recharge, reward information B: reward value 1;

[0085] The second group of initial information includes status information: loading point 1 currently has 50t of ore in stock, loading point 2 currently has 10t of ore in stock, unloading point 2 currently needs 20t of ore, there is a dump truck currently swapping batteries at battery swap station 3, and the swapping is expected to be completed in 2 minutes, there is a battery swapping queue at battery swap station 4, action information C: dump truck 1 goes to loading point 2 to load, reward information: reward value -1; action information D: dump truck 1 goes to battery swap station 4 to recharge, reward information: reward value -1.

[0086] S212: The central integrated intelligent agent calculates the state information, action information and reward information in each group of initial information groups according to the target task information to obtain the first predicted completion information of each group of initial information groups for the target task information.

[0087] The central integrated intelligent agent decision model analyzes the advantages and disadvantages of multiple groups of initial information groups according to the input prompt text. The input prompt text can be the initial information groups of multiple groups of equipment agents currently given, and analyzes the advantages and disadvantages of multiple groups of initial information groups, and requires the disadvantage analysis to be based on the decision preview of the equipment agent after receiving the initial information group, previewing the gains and risks of subsequent decisions of the equipment for the overall goals of the mine.

[0088] The central integrated intelligent agent calculates the state information, action information and reward information in each group of initial information groups according to the target task information, and can obtain the overall reward information in each group of initial information groups. According to the overall reward information of each group of initial information groups, the first predicted completion information of each group of initial information groups for the target task information is obtained.

[0089] In some implementations, the first predicted completion information is the total reward information calculated based on a certain initial information group to determine the progress percentage of the initial information group in completing the target task information.

[0090] S213: The central integrated intelligent agent selects, based on the first preset completion condition, an initial information group corresponding to the first predicted completion information matching the first preset completion condition from among the plurality of first predicted completion information as the target information group.

[0091] The central integrated agent decision model sends the optimal initial information group (target information group) to the device agent according to the input prompt text. The input prompt text can be the optimal initial information group (target information group) for each device agent based on the candidate decision, i.e., multiple groups of initial information groups and the advantages and disadvantages analysis of the multiple groups of initial information groups.

[0092] A first preset completion condition is set in advance, and the first predicted completion information of the target task information of each initial information group is compared with the first preset completion condition respectively, and the first predicted completion information matching the first preset completion condition is selected therefrom, and the initial information group corresponding to the first predicted completion information matching the first preset completion condition is set as the target information group.

[0093] In some embodiments of the present application, Figure 3 As shown, in S211, the central integrated intelligent agent generates multiple groups of initial information groups according to the overall state data, target task information and actual state data. The specific steps may include:

[0094] S2111: The central integrated agent generates the current status information of the target device agent based on the overall status data, target task information and actual status data.

[0095] The central integrated agent obtains the overall status data of the mine, target task information, and the actual status data obtained by the target equipment agent when constructing in the mine construction scene, and generates the current status information of the target equipment agent based on the overall status data of the mine, target task information, and the actual status data obtained by the target equipment agent when constructing in the mine construction scene. The current status information of the target equipment agent includes: the current actual status data of the target equipment agent, the current overall status data of the mine, and the completion status of the target task information.

[0096] S2112: Generate multiple predicted action information of the target device agent based on the current state information of the target device agent.

[0097] In some embodiments, the predicted action information is the next possible scheduling decision of the target device agent. For example, the current state information a of the target device agent: the current inventory of loading point 1 is 50 tons of ore, the current inventory of loading point 2 is 10 tons of ore, the current demand of unloading point 2 is 20 tons of ore, there is a dump truck currently swapping batteries at swap station 3, and the swapping is expected to be completed in 2 minutes, and there is a queue for swapping batteries at swap station 4. The multiple predicted action information of the target device agent can be generated in the following situations:

[0098] Predicted action information a: Dump truck 1 goes to loading point 1 to load;

[0099] Predicted action information b: Dump truck 1 goes to loading point 2 to load materials;

[0100] Predicted action information c: dump truck 1 goes to battery swap station 3 to recharge;

[0101] Predicted action information d: Dump truck 1 goes to battery swap station 4 to recharge.

[0102] S2113: The central integrated intelligent agent generates multiple predicted reward information for the target device intelligent agent based on multiple predicted action information and reward mechanisms.

[0103] The reward mechanism is set up in advance. The predicted reward information is used to characterize the pros and cons of executing a certain predicted action information. One predicted action information corresponds to one predicted reward information. The central integrated intelligent agent generates a predicted reward information for the target device intelligent agent based on one predicted action information and reward mechanism.

[0104] In some embodiments of the present application, the process of generating the reward mechanism includes: constructing a reward design model in the central integrated intelligent agent based on the central integrated intelligent agent, and the reward design model generates a reward mechanism according to preset safety requirements, target requirements, the target task information, and the progress of completion of the target task information.

[0105] Preset safety requirements and target demands can be set manually. For example, preset safety requirements may include not carrying out construction in rainy weather and shutting down for repairs when equipment fails. Target demands may include charging equipment batteries during periods of low electricity consumption.

[0106] In some embodiments of the present application, Figure 4 As shown, the specific steps of sending the target information group to the target device agent in S200 may include:

[0107] S221: Central integrated agent constructs a strategy explanation model in the central integrated agent.

[0108] The central integrated agent builds a policy explanation model based on the large language model.

[0109] S222: The policy interpretation model parses the target information group into target information groups corresponding to each target device agent, and sends the parsed target information groups to the corresponding target device agents.

[0110] The central integrated intelligent agent obtains the overall status data and target task information of the mine construction scene, as well as the actual status data obtained by multiple target device intelligent agents during construction in the mine construction scene. The central integrated intelligent agent generates multiple target information groups based on the overall status data, target task information, and the actual status data obtained by multiple target device intelligent agents during construction in the mine construction scene. However, the central integrated intelligent agent does not know the target device intelligent agent corresponding to each target information group. The strategy interpretation model serves as a bridge between the central integrated intelligent agent and multiple target device intelligent agents, parses multiple target information groups into target information groups corresponding to each target device intelligent agent, and sends the parsed target information groups to the corresponding target device intelligent agents, ensuring that each target device intelligent agent receives the information data it needs.

[0111] The strategy interpretation model of the central integrated intelligent agent parses multiple target information groups into target information groups corresponding to each target device intelligent agent based on the input prompt text, and sends the parsed target information groups to the corresponding target device intelligent agent. The input prompt text can be used to understand the current multiple target information groups and standardize the multiple target information groups into a standard format so that they can be fed back to each device intelligent agent in a standardized manner.

[0112] In some embodiments of the present application, the target device agent in S300 generates a scheduling decision for the target device agent according to the target information group and the actual status data, and the specific steps may include:

[0113] S311: The target device agent generates multiple candidate scheduling decisions based on the target information group and actual status data.

[0114] The decision model of the device agent generates multiple next-step candidate scheduling decisions based on the received prompt text. The input prompt text can be to generate multiple next-step candidate scheduling decisions based on the currently received target information group, and it is required to generate multiple next-step candidate scheduling decisions, expecting to obtain the maximum reward, and more than 2 candidate scheduling decisions.

[0115] After receiving the target information group sent by the central integration agent, the target device agent generates multiple candidate scheduling decisions based on the current actual status data obtained by the target device agent itself. The current actual status data includes the scene data and operation data of the target device agent. The multiple candidate scheduling decisions generated by the target device agent may or may not correspond to the target action information in the target information group. Specifically, it is necessary to compare the target information group sent by the central integration agent with the current actual status data obtained by the target device agent itself.

[0116] S312: The target equipment agent calculates each candidate decision based on the construction efficiency, cost, equipment loss, construction period and profit level according to the target task information to obtain the second predicted completion information of each candidate decision.

[0117] The decision-making model of the device agent analyzes the advantages and disadvantages of multiple candidate scheduling decisions based on the received prompt text. The input prompt text can be to analyze the advantages and disadvantages of multiple candidate scheduling decisions based on the multiple candidate scheduling decisions currently given, and the advantages and disadvantages analysis is required to be based on the reward preview obtained after the scheduling decision, and preview the reward feedback given by the central agent based on the current target information group.

[0118] On the premise that the target task information is determined, the target equipment intelligent agent scores each candidate decision according to construction efficiency, cost, equipment loss, construction period and profit level, and obtains the overall score of each candidate decision. According to the overall score of each candidate decision, the second predicted completion information of each candidate decision is obtained.

[0119] In some implementations, the second predicted completion information is an overall score calculated based on a candidate decision to determine the progress percentage of the candidate decision in completing the target task information.

[0120] S313: The target device agent selects, based on the second preset completion condition, a candidate decision corresponding to the second predicted completion information matching the second preset completion condition from among a plurality of second predicted completion information as a scheduling decision for the target device agent.

[0121] The decision model of the target device agent selects the best candidate decision according to the input prompt text. The input prompt text can be to select the best candidate decision according to the analysis of the advantages and disadvantages of the candidate decision and multiple candidate decisions. The second preset completion condition is set in advance, and the second predicted completion information of each candidate decision is compared with the second preset completion condition respectively, and the second predicted completion information matching the second preset completion condition is selected, and the candidate decision corresponding to the second predicted completion information matching the second preset completion condition is set as the scheduling decision of the target device agent.

[0122] In some embodiments of the present application, the multi-agent construction equipment scheduling method further includes: the target equipment agent sends the scheduling decision to the central integration agent.

[0123] All target device agents send scheduling decisions to the central integrated agent, which then performs self-reinforcement learning after receiving these scheduling decisions. The central integrated agent can perform self-reinforcement learning by using reinforcement learning algorithms such as TD (Temporal Difference), Q learning algorithm, Sarsa algorithm, etc.

[0124] In some embodiments of the present application, after the central integrated agent obtains the overall state data and target task information of the target construction scene, and the actual state data obtained by the target equipment agent when constructing in the target construction scene, and before the central integrated agent generates a target information group based on the overall state data, the target task information and the actual state data, and sends the target information group to the target equipment agent, the multi-agent construction equipment scheduling method further includes:

[0125] The central integrated agent constructs a state understanding model in the central integrated agent, and the state understanding model converts the overall state data, target task information and actual state data into data in a target format.

[0126] The overall status data, target task information and actual status data obtained by the central integrated intelligent agent are converted into data in a target format, that is, data in a unified format.

[0127] The central integration agent generates a target information group according to the overall status data, target task information and actual status data converted into a target format, and sends the target information group to the target device agent.

[0128] The generated target information group is also data in a unified format, which ultimately ensures a high degree of data consistency during the decision-making process.

[0129] Exemplary Systems

[0130] In some embodiments of the present application, Figure 6 As shown, a multi-agent construction equipment scheduling system is provided, including: a central integration agent 601 and a target equipment agent 602, wherein:

[0131] The central integration agent 601 is used to obtain the overall state data of the target construction scene, target task information, and the actual state data obtained by the target equipment agent when constructing in the target construction scene, wherein the actual state data includes: the scene data and operation data of the target equipment agent;

[0132] The central integrated agent generates a target information group according to the overall state data, target task information and actual state data, and sends the target information group to the target device agent. The target information group includes target state information, target action information and target reward information.

[0133] The target device agent 602 is used to generate a scheduling decision of the target device agent according to the target information group and the actual status data of the target device agent.

[0134] The construction equipment scheduling system provided in this embodiment belongs to the same application concept as the multi-agent construction equipment scheduling method provided in the above embodiments of this application, and can execute the multi-agent construction equipment scheduling method provided in any of the above embodiments of this application, and has the corresponding functional units and beneficial effects of executing the multi-agent construction equipment scheduling method. For technical details not fully described in this embodiment, please refer to the specific processing content of the multi-agent construction equipment scheduling method provided in the above embodiments of this application, which will not be repeated here.

[0135] Exemplary Construction Equipment

[0136] In some embodiments of the present application, a construction equipment is provided, including: the above-mentioned multi-agent construction equipment scheduling system.

[0137] Exemplary Electronic Devices

[0138] In some embodiments of the present application, an electronic device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, any of the above-mentioned multi-agent construction equipment scheduling methods is implemented.

[0139] Specifically, the electronic device may further include: a bus, a communication interface, an input device and an output device. The processor, the memory, the communication interface, the input device and the output device are interconnected through the bus. The bus may include a path to transmit information between various components of the computer system.

[0140] The processor may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present application. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component.

[0141] The processor may include a main processor and may also include a baseband chip, a modem, etc.

[0142] The memory stores a computer program for executing the technical solution of the present application, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes computer operation instructions. More specifically, the memory may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.

[0143] The computer program may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the embodiments of the present specification, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0144] The input device may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0145] Output devices may include means that allow information to be output to a user, such as display screens, printers, speakers, etc.

[0146] The communication interface may include using any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0147] The processor executes the program stored in the memory and calls other devices, which can be used to implement the various steps of the multi-agent construction equipment scheduling method provided in the above-mentioned implementation mode of the present application.

[0148] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, etc. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of a software functional module. The present disclosure is not limited to any particular form of combination of hardware and software.

[0150] Unless otherwise defined, all terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present disclosure belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or highly formal sense, unless explicitly defined as such herein.

[0151] The above is an explanation of the present disclosure and should not be considered as a limitation thereof. Although several exemplary embodiments of the present disclosure are described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present disclosure. Therefore, all such modifications are intended to be included within the scope of the present disclosure as defined in the claims. It should be understood that the above is an explanation of the present disclosure and should not be considered to be limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present disclosure is defined by the claims and their equivalents.

Claims

1. A multi-agent construction equipment scheduling method, characterized in that: The method comprises: The central integration agent obtains the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment agent when constructing in the target construction scene, wherein the actual status data includes the scene data and operation data of the target equipment agent; The central integration agent generates a target information group according to the overall state data, the target task information and the actual state data, and sends the target information group to the target device agent, wherein the target information group includes target state information, target action information and target reward information, one target state information corresponds to at least one target action information, and one target action information corresponds to one target reward information; The target device agent generates a scheduling decision for the target device agent according to the target information group and the actual status data.

2. The multi-agent construction equipment scheduling method according to claim 1, characterized in that: The central integrated intelligent agent generates a target information group according to the overall status data, the target task information and the actual status data, including: The central integrated intelligent agent generates a plurality of initial information groups according to the overall state data, the target task information and the actual state data, each initial information group including state information, action information and reward information; The central integrated intelligent agent calculates the state information, action information and reward information in each group of initial information according to the target task information to obtain first predicted completion information of each group of initial information for the target task information; The central integration intelligent agent selects, based on the first preset completion condition, an initial information group corresponding to the first predicted completion information matching the first preset completion condition from among a plurality of the first predicted completion information as the target information group.

3. The multi-agent construction equipment scheduling method according to claim 2 is characterized in that: The central integrated intelligent agent generates a plurality of initial information groups according to the overall state data, the target task information and the actual state data, including: The central integrated agent generates the current state information of the target device agent according to the overall state data, the target task information and the actual state data; Generate multiple predicted action information of the target device agent according to the current state information of the target device agent, wherein the current state information includes the current actual state data, the overall state data and the completion status of the target task information; The central integrated intelligent agent generates multiple predicted reward information of the target device intelligent agent based on the multiple predicted action information and reward mechanism, and the predicted reward information is used to characterize the pros and cons of executing a certain predicted action information.

4. The multi-agent construction equipment scheduling method according to claim 3 is characterized in that: The generation process of the reward mechanism includes: Constructing a reward design model in a central integrated intelligent agent based on the central integrated intelligent agent; The reward design model generates a reward mechanism according to preset safety requirements, target requirements, the target task information, and the progress of completion of the target task information.

5. The multi-agent construction equipment scheduling method according to claim 1, characterized in that: The target device agent generates a scheduling decision of the target device agent according to the target information group and the actual status data, including: The target device agent generates a plurality of candidate scheduling decisions according to the target information group and the actual status data; The target equipment agent calculates each candidate decision according to the construction efficiency, cost, equipment loss, construction period and profit level according to the target task information to obtain second predicted completion information of each candidate decision; The target device agent selects a candidate decision corresponding to the second predicted completion information matching the second preset completion condition from a plurality of the second predicted completion information as a scheduling decision of the target device agent based on the second preset completion condition.

6. The multi-agent construction equipment scheduling method according to claim 1, characterized in that: The method further comprises: The target device agent sends the scheduling decision to the central integration agent.

7. The multi-agent construction equipment scheduling method according to claim 1, characterized in that: After the central integration agent acquires the overall status data and target task information of the target construction scene, and the actual status data acquired by the target device agent when constructing in the target construction scene, and before the central integration agent generates a target information group based on the overall status data, the target task information and the actual status data, and sends the target information group to the target device agent, the method further includes: The central integrated intelligent agent constructs a state understanding model in the central integrated intelligent agent; The state understanding model converts the overall state data, the target task information and the actual state data into data in a target format; The central integration agent generates a target information group based on the overall status data, target task information and actual status data converted into a target format, and sends the target information group to the target device agent.

8. The multi-agent construction equipment scheduling method according to claim 1, characterized in that: The step of sending the target information group to the target device agent comprises: The central integrated intelligent agent constructs a strategy explanation model in the central integrated intelligent agent; The policy interpretation model parses the target information group into target information groups corresponding to each target device agent, and sends the parsed target information groups to the corresponding target device agents.

9. A multi-agent construction equipment scheduling system, characterized in that: include: Central integration agent and target device agent, where The central integrated agent is used to obtain the overall status data of the target construction scene, the target task information, and the actual status data obtained by the target equipment agent when constructing in the target construction scene, wherein the actual status data includes the scene data and operation data of the target equipment agent; The central integration agent generates a target information group according to the overall state data, the target task information and the actual state data, and sends the target information group to the target device agent, wherein the target information group includes target state information, target action information and target reward information; The target device agent is used to generate a scheduling decision for the target device agent according to the target information group and the actual status data of the target device agent.

10. A construction equipment, characterized in that: include: The dispatching system for construction equipment as claimed in claim 9.