A multi-load automatic tractor online scheduling method and system

Abstract

The present application relates to multi-agent system and wisdom logistics technical field, specifically provide a kind of multi-load automatic tractor online scheduling method and system, comprising: constructing e-commerce intelligent warehouse scene model;Establish multi-load AGV online order scheduling problem model;Define agent type, based on the improved contract net protocol mechanism is combined with all agents to build multi-agent system;Establish order packing mechanism, establish distribution transport mechanism, based on order packing mechanism and distribution transport mechanism to realize order task allocation;Get the task execution order of each AGV agent, configure the work priority for each AGV agent, combine AGV agent work priority and task execution order to carry out conflict-free scheduling.When multiple AGVs simultaneously execute carrying task, it is not easy to produce conflict, improve the scheduling efficiency of e-commerce intelligent warehouse.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-agent systems and intelligent logistics, and specifically relates to an online scheduling method and system for multi-load automatic tractors. Background Art

[0002] In recent years, with the rapid development of e-commerce logistics, the demand for warehousing and logistics has continued to increase. Traditional manual sorting can no longer meet the high efficiency and high precision requirements. Currently, with the continuous expansion of the application of automatic guided vehicles (AGVs), they have been widely used as intelligent logistics equipment in e-commerce smart warehouses, providing a new solution for warehousing and logistics.

[0003] In the online order task scheduling scenario of an e-commerce smart warehouse, during the operation of the smart warehouse, users place orders through the online shopping platform. After the smart warehouse receives the order, it mobilizes AGVs to perform transportation tasks according to the order.

[0004] As the scale and order volume of e-commerce smart warehouses gradually increase, the number of AGVs required gradually increases. When multiple AGVs perform handling tasks at the same time, conflicts are likely to occur, resulting in low scheduling efficiency of e-commerce smart warehouses. Summary of the Invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides an online dispatching method and system for multi-load automatic tractors to solve the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides an online scheduling method for multi-load automatic tractors, comprising:

[0007] S1, obtain the layout rules of the e-commerce smart warehouse, and build an e-commerce smart warehouse scenario model based on the grid method and topology method combined with the layout rules;

[0008] S2: Obtain online orders from the e-commerce smart warehouse, build a multi-load AGV online order scheduling problem model based on the e-commerce smart warehouse model and online orders, and obtain key parameters and objective functions based on the problem model;

[0009] S3, based on the multi-load AGV online order scheduling model, defines the agent types as order management agent, packaging station agent, AGV agent, and information management agent. Based on the improved contract network protocol mechanism, all agents are combined to build a multi-agent system. The party that provides task allocation is regarded as the tenderer, and the party that receives and executes the task is regarded as the bidder.

[0010] S4: Establish an order packaging mechanism with the order management agent as the tenderer and the packaging station agent as the bidder. Establish an allocation and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder. Order task allocation is achieved based on the order packaging mechanism and the allocation and transportation mechanism.

[0011] S5, based on key parameters and objective functions, obtains the task execution order of each AGV agent, configures the work priority for each AGV agent based on the power of the AGV and the number of its load tasks, and performs conflict-free scheduling by combining the AGV agent's work priority and task execution order.

[0012] In an optional embodiment, step S1 specifically includes:

[0013] Layout rules include shelf layout rules, AGV docking area layout rules, and packaging station layout rules;

[0014] Represent roads, shelves, packing tables and AGV docking areas in the warehouse environment based on the grid method;

[0015] Based on the topological method, the warehouse is divided into multiple areas with the intersection as the benchmark, the center of the intersection grid as the node, and the dividing lines of the two-way roads as the edges connecting the nodes.

[0016] In an optional embodiment, step S2 specifically includes:

[0017] Configure loading points, unloading points and loads for each online order task;

[0018] Combine all tasks currently undertaken by each AGV into a task group;

[0019] Calculate the time cost for each AGV to complete the task group. The specific calculation is:

[0020]

[0021] Among them, t Gg Indicates the completion of task group G g The time cost required, g represents the g-th task group, N g represents the number of tasks contained in the g-th task group, m represents the number of AGVs, M represents the total number of transportable AGVs, n represents the task number in the task group, t0 represents the time for each operation at the loading point or unloading point, t w represents the waiting time of AGV due to conflict avoidance when executing tasks, a0, a n-1 and a n are the loading points of the 1st, nth and n+1th tasks in the task group, b0, b n-1 and b nare the unloading points of the 1st, nth, and n+1th tasks in the task group, respectively. represents the time cost between node m and node a0, Represents node a n-1 and a n The time cost between Represents node a n The time cost between and b0, Represents node b n-1 and b n The time cost between

[0022] Obtain the sum of the path costs between the loading points and the unloading points of each AGV completing the task group and combine them to obtain the total path cost;

[0023] The scheduling optimization objective function of each AGV is calculated based on the total time cost and path cost of each AGV completing the task group. The specific calculation is:

[0024]

[0025] Among them, min() represents the operation of finding the minimum value.

[0026] In an optional embodiment, in step S4, establishing an order packaging mechanism with the order management agent as the tenderer and the packaging platform agent as the bidder specifically includes:

[0027] When there is a new order, the order management agent is set as the bidder and a tender document is prepared, which contains the location information of the ordered goods;

[0028] The packaging agent is the bidder. After receiving the bid information from the order management agent, it first determines whether the number of its currently unfinished packaging order tasks exceeds the threshold for accepting new orders.

[0029] If the conditions for accepting the order are met, a bid is submitted to the tenderer's order management agent, and the remaining unfinished packaging task information and the path length indicator from the location of the ordered goods are provided. The tenderer's order management agent then calculates the bid value and evaluates the bid. The bid value calculation formula is as follows:

[0030]

[0031] Among them, Val1 is the bid value of the bidder's packaging agent, n1 is the number of orders currently accepted by the bidder's packaging agent, N1 is the maximum number of orders that can be received, d xx′ is the Manhattan distance between the bidder and the location of the ordered goods, d maxis the maximum distance between the packing station agent and the goods, and α and β are used to adjust the weights of the number of orders undertaken and the distance to the goods;

[0032] The order management agent evaluates all bidders, selects the bidder with the largest bid value, issues a winning notice, and waits for the winning bidder to confirm. If the winning bidder confirms, the contract is signed and the task assignment is completed.

[0033] In an optional embodiment, in step S4, establishing a transportation allocation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder specifically includes:

[0034] When the packing station agent accepts the order task assigned by the order management agent, the packing station agent will act as the bidder to specify the bidding document. The bidding document information includes the location of the ordered goods and the location of the packing station.

[0035] As a bidder, the AGV agent, after receiving the bid information from the packaging station agent, first determines whether its power is sufficient to accept the new task and whether the current capacity can accommodate the new order goods;

[0036] If the conditions are met, a bid is initiated to the tenderer's packaging agent. The difference between the total length of the path planned after and before accepting the task, as well as the current number of accepted order tasks, is calculated as the bid information and sent to the packaging agent for calculating the bid value. The bidder's bid value calculation formula is as follows:

[0037]

[0038] Among them, Val2 is the bid value of the bidder's AGV intelligent body, n2 is the number of orders currently accepted by the bidder's AGV intelligent body, N2 is the maximum number of orders that can be received, l g The total length of the path planned by the bidder AGV agent before accepting the task, l g′ The total length of the path planned by the bidder's AGV agent after accepting the task, Δl max is the maximum difference between the total length of the path planned by the AGV agent before and after accepting the task and the distance to the goods. α′ and β′ are used to adjust the weights of the number of orders received by the AGV agent and the maximum difference between the total length of the path planned by the AGV agent before and after accepting the task and the distance to the goods.

[0039] The packaging station agent evaluates all bidders, selects the bidder with the largest bid value, issues a winning notice, and waits for the winning bidder to confirm; if the winning bidder confirms, a contract is signed, completing the task allocation from the packaging station agent to the AGV agent.

[0040] In an optional embodiment, in step S5, the task execution order of each AGV agent is obtained based on the key parameters and the objective function, specifically, when the number of tasks currently accepted by the AGV is not greater than 6, a two-stage branch and bound algorithm based on a priority queue is used to solve the problem:

[0041] S5-1, construct the adjacency matrix between the AGV's current position and all task loading points, and randomly generate a loading segment execution order as a limit, create a priority queue and put the AGV's current position and current path into the priority queue;

[0042] S5-2, take the first node of the queue for expansion and calculate the current path length. If the path value of a certain extended loading point is greater than the path of the current optimal loading segment execution order, then it will not be added to the priority queue. Otherwise, it will be added to the priority queue.

[0043] S5-3, if the current path already contains all loading points and the path length is less than the optimal solution of the current loading segment, then update the optimal solution of the loading segment;

[0044] S5-4, if the priority queue is empty, record the optimal loading segment execution order and the last loading point, otherwise jump to S5-2.

[0045] S5-5: Construct an adjacency matrix between the last loading point of the optimal loading segment execution order and all unloading points, randomly generate an unloading segment execution order as a limit, and put the last loading point and the current path into the priority queue;

[0046] S5-6 takes out the node at the head of the queue, expands it, and calculates the current path length. If the path value of a certain expanded unloading point is greater than the path of the currently found optimal unloading segment execution order, then it is not added to the priority queue. Otherwise, it is added to the priority queue.

[0047] S5-7 If the current path already contains all unloading points and the path length is less than the optimal solution of the current unloading segment, then update the optimal solution of the unloading segment;

[0048] If the priority queue is not empty in S5-8, jump to S5-6. Otherwise, record the optimal unloading segment execution order and merge it with the optimal loading segment execution order to obtain the final task execution order.

[0049] In an optional embodiment, in step S5, the task execution order of each AGV agent is obtained based on the key parameters and the objective function, specifically, when the number of tasks currently accepted by the AGV is greater than 6, a two-stage genetic algorithm is used to solve:

[0050] Initialize the genetic algorithm crossover probability, mutation probability, and population evolution generation, and generate the initial population based on the task execution order of the initial loading phase and the task execution order of the initial unloading phase as chromosomes;

[0051] Based on the tournament selection method, the current population is selected to generate a new population;

[0052] Perform crossover and mutation operations on individuals in the new population;

[0053] Calculate the fitness value of the new population based on the parameters and objective function in the problem model, record the execution order and path length of the optimal individuals in the population, and update the global optimal execution order and path length;

[0054] If the evolutionary generation reaches the preset generation, the optimal execution order and path length of the AGV are output, otherwise the crossover operation and mutation operation steps are returned.

[0055] In an optional embodiment, the step of configuring the work priority for each AGV agent based on the power of the AGV and the number of its load tasks in step S5 includes:

[0056] According to the power of AGV and the number of its load tasks, the working states of AGV intelligent body include idle state, risk state, loading state and unloading state;

[0057] When a risky AGV conflicts with other AGVs, the risky AGV has the highest priority. When a risky AGV conflicts with another risky AGV, one of the AGVs is randomly assigned a higher priority.

[0058] When an idle AGV conflicts with other AGVs, the idle AGV has the lowest priority. When an idle AGV conflicts with another idle AGV, one of the AGVs is randomly assigned a higher priority.

[0059] When a conflict occurs between the loading AGV and the unloading AGV, the unloading AGV has a higher priority;

[0060] When a loading AGV conflicts with another loading AGV, or an unloading AGV conflicts with another unloading AGV, the AGV with fewer remaining tasks to be completed has a higher priority. Otherwise, the AGV that is expected to complete the order task first has a higher priority. Otherwise, the AGV with a longer cumulative waiting time has a higher priority. If the above conditions are all the same, the randomly designated AGV has a higher priority.

[0061] In an optional embodiment, the specific steps for conflict-free scheduling based on the AGV agent's work priority and task execution order are as follows:

[0062] AGV puts the nodes in the order of execution of its own loading phase into the list one by one.

[0063] Take the first loading node in the list as the current task point, calculate the global planning path based on the topological dimension, and select the driving road;

[0064] The AGV agent detects whether a conflict occurs based on the grid dimension. If a conflict occurs, it compares the priority with the conflicting AGV agent and resolves the conflict based on the priority of the AGV agent, choosing to drive or stop and wait;

[0065] If the AGV has not reached the current loading point, it will jump to the step of whether a conflict occurs. Otherwise, the current loading node will be removed from the list. If the number of nodes in the list is not empty, it will jump to the step of taking the first loading node in the list as the current task point.

[0066] If a new order task is accepted, the execution order is replanned and the process is jumped to the step of adding the tasks to the list in sequence.

[0067] AGV puts the nodes in the order of execution of its own unloading phase into the list in sequence;

[0068] Take the first unloading node in the list as the current task point, calculate the global planning path based on the topological dimension, and select the driving road;

[0069] The AGV agent detects whether a conflict occurs based on the grid dimension. If a conflict occurs, it compares the priority with the conflicting AGV agent and resolves the conflict based on the priority of the AGV agent, choosing to drive or stop and wait;

[0070] If the AGV has not reached the current unloading node, it will jump to the step of detecting whether a conflict occurs. Otherwise, it will remove the current unloading node from the list. If the number of nodes in the list is not empty, it will jump to the step of taking the first unloading node in the list as the current task point.

[0071] If a new order task is accepted, it is placed in the buffer list and the task execution sequence is planned after all current unloading tasks are completed.

[0072] In a second aspect, the present invention provides an online dispatching system for multi-load automatic tractors. When the system is implemented, the above-mentioned online dispatching method for multi-load automatic tractors is executed. The system includes:

[0073] The scenario model construction module obtains the layout rules of the e-commerce smart warehouse and constructs the e-commerce smart warehouse scenario model based on the grid method and topology method combined with the layout rules;

[0074] The problem model construction module obtains online orders from e-commerce smart warehouses, builds a multi-load AGV online order scheduling problem model based on the e-commerce smart warehouse model and online orders, and obtains key parameters and objective functions based on the problem model;

[0075] The agent construction module defines agent types as order management agent, packaging station agent, AGV agent, and information management agent based on the multi-load AGV online order scheduling model. Based on the improved contract network protocol mechanism, all agents are combined to build a multi-agent system. The party that provides task allocation is regarded as the tenderer, and the party that receives and executes the task is regarded as the bidder.

[0076] The order task allocation module establishes an order packaging mechanism with the order management agent as the tenderer and the packaging station agent as the bidder. It also establishes an allocation and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder. Order task allocation is achieved based on the order packaging mechanism and the allocation and transportation mechanism.

[0077] The scheduling execution module obtains the task execution order of each AGV agent based on key parameters and objective functions, configures the work priority for each AGV agent based on the power of the AGV and the number of its load tasks, and performs conflict-free scheduling based on the work priority of the AGV agent and the task execution order.

[0078] The beneficial effects of the present invention are that the online scheduling method and system for multi-load automatic tractors provided by the present invention construct a scenario model by acquiring the layout rules of the e-commerce intelligent warehouse, establish a scheduling problem model in combination with online orders and obtain key parameters and objective functions, define multiple agent types to build a multi-agent system, establish an order packaging and distribution transportation mechanism to realize task allocation, obtain the task execution order based on key factors and configure work priority for conflict-free scheduling, effectively improve the order task scheduling efficiency of the e-commerce intelligent warehouse, optimize the AGV operation path, reasonably allocate tasks, adapt to dynamic changes in the warehouse, enhance the system robustness, improve the overall operational efficiency, and provide strong support for the efficient operation of e-commerce logistics.

[0079] In addition, the present invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0081] Figure 1The figure is a schematic flow chart of an online dispatching method for multi-load automatic tractors according to an embodiment of the present invention.

[0082] Figure 2 This is a schematic diagram of an e-commerce smart warehouse scenario construction based on the grid-topology dual dimensions according to an embodiment of the present invention.

[0083] Figure 3 It is a schematic diagram of a multi-agent system protocol mechanism according to an embodiment of the present invention.

[0084] Figure 4 It is a schematic diagram of a multi-agent and multi-agent system according to an embodiment of the present invention.

[0085] Figure 5 1 is a chromosome diagram of a two-stage genetic algorithm according to an embodiment of the present invention.

[0086] Figure 6 Schematic diagram of a crossover strategy of a two-stage genetic algorithm according to an embodiment of the present invention.

[0087] Figure 7 Schematic diagram of a mutation strategy of a two-stage genetic algorithm according to an embodiment of the present invention.

[0088] Figure 8 The figure is a schematic block diagram of an online dispatching system for automatic multi-load tractors according to an embodiment of the present invention. DETAILED DESCRIPTION

[0089] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0090] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0091] The multi-load automatic tractor online dispatching method provided in the embodiment of the present invention is executed by a computer device. Accordingly, the multi-load automatic tractor online dispatching system runs in the computer device.

[0092] Figure 1 This is a schematic flow chart of an online dispatching method for multi-load automatic tractors according to an embodiment of the present invention. Figure 1The execution subject can be a multi-load automatic tractor online dispatching system. According to different requirements, the order of the steps in the flow chart can be changed, and some can be omitted.

[0093] like Figure 1 As shown, the method includes:

[0094] S1, obtain the layout rules of the e-commerce smart warehouse, and build an e-commerce smart warehouse scenario model based on the grid method and topology method combined with the layout rules;

[0095] It can accurately reflect the actual layout structure within the warehouse, including the distribution of shelves, AGV docking areas, and packaging tables, providing a basic framework for subsequent operations.

[0096] S2: Obtain online orders from the e-commerce smart warehouse, build a multi-load AGV online order scheduling problem model based on the e-commerce smart warehouse model and online orders, and obtain key parameters and objective functions based on the problem model;

[0097] By incorporating the dynamic information of orders into the model, the key parameters (such as loading points, unloading points, AGV load capacity, etc.) and objective functions (such as minimizing the sum of total path and time) obtained based on such a model can accurately reflect the essence of the actual scheduling problem and provide a reliable basis for formulating scientific and reasonable scheduling strategies.

[0098] S3, based on the multi-load AGV online order scheduling model, defines the agent types as order management agent, packaging station agent, AGV agent, and information management agent. Based on the improved contract network protocol mechanism, all agents are combined to build a multi-agent system. The party that provides task allocation is regarded as the tenderer, and the party that receives and executes the task is regarded as the bidder.

[0099] It realizes the clear division of labor and efficient collaboration of different functional modules in the operation of e-commerce smart warehouses. The system selects the most suitable intelligent agent to perform the task and realizes the rational allocation of resources.

[0100] S4: Establish an order packaging mechanism with the order management agent as the tenderer and the packaging station agent as the bidder. Establish an allocation and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder. Order task allocation is achieved based on the order packaging mechanism and the allocation and transportation mechanism.

[0101] By establishing an order packaging mechanism, the order management agent is used as the tenderer and the packaging platform agent as the bidder, and the reasonable allocation of order tasks in the packaging link is achieved.

[0102] S5, based on key parameters and objective functions, obtains the task execution order of each AGV agent, configures the work priority for each AGV agent based on the power of the AGV and the number of its load tasks, and performs conflict-free scheduling by combining the AGV agent's work priority and task execution order.

[0103] The task execution sequence of each AGV intelligent body is obtained based on key parameters and objective functions, and scientific planning can be carried out according to the specific requirements of the task (such as loading point, unloading point location, cargo load, etc.) and the warehouse environment (such as road conditions, path cost between nodes, etc.).

[0104] Optionally, as an embodiment of the present invention, the specific steps of step S1 are as follows:

[0105] Layout rules include shelf layout rules, AGV docking area layout rules, and packaging station layout rules;

[0106] Represent roads, shelves, packing tables and AGV docking areas in the warehouse environment based on the grid method;

[0107] Based on the topological method, the warehouse intersection is used as the benchmark, the center of the intersection grid is the node, and the dividing line of the two-way road is used as the edge connecting the nodes. The warehouse is divided into multiple areas. The scene diagram obtained after modeling is as follows Figure 2 shown.

[0108] Optionally, as an embodiment of the present invention, step S2 specifically includes:

[0109] Let the length of each grid be d, then the distance between the centers of two adjacent grids is also d. Define that there are N L Loading points are located in adjacent grids of the rack area, and the collection of loading points is represented by the collection Indicates. There is N u unloading points, the set of unloading points is composed of the set express.

[0110] It is defined that there are M transportable AGVs in the system, which are represented by the set A = {AGV1, AGV2, ..., AGV M} indicates that AGV k The corresponding load capacity is given by C k The maximum load capacity is represented by a constant V. Under the premise of meeting the maximum load requirement, the warehouse can simultaneously undertake the delivery of goods at multiple loading points. Online orders from e-commerce warehouses arrive in real time, and the processed orders are converted into a set of task points (loading points and unloading points, where the loading point is the AGV pickup location and the unloading point is the location where the AGV unloads to the packaging station) and are identified by the loading point and unloading point. For example, a i and b i To identify task Ti , where a i Indicates its loading point, b i Indicates its unloading point. T ={a1, a2, ..., a n} represents the set of loading points in the task set, U T ={b1, b2, ..., b n} represents the set of unloading points in the task set. Each order task contains the following information: loading point a i , Uninstall point b i , load d i , that is, task T i ={a i , b i , d i}.

[0111] AGV regulations k It can receive new orders when the current load is less than V and re-plan the order of executing tasks. k After each task is accepted, all tasks currently received are considered as a task group. g Task Group G g The number of tasks included is N g .

[0112] Calculate the time cost for each AGV to complete the task group. The specific calculation is:

[0113]

[0114] Among them, t Gg Indicates the completion of task group G g The time cost required, g represents the g-th task group, N g represents the number of tasks contained in the g-th task group, m represents the number of AGVs, M represents the total number of transportable AGVs, n represents the task number in the task group, t0 represents the time for each operation at the loading point or unloading point, t w represents the waiting time of AGV due to conflict avoidance when executing tasks, a0, a n-1 and a n are the loading points of the 1st, nth and n+1th tasks in the task group, b0, b n-1 and b n are the unloading points of the 1st, nth, and n+1th tasks in the task group, respectively. represents the time cost between node m and node a0, Represents node a n-1 and a n The time cost between Represents node an The time cost between and b0, Represents node b n-1 and b n The time cost between

[0115] Obtain the sum of the path costs between the loading points and the unloading points of each AGV completing the task group and combine them to obtain the total path cost;

[0116] The scheduling optimization objective function of each AGV is calculated based on the total time cost and path cost of each AGV completing the task group. The specific calculation is:

[0117]

[0118] Among them, min() represents the operation of finding the minimum value.

[0119] The formula calculation method and constraints of the multi-load AGV scheduling solution model for e-commerce smart warehouses are as follows:

[0120]

[0121] The above formula means that each task T i Can only be assigned to one AGV.

[0122]

[0123] The above formula refers to AGV m The load capacity of a task group must be less than or equal to the sum of all cargo loads in its task group.

[0124]

[0125] The above formula means that at any time, the same grid in the warehouse can only accommodate one AGV at most.

[0126] Optionally, as an embodiment of the present invention, step S3 specifically includes:

[0127] By analyzing the scenario requirements and problem-solving characteristics of e-commerce smart warehouses, we divide the system into multiple classes of agents at different levels of abstraction. The online multi-load AGV scheduling problem in e-commerce smart warehouses can be decomposed into the process of the order system receiving external orders, the order system assigning order tasks to the packing station, the packing station assigning order tasks to the AGV, and the AGV executing the transport task. This process involves task collaboration among multiple agents. Therefore, through abstraction, the system can be divided into four classes of agents: the order management agent, the packing station agent, the AGV agent, and the information management agent, forming a multi-agent system.

[0128] A detailed analysis and design of four intelligent agents are conducted, namely the order management agent, the packing station agent, the AGV agent, and the information management agent. The functions of the order management agent include receiving orders from the external system, adding the corresponding pickup points for the ordered goods to the order task information, and allocating and tracking orders with the packing station agent. The functions of the packing station agent include accepting order tasks assigned from the order management agent, allocating order tasks between AGV agents, and receiving goods delivered by the AGV agent. The AGV agent is responsible for accepting order tasks assigned from the packing station agent, task execution sequence planning, conflict-free path planning, loading goods, transporting goods, and unloading goods. The information management agent is mainly responsible for recording and synchronizing feedback data from the execution of tasks by the order management agent, the packing station agent, and the AGV agent.

[0129] Determine the protocol mechanism of the multi-agent system, optimize the collaboration process between agents, improve coordination efficiency, improve the evaluation mechanism, and complete the construction of the multi-agent system. This step adopts an improved contract network protocol mechanism to divide the agents in the multi-agent system into two roles: tenderers and bidders. Among them, the party that provides task assignments is the tenderer, and the party that receives and executes tasks is the bidder. The protocol mechanism mainly includes processes such as bidding, bidding, bid evaluation, authorization, and contract signing. When the tenderer receives a task and needs to assign it, it will publish a message to all bidders and wait for a certain period of time to receive the bids from the bidders. After receiving the information, the bidder will choose whether to bid according to its own situation. If it chooses to bid, it will also send a bid containing bidding information for calculating the bid value. After receiving the bid, the tenderer will evaluate all the bids received and select the bidder with the largest bid value, that is, the bidder who is most suitable to perform the task, and initiate a contract invitation. The two parties sign the contract to complete the task assignment from the tenderer to the bidder. Figure 3 As shown. The multi-agent construction is completed, as shown Figure 4 shown.

[0130] Optionally, as an embodiment of the present invention, in step S4, establishing an order packaging mechanism with the order management agent as the tenderer and the packaging platform agent as the bidder specifically includes:

[0131] When there is a new order, the order management agent is set as the bidder and a tender document is prepared, which contains the location information of the ordered goods;

[0132] The packaging agent is the bidder. After receiving the bid information from the order management agent, it first determines whether the number of its currently unfinished packaging order tasks exceeds the threshold for accepting new orders.

[0133] If the conditions for accepting the order are met, a bid is submitted to the tenderer's order management agent, and the remaining unfinished packaging task information and the path length indicator from the location of the ordered goods are provided. The tenderer's order management agent then calculates the bid value and evaluates the bid. The bid value calculation formula is as follows:

[0134]

[0135] Among them, Val1 is the bid value of the bidder's packaging agent, n1 is the number of orders currently accepted by the bidder's packaging agent, N1 is the maximum number of orders that can be received, d xx′ is the Manhattan distance between the bidder and the location of the ordered goods, d max is the maximum distance between the packing station agent and the goods, and α and β are used to adjust the weights of the number of orders undertaken and the distance to the goods;

[0136] The order management agent evaluates all bidders, selects the bidder with the largest bid value, issues a winning notice, and waits for the winning bidder to confirm. If the winning bidder confirms, the contract is signed and the task assignment is completed.

[0137] Optionally, as an embodiment of the present invention, in step S4, establishing a distribution and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder specifically includes:

[0138] When the packing station agent accepts the order task assigned by the order management agent, the packing station agent will act as the bidder to specify the bidding document. The bidding document information includes the location of the ordered goods and the location of the packing station.

[0139] As a bidder, the AGV agent, after receiving the bid information from the packaging station agent, first determines whether its power is sufficient to accept the new task and whether the current capacity can accommodate the new order goods;

[0140] If the conditions are met, a bid is initiated to the tenderer's packaging agent. The difference between the total length of the path planned after and before accepting the task, as well as the current number of accepted order tasks, is calculated as the bid information and sent to the packaging agent for calculating the bid value. The bidder's bid value calculation formula is as follows:

[0141]

[0142] Among them, Val2 is the bid value of the bidder's AGV intelligent body, n2 is the number of orders currently accepted by the bidder's AGV intelligent body, N2 is the maximum number of orders that can be received, l g The total length of the path planned by the bidder AGV agent before accepting the task, l g′ The total length of the path planned by the bidder's AGV agent after accepting the task, Δlmax is the maximum difference between the total length of the path planned by the AGV agent before and after accepting the task and the distance to the goods. α′ and β′ are used to adjust the weights of the number of orders received by the AGV agent and the maximum difference between the total length of the path planned by the AGV agent before and after accepting the task and the distance to the goods.

[0143] The packaging station agent evaluates all bidders, selects the bidder with the largest bid value, issues a winning notice, and waits for the winning bidder to confirm; if the winning bidder confirms, a contract is signed, completing the task allocation from the packaging station agent to the AGV agent.

[0144] Optionally, as an embodiment of the present invention, task execution sequence planning means that when a multi-load AGV agent receives multiple tasks, it needs to plan the order in which the tasks are executed, so as to optimize the total length of the path. In a multi-agent system, the AGV agent is set to first execute the order picking and loading task, and then execute the unloading task, to prevent the goods of certain order tasks from being unloaded and packaged by the AGV load all the time. Therefore, the task execution sequence of the AGV agent is divided into two stages: the loading stage and the unloading stage. The loading stage refers to the process of the AGV traversing all task loading points from the current position, and the unloading stage refers to the process of traversing all task unloading points from the last loading point. Loading must precede unloading, and the two cannot intersect. The AGV can accept new orders during the loading stage and re-plan the execution sequence, but it is allowed to accept new orders during the unloading stage but will be placed in the buffer list, and continue to execute the current transportation task until all unloading tasks are completed. The present invention solves the following two situations based on the number of tasks currently accepted by the AGV:

[0145] When the number of tasks currently accepted by the AGV is no more than 6, a two-stage branch and bound algorithm based on the priority queue is used to solve the problem.

[0146] When the number of tasks currently accepted by AGV is greater than 6, a two-stage genetic algorithm is used to solve the problem.

[0147] Optionally, as an embodiment of the present invention, in step S5, the task execution order of each AGV agent is obtained based on the key parameters and the objective function, specifically, when the number of tasks currently accepted by the AGV is not greater than 6, a two-stage branch and bound algorithm based on a priority queue is used to solve the problem:

[0148] S5-1, construct the adjacency matrix between the AGV's current position and all task loading points, and randomly generate a loading segment execution order as a limit, create a priority queue and put the AGV's current position and current path into the priority queue;

[0149] S5-2, take the first node of the queue for expansion and calculate the current path length. If the path value of a certain extended loading point is greater than the path of the current optimal loading segment execution order, then it will not be added to the priority queue. Otherwise, it will be added to the priority queue.

[0150] S5-3, if the current path already contains all loading points and the path length is less than the optimal solution of the current loading segment, then update the optimal solution of the loading segment;

[0151] S5-4, if the priority queue is empty, record the optimal loading segment execution order and the last loading point, otherwise jump to S5-2.

[0152] S5-5: Construct an adjacency matrix between the last loading point of the optimal loading segment execution order and all unloading points, randomly generate an unloading segment execution order as a limit, and put the last loading point and the current path into the priority queue;

[0153] S5-6 takes out the node at the head of the queue, expands it, and calculates the current path length. If the path value of a certain expanded unloading point is greater than the path of the currently found optimal unloading segment execution order, then it is not added to the priority queue. Otherwise, it is added to the priority queue.

[0154] S5-7 If the current path already contains all unloading points and the path length is less than the optimal solution of the current unloading segment, then update the optimal solution of the unloading segment;

[0155] If the priority queue is not empty in S5-8, jump to S5-6. Otherwise, record the optimal unloading segment execution order and merge it with the optimal loading segment execution order to obtain the final task execution order.

[0156] Optionally, as an embodiment of the present invention, in step S5, the task execution order of each AGV agent is obtained based on the key parameters and the objective function, specifically, when the number of tasks currently accepted by the AGV is greater than 6, a two-stage genetic algorithm is used to solve:

[0157] The two-stage genetic algorithm adopts a specific segmented encoding method. The chromosome consists of two parts, A and B, where A represents the task execution order in the loading phase (LS) and B represents the task execution order in the unloading phase (US). Their lengths are both n i (The number of tasks currently accepted by the AGV).

[0158] The numbers represent the subscript index of the tasks within the task group. The LS part indicates that the AGV traverses the loading points of these tasks in sequence. The US part indicates that after the AGV completes all loading tasks, it traverses the unloading points of these tasks in sequence.

[0159] The selection strategy is to use the tournament selection method to select (1-P e)*N individuals, use the elite retention method to select P e *N individuals (P e proportion of elite individuals).

[0160] The crossover strategy is to randomly generate a length of 2*n i , and an array containing only 0 and 1. For the position with a value of 1 in the array, the tasks of the corresponding positions of the two individuals are cross-operated, and two child individuals are generated according to the mapping relationship.

[0161] The mutation operation is to randomly select a point in the loading section and a point in the unloading section as the mutation point, and randomly generate a new task point to exchange with it.

[0162] Reference Figure 5 , initialize the genetic algorithm crossover probability, mutation probability, population number evolution generation, and generate the initial population based on the task execution order of the initial loading phase and the task execution order of the initial unloading phase as chromosomes;

[0163] Based on the tournament selection method, the current population is selected to generate a new population;

[0164] Reference Figure 6 and Figure 7 Perform crossover and mutation operations on individuals in the new population;

[0165] Calculate the fitness value of the new population based on the parameters and objective function in the problem model, record the execution order and path length of the optimal individuals in the population, and update the global optimal execution order and path length;

[0166] If the evolutionary generation reaches the preset generation, the optimal execution order and path length of the AGV are output, otherwise the crossover operation and mutation operation steps are returned.

[0167] Optionally, as an embodiment of the present invention, the step of configuring the work priority for each AGV agent based on the power consumption of the AGV and the number of its load tasks in step S5 includes:

[0168] Since the operation cycle of e-commerce smart warehouses is relatively long, the power constraints of the AGV itself need to be considered. Therefore, a power threshold of the AGV is set. When the power threshold is lower than the AGV power threshold, in order to prevent the AGV from losing power during work, causing congestion or deadlock in the warehouse road, no new tasks will be assigned to the AGV. In addition, in order to ensure the orderliness of the AGV's work, the AGV cannot perform loading and unloading tasks at the same time. In order to more conveniently handle the task allocation problem based on the AGV's power and the current load task volume, a method for defining the working status of the AGV is proposed. According to the AGV's power and the number of its load tasks, the working status of the AGV is defined as follows:

[0169] (1) Idle state: The AGV has not been assigned a task and is in the docking area or on its way to the docking area. The battery level is above the threshold. A new order task can be assigned and executed immediately.

[0170] (2) Risky state: The AGV may be performing a loading task, an unloading task, or has not been assigned a task, but the current AGV power level is below the power threshold. No new tasks can be assigned, and the final destination of the risky AGV is the docking area.

[0171] (3) Loading state: The AGV is executing the loading task and the current power level is higher than the power threshold. New tasks can be assigned based on the current capacity of the AGV and the execution order can be replanned.

[0172] (4) Unloading state: The AGV is executing the unloading segment task, and the current power level is higher than the power threshold. A new task can be assigned and placed in the buffer pool. After all unloading segment tasks are completed, the execution sequence is planned.

[0173] Since multiple AGVs may conflict during mission execution, it is necessary to set priorities based on the AGV's working status. When a conflict occurs, obstacle avoidance can be performed based on the AGV's priority. For the above working status, the AGV priority in different states is defined.

[0174] (1) When a risky AGV collides with other AGVs, the risky AGV has the highest priority. When a risky AGV collides with another risky AGV, one of the AGVs is randomly assigned a higher priority.

[0175] (2) When an idle AGV conflicts with other AGVs, the idle AGV has the lowest priority. When an idle AGV conflicts with another idle AGV, one of the AGVs is randomly assigned a higher priority.

[0176] (3) When a conflict occurs between a loading AGV and an unloading AGV, the unloading AGV has a higher priority.

[0177] (4) When a loading AGV collides with another loading AGV, or when an unloading AGV collides with another unloading AGV, the AGV with the least number of tasks remaining to be completed has a higher priority. Otherwise, the AGV that is expected to complete the order task first has a higher priority. Otherwise, the AGV with the longest cumulative waiting time has a higher priority. If all the above conditions are the same, the randomly assigned AGV has a higher priority.

[0178] Optionally, as an embodiment of the present invention, the conflicts in conflict-free path planning mainly include three types: opposite conflicts, encounter conflicts and rear-end conflicts. An opposite conflict refers to a collision between two AGVs traveling in opposite directions in the same lane; an encounter conflict refers to a collision between two AGVs at the same time; a rear-end conflict refers to a situation where one AGV stops and waits, hindering the travel of another AGV following behind. Since each path in the warehouse is a two-way passage, vehicles travel along the right side of the road, so there is no opposite conflict. When two AGVs have a rear-end collision, the AGV at the rear needs to wait for the AGV in front to leave before it can travel. It is stipulated that when the AGV is at the starting point, loading point and unloading point, or is only one unit away from these three positions, the AGV is allowed to travel directly in the vertical direction for one unit and enter the grid with the same travel direction as the planned path.

[0179] In response to encounter conflicts, the present invention proposes a conflict resolution method based on AGV priority. If an encounter conflict occurs at an intersection and the waiting time of the two AGVs is the same, the AGV with a higher priority will have priority, otherwise the AGV with a longer waiting time will have priority. This ensures that the waiting time for the AGV to pass through the intersection is average and shortest due to conflict avoidance. If an encounter conflict occurs on a two-way road, when the driving direction of the AGV in the grid where the starting point or task point is located is different from the direction it is heading to the next task point, or when it is located in an adjacent grid with a driving direction opposite to the task point grid, if the target grid is not occupied by other AGVs, the priority of the AGV is set to the highest, and the AGV is allowed to shuttle through the road and enter the target grid. If the target grid is occupied by other AGVs, the AGV needs to wait for the grid to be free, and if the next target of the same AGV needs to enter this grid, the AGV has the highest priority and enters this grid first. Otherwise, according to the above-mentioned definition of AGV working status and priority, the AGV with a higher priority will have priority.

[0180] The specific steps for conflict-free scheduling based on the AGV agent's work priority and task execution order are as follows:

[0181] AGV puts the nodes in the order of execution of its own loading phase into the list one by one.

[0182] Take the first loading node in the list as the current task point, calculate the global planning path based on the topological dimension, and select the driving road;

[0183] The AGV agent detects whether a conflict occurs based on the grid dimension. If a conflict occurs, it compares the priority with the conflicting AGV agent and resolves the conflict based on the priority of the AGV agent, choosing to drive or stop and wait;

[0184] If the AGV has not reached the current loading point, it will jump to the step of whether a conflict occurs. Otherwise, the current loading node will be removed from the list. If the number of nodes in the list is not empty, it will jump to the step of taking the first loading node in the list as the current task point.

[0185] If a new order task is accepted, the execution order is replanned and the process is jumped to the step of adding the tasks to the list in sequence.

[0186] AGV puts the nodes in the order of execution of its own unloading phase into the list in sequence;

[0187] Take the first unloading node in the list as the current task point, calculate the global planning path based on the topological dimension, and select the driving road;

[0188] The AGV agent detects whether a conflict occurs based on the grid dimension. If a conflict occurs, it compares the priority with the conflicting AGV agent and resolves the conflict based on the priority of the AGV agent, choosing to drive or stop and wait;

[0189] If the AGV has not reached the current unloading node, it will jump to the step of detecting whether a conflict occurs. Otherwise, it will remove the current unloading node from the list. If the number of nodes in the list is not empty, it will jump to the step of taking the first unloading node in the list as the current task point.

[0190] If a new order task is accepted, it is placed in the buffer list and the task execution sequence is planned after all current unloading tasks are completed.

[0191] In some embodiments, the multi-load automatic tractor online dispatching system may include multiple functional modules composed of computer program segments. The computer program of each program segment in the multi-load automatic tractor online dispatching system may be stored in a memory of a computer device and executed by at least one processor to perform (see Figure 1 Description) Function of online dispatching of multi-load automatic tractors.

[0192] The system includes:

[0193] The scenario model construction module obtains the layout rules of the e-commerce smart warehouse and constructs the e-commerce smart warehouse scenario model based on the grid method and topology method combined with the layout rules;

[0194] The problem model construction module obtains online orders from e-commerce smart warehouses, builds a multi-load AGV online order scheduling problem model based on the e-commerce smart warehouse model and online orders, and obtains key parameters and objective functions based on the problem model;

[0195] The agent construction module defines agent types as order management agent, packaging station agent, AGV agent, and information management agent based on the multi-load AGV online order scheduling model. Based on the improved contract network protocol mechanism, all agents are combined to build a multi-agent system. The party that provides task allocation is regarded as the tenderer, and the party that receives and executes the task is regarded as the bidder.

[0196] The order task allocation module establishes an order packaging mechanism with the order management agent as the tenderer and the packaging station agent as the bidder. It also establishes an allocation and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder. Order task allocation is achieved based on the order packaging mechanism and the allocation and transportation mechanism.

[0197] The scheduling execution module obtains the task execution order of each AGV agent based on key parameters and objective functions, configures the work priority for each AGV agent based on the power of the AGV and the number of its load tasks, and performs conflict-free scheduling based on the work priority of the AGV agent and the task execution order.

[0198] In this embodiment, the multi-load automatic tractor online dispatching system can be divided into multiple functional modules according to the functions it performs, such as Figure 8 As shown. The functional modules of the system may include: a scenario model construction module, a problem model construction module, an agent construction module, an order task allocation module, and a scheduling execution module. A module as referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can perform fixed functions, and is stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0199] Through the scenario model construction module, the layout construction model of the e-commerce intelligent warehouse is accurately grasped. The problem model construction module combines orders to establish a scheduling model and clarify key elements. The intelligent agent construction module defines multiple intelligent agents to build an efficient multi-agent system. The order task allocation module establishes a reasonable mechanism to achieve accurate allocation. The scheduling execution module scientifically plans the task sequence and configures priorities to ensure conflict-free scheduling. It comprehensively improves the order processing efficiency of the e-commerce intelligent warehouse, optimizes resource allocation, enhances system adaptability and stability, reduces operating costs, improves customer satisfaction, provides solid support for the efficient operation of e-commerce logistics, and promotes e-commerce companies to gain an advantage in the fierce market competition.

[0200] Therefore, the technical effects that can be achieved by this embodiment can be found in the description above and will not be repeated here.

[0201] Those skilled in the art will clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general-purpose hardware platform. Based on this understanding, the technical solutions in the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or other medium that can store program code, and includes a number of instructions for enabling a computer terminal (which can be a personal computer, a server, or a second terminal, a network terminal, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention.

[0202] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the terminal embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0203] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or modules, and can be electrical, mechanical or other forms.

[0204] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of the present embodiment according to actual needs.

[0205] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0206] Although the present invention has been described in detail with reference to the accompanying drawings and in conjunction with preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, persons of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and such modifications or substitutions shall be within the scope of the present invention. Any changes or substitutions that can be easily conceived by persons skilled in the art within the technical scope disclosed in the present invention shall be within the scope of protection of the present invention.

Claims

1. A method for online dispatching of multi-load automatic tractors, characterized in that: The following steps are involved: S1, obtain the layout rules of the e-commerce smart warehouse, and build an e-commerce smart warehouse scenario model based on the grid method and topology method combined with the layout rules; S2: Obtain online orders from the e-commerce smart warehouse, build a multi-load AGV online order scheduling problem model based on the e-commerce smart warehouse model and online orders, and obtain key parameters and objective functions based on the problem model; S3, based on the multi-load AGV online order scheduling model, defines the agent types as order management agent, packaging station agent, AGV agent, and information management agent. Based on the improved contract network protocol mechanism, all agents are combined to build a multi-agent system. The party that provides task allocation is regarded as the tenderer, and the party that receives and executes the task is regarded as the bidder. S4: Establish an order packaging mechanism with the order management agent as the tenderer and the packaging station agent as the bidder. Establish an allocation and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder. Order task allocation is achieved based on the order packaging mechanism and the allocation and transportation mechanism. S5, based on key parameters and objective functions, obtains the task execution order of each AGV agent, configures the work priority of each AGV agent based on the AGV's power and the number of load tasks, and performs conflict-free scheduling based on the AGV agent's work priority and task execution order; The improved contract network protocol mechanism includes the process of bidding, tendering, bid evaluation, authorization, and contract signing. When the tendering party receives a task and needs to assign it, it will publish a message to all bidders and wait for receiving the bids from the bidders within a certain period of time; after receiving the information, the bidder will choose whether to bid according to its own situation. If it chooses to bid, it will also send a bid containing bid information for calculating the bid value; after receiving the bids, the tendering party will evaluate all the bids received, select the bidder with the largest bid value, and initiate a contract signing invitation. The two parties sign the contract to complete the task allocation from the tendering party to the bidder.

2. The online dispatching method for multi-load automatic tractors according to claim 1, characterized in that: Step S1 specifically includes: Layout rules include shelf layout rules, AGV docking area layout rules, and packaging station layout rules; Represent roads, shelves, packing tables and AGV docking areas in the warehouse environment based on the grid method; Based on the topological method, the warehouse is divided into multiple areas with the intersection as the benchmark, the center of the intersection grid as the node, and the dividing lines of the two-way roads as the edges connecting the nodes.

3. The online dispatching method for multi-load automatic tractors according to claim 1, characterized in that: The step S2 specifically includes: Configure loading points, unloading points and loads for each online order task; Combine all tasks currently undertaken by each AGV into a task group; Calculate the time cost for each AGV to complete the task group. The specific calculation is: Among them, t Gg Indicates the completion of task group G g The time cost required, g represents the g-th task group, N g represents the number of tasks contained in the g-th task group, m represents the number of AGVs, M represents the total number of transportable AGVs, n represents the task number in the task group, t0 represents the time for each operation at the loading point or unloading point, t w represents the waiting time of AGV due to conflict avoidance when executing tasks, a0, a n-1 and a n are the loading points of the 1st, nth and n+1th tasks in the task group, b0, b n-1 and b n are the unloading points of the 1st, nth, and n+1th tasks in the task group, respectively. represents the time cost between node m and node a0, Represents node a n-1 and a n The time cost between Represents node a n The time cost between and b0, Represents node b n-1 and b n The time cost between Obtain the sum of the path costs between the loading points and the unloading points of each AGV completing the task group and combine them to obtain the total path cost; The scheduling optimization objective function of each AGV is calculated based on the total time cost and path cost of each AGV completing the task group. The specific calculation is: Among them, min() represents the operation of finding the minimum value.

4. The online dispatching method for multi-load automatic tractors according to claim 1, characterized in that: In step S4, the order management agent is used as the tenderer and the packaging agent is used as the bidder to establish an order packaging mechanism, which specifically includes: When there is a new order, the order management agent is set as the bidder and a tender document is prepared, which contains the location information of the ordered goods; The packaging agent is the bidder. After receiving the bid information from the order management agent, it first determines whether the number of its currently unfinished packaging order tasks exceeds the threshold for accepting new orders. If the conditions for accepting the order are met, a bid is submitted to the tenderer's order management agent, and the remaining unfinished packaging task information and the path length indicator from the location of the ordered goods are provided. The tenderer's order management agent then calculates the bid value and evaluates the bid. The bid value calculation formula is as follows: Among them, Val1 is the bid value of the bidder's packaging agent, n1 is the number of orders currently accepted by the bidder's packaging agent, N1 is the maximum number of orders that can be received, d xx′ is the Manhattan distance between the bidder and the location of the ordered goods, d max is the maximum distance between the packing station agent and the goods, and α and β are used to adjust the weights of the number of orders undertaken and the distance to the goods; The order management agent evaluates all bidders, selects the bidder with the largest bid value, issues a winning notice, and waits for the winning bidder to confirm. If the winning bidder confirms, the contract is signed and the task assignment is completed.

5. The online dispatching method for multi-load automatic tractors according to claim 1, characterized in that: In step S4, the packaging station agent is used as the tenderer and the AGV agent is used as the bidder to establish a distribution and transportation mechanism, which specifically includes: When the packing station agent accepts the order task assigned by the order management agent, the packing station agent will act as the bidder to specify the bidding document. The bidding document information includes the location of the ordered goods and the location of the packing station. As a bidder, the AGV agent, after receiving the bid information from the packaging station agent, first determines whether its power is sufficient to accept the new task and whether the current capacity can accommodate the new order goods; If the conditions are met, a bid is initiated to the tenderer's packaging agent. The difference between the total length of the path planned after and before accepting the task, as well as the current number of accepted order tasks, is calculated as the bid information and sent to the packaging agent for calculating the bid value. The bidder's bid value calculation formula is as follows: Among them, Val2 is the bid value of the bidder's AGV intelligent body, n2 is the number of orders currently accepted by the bidder's AGV intelligent body, N2 is the maximum number of orders that can be received, l g The total length of the path planned by the bidder AGV agent before accepting the task, l g′ The total length of the path planned by the bidder's AGV agent after accepting the task, Δl max The maximum difference between the total length of the path planned by the AGV agent before and after accepting the task and the distance to the cargo, α ′ and β ′ The weight used to adjust the maximum difference between the number of orders received by the AGV agent and the total length of the path planned before and after the AGV agent accepts the task and the distance to the goods; The packaging station agent evaluates all bidders, selects the bidder with the largest bid value, issues a winning notice, and waits for the winning bidder to confirm; if the winning bidder confirms, a contract is signed, completing the task allocation from the packaging station agent to the AGV agent.

6. The online dispatching method for multi-load automatic tractors according to claim 1, characterized in that: In step S5, the task execution order of each AGV agent is obtained based on the key parameters and the objective function. Specifically, when the number of tasks currently accepted by the AGV is not greater than 6, a two-stage branch and bound algorithm based on a priority queue is used to solve the problem: S5-1, construct the adjacency matrix between the AGV's current position and all task loading points, and randomly generate a loading segment execution order as a limit, create a priority queue and put the AGV's current position and current path into the priority queue; S5-2, take the first node of the queue for expansion and calculate the current path length. If the path value of a certain extended loading point is greater than the path of the current optimal loading segment execution order, then it will not be added to the priority queue. Otherwise, it will be added to the priority queue. S5-3, if the current path already contains all loading points and the path length is less than the optimal solution of the current loading segment, then update the optimal solution of the loading segment; S5-4, if the priority queue is empty, record the optimal loading segment execution order and the last loading point, otherwise jump to S5-2; S5-5: Construct an adjacency matrix between the last loading point of the optimal loading segment execution order and all unloading points, randomly generate an unloading segment execution order as a limit, and put the last loading point and the current path into the priority queue; S5-6 takes out the node at the head of the queue, expands it, and calculates the current path length. If the path value of a certain expanded unloading point is greater than the path of the currently found optimal unloading segment execution order, then it is not added to the priority queue. Otherwise, it is added to the priority queue. S5-7 If the current path already contains all unloading points and the path length is less than the optimal solution of the current unloading segment, then update the optimal solution of the unloading segment; If the priority queue is not empty in S5-8, jump to S5-6. Otherwise, record the optimal unloading segment execution order and merge it with the optimal loading segment execution order to obtain the final task execution order.

7. The online dispatching method for multi-load automatic tractors according to claim 6, characterized in that: In step S5, the task execution order of each AGV agent is obtained based on the key parameters and the objective function. Specifically, when the number of tasks currently accepted by the AGV is greater than 6, a two-stage genetic algorithm is used to solve the problem: Initialize the genetic algorithm crossover probability, mutation probability, and population evolution generation, and generate the initial population based on the task execution order of the initial loading phase and the task execution order of the initial unloading phase as chromosomes; Based on the tournament selection method, the current population is selected to generate a new population; Perform crossover and mutation operations on individuals in the new population; Calculate the fitness value of the new population based on the parameters and objective function in the problem model, record the execution order and path length of the optimal individuals in the population, and update the global optimal execution order and path length; If the evolutionary generation reaches the preset generation, the optimal execution order and path length of the AGV are output, otherwise the crossover operation and mutation operation steps are returned.

8. The online dispatching method for multi-load automatic tractors according to claim 7, characterized in that: The step of configuring the work priority for each AGV agent based on the power consumption of the AGV and the number of its load tasks in step S5 includes: According to the power of AGV and the number of its load tasks, the working states of AGV intelligent body include idle state, risk state, loading state and unloading state; When a risky AGV conflicts with other AGVs, the risky AGV has the highest priority. When a risky AGV conflicts with another risky AGV, one of the AGVs is randomly assigned a higher priority. When an idle AGV conflicts with other AGVs, the idle AGV has the lowest priority. When an idle AGV conflicts with another idle AGV, one of the AGVs is randomly assigned a higher priority. When a conflict occurs between the loading AGV and the unloading AGV, the unloading AGV has a higher priority; When a loading AGV conflicts with another loading AGV, or an unloading AGV conflicts with another unloading AGV, the AGV with fewer remaining tasks to be completed has a higher priority. Otherwise, the AGV that is expected to complete the order task first has a higher priority. Otherwise, the AGV with a longer cumulative waiting time has a higher priority. If the above conditions are all the same, the randomly designated AGV has a higher priority.

9. The online dispatching method for multi-load automatic tractors according to claim 8, characterized in that: The specific steps for conflict-free scheduling based on the AGV agent's work priority and task execution order are as follows: The AGV puts the nodes in the order of execution of its own loading phase into the list one by one; Take the first loading node in the list as the current task point, calculate the global planning path based on the topological dimension, and select the driving road; The AGV agent detects whether a conflict occurs based on the grid dimension. If a conflict occurs, it compares the priority with the conflicting AGV agent and resolves the conflict based on the priority of the AGV agent, choosing to drive or stop and wait; If the AGV has not reached the current loading point, it will jump to the step of whether a conflict occurs. Otherwise, the current loading node will be removed from the list. If the number of nodes in the list is not empty, it will jump to the step of taking the first loading node in the list as the current task point. If a new order task is accepted, the execution order is replanned and the process is jumped to the step of adding the tasks to the list in sequence. AGV puts the nodes in the order of execution of its own unloading phase into the list in sequence; Take the first unloading node in the list as the current task point, calculate the global planning path based on the topological dimension, and select the driving road; The AGV agent detects whether a conflict occurs based on the grid dimension. If a conflict occurs, it compares the priority with the conflicting AGV agent and resolves the conflict based on the priority of the AGV agent, choosing to drive or stop and wait; If the AGV has not reached the current unloading node, it will jump to the step of detecting whether a conflict occurs. Otherwise, it will remove the current unloading node from the list. If the number of nodes in the list is not empty, it will jump to the step of taking the first unloading node in the list as the current task point. If a new order task is accepted, it is placed in the buffer list and the task execution sequence is planned after all current unloading tasks are completed.

10. A method for online dispatching of multi-load automatic tractors, characterized in that: When the system is implemented, the online dispatching method for multi-load automatic tractors according to any one of claims 1 to 7 is executed, and the system includes: The scenario model construction module obtains the layout rules of the e-commerce smart warehouse and constructs the e-commerce smart warehouse scenario model based on the grid method and topology method combined with the layout rules; The problem model construction module obtains online orders from e-commerce smart warehouses, builds a multi-load AGV online order scheduling problem model based on the e-commerce smart warehouse model and online orders, and obtains key parameters and objective functions based on the problem model; The agent construction module defines agent types as order management agent, packaging station agent, AGV agent, and information management agent based on the multi-load AGV online order scheduling model. Based on the improved contract network protocol mechanism, all agents are combined to build a multi-agent system. The party that provides task allocation is regarded as the tenderer, and the party that receives and executes the task is regarded as the bidder. The order task allocation module establishes an order packaging mechanism with the order management agent as the tenderer and the packaging station agent as the bidder. It also establishes an allocation and transportation mechanism with the packaging station agent as the tenderer and the AGV agent as the bidder. Order task allocation is achieved based on the order packaging mechanism and the allocation and transportation mechanism. The scheduling execution module obtains the task execution order of each AGV agent based on key parameters and objective functions, configures the work priority for each AGV agent based on the power of the AGV and the number of its load tasks, and performs conflict-free scheduling based on the work priority of the AGV agent and the task execution order.

Images