CN114326608B - AGV group system based on multiple agents - Google Patents
AGV group system based on multiple agents Download PDFInfo
- Publication number
- CN114326608B CN114326608B CN202111438028.XA CN202111438028A CN114326608B CN 114326608 B CN114326608 B CN 114326608B CN 202111438028 A CN202111438028 A CN 202111438028A CN 114326608 B CN114326608 B CN 114326608B
- Authority
- CN
- China
- Prior art keywords
- agv
- task
- layer
- agvs
- cost
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004891 communication Methods 0.000 claims abstract description 53
- 230000003993 interaction Effects 0.000 claims abstract description 16
- 230000007246 mechanism Effects 0.000 claims description 39
- 230000001360 synchronised effect Effects 0.000 claims description 35
- 238000013439 planning Methods 0.000 claims description 28
- 238000007726 management method Methods 0.000 claims description 24
- 238000012546 transfer Methods 0.000 claims description 22
- 230000008901 benefit Effects 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000000034 method Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 7
- 230000003068 static effect Effects 0.000 claims description 7
- 230000036541 health Effects 0.000 claims description 6
- 230000000903 blocking effect Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims 1
- 238000004364 calculation method Methods 0.000 abstract description 3
- 230000002860 competitive effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 8
- 238000012795 verification Methods 0.000 description 5
- 238000009434 installation Methods 0.000 description 3
- 230000008447 perception Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/60—Electric or hybrid propulsion means for production processes
Landscapes
- Multi Processors (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses an AGV group system based on multiple intelligent agents, which comprises at least one central intelligent agent and a plurality of AGVs which are in communication connection with each other, wherein the central intelligent agent is in communication connection with the AGVs; the central agent includes: external system communication layer, AGV communication service layer and operational decision layer, AGV includes: the system comprises an inference decision layer, a network interaction layer and a control layer, wherein the network interaction layer comprises: a central agent communication layer for with central agent communication and be used for the AGV communication layer with other AGVs communication, central agent communication layer is connected with AGV communication service layer, has weakened central agent's function, and give the more decision-making ability of AGV, makes the AGV have the optimal task cost of autonomous calculation to select optimal task with the mode of competitive bidding, independently plan the route of traveling, make entire system more intelligent, simultaneously, a plurality of AGVs can self-group, the cooperation accomplishes the task, cooperation operation ability has obtained obvious promotion.
Description
Technical Field
The invention relates to the technical field of AGV scheduling, in particular to an AGV group system based on multiple intelligent agents.
Background
At present, an AGV system adopts a mode of centralized control by an AGV dispatching system, and comprises task management, vehicle management, traffic management, path search, fault diagnosis and the like, which are all completed by the AGV dispatching system, and the AGV (AGV agent) has no self-learning, reasoning, operation and decision making capability, cannot autonomously select task execution, cannot communicate with other AGVs and the like. When the AGV is in a motion state, the parameters of the AGV are changed at all times, the task state is changed along with time, and the path cost is dynamically changed along with time, so that the task distribution result obtained by instantaneous calculation of the AGV dispatching system is not an optimal solution, and the efficiency is low. Meanwhile, when the number of AGVs is large, the information processing amount of the central intelligent agent is large, overload operation can occur, and information processing is not timely.
Disclosure of Invention
The invention aims to improve the control mode of an AGV system in the prior art, release the function of a central intelligent agent, ensure that the central intelligent agent is only responsible for the communication interface of an external system and the registration management, the task bidding, the task arbitration and the like of an internal system, namely the AGV, further endow the AGV with more decision-making capability, and the central intelligent agent is not responsible for the allocation of tasks, path planning, path resource allocation, traffic management and the like, and the functions are independently completed by the AGV intelligent agent, so that the AGV has more wisdom, calculation and decision-making capability and becomes the AGV intelligent agent, and the efficiency, the stability and the fault tolerance of the AGV system are improved.
According to a first aspect, the present invention provides a multi-agent based AGV cluster system comprising at least one central agent and a plurality of AGVs communicatively coupled to each other, the central agent being communicatively coupled to the AGVs;
The central agent comprises: the system comprises an external system communication layer, an AGV communication service layer and an operation decision layer, wherein the external system communication layer, the AGV communication service layer and the operation decision layer are connected through an internal bus;
The AGV includes: the system comprises an inference decision layer, a network interaction layer and a control layer, wherein the inference decision layer, the network interaction layer and the control layer are connected through an internal bus;
the network interaction layer comprises: the system comprises a central agent communication layer for communicating with a central agent and an AGV communication layer for communicating with other AGVs, wherein the central agent communication layer is connected with the AGV communication service layer;
The external system communication layer is used for receiving tasks issued by an external system;
The operation decision layer is used for analyzing tasks issued by the external system, generating task bidding notification sheets, and issuing the bidding notification sheets to each AGV;
the reasoning decision layer is used for calculating the task cost of the self according to the bidding notification bill;
the network interaction layer is used for notifying the task cost of the network interaction layer to a central intelligent agent and other AGVs;
The operation decision layer is also used for selecting the AGV with the lowest task cost as the bid-winning candidate AGV, marking the AGV as the bid-winning state, giving the execution authority of the bid-winning notification bill to the bid-winning candidate AGV according to the bid result, and modifying the bid-winning state into the bid-winning state;
the reasoning decision layer is also used for judging task types after the AGVs are bound with the winning-bid task, if the tasks are collaborative tasks, a collaborative operation mechanism is adopted, and the number of collaborative AGVs required by the collaborative tasks are summoned to form a collaborative operation group together;
The control layer is used for controlling the AGV to execute tasks.
Further, the task cost includes: static cost and dynamic cost; the static cost includes: theoretical time cost required by an AGV driving path, predictable acceleration and deceleration time cost of the AGV in the driving process and other time cost for executing operation; the task dynamic costs include unpredictable time costs caused by obstacles, traffic jams, AGV failure, and human intervention during the travel of the AGV.
Further, the theoretical time cost C theorypath is:
Where n is the total number of all path segments that need to be traversed, L i is the distance of each path segment, V i is the speed of each path segment, A calibration factor for each path segment.
Further, the operational decision layer includes: the system comprises a task management module and a task arbitration service module, wherein the task management module is used for analyzing tasks issued by an external system, generating task bidding notification sheets, and issuing the bidding notification sheets to each AGV;
The reasoning decision layer is also used for proposing objections to the task arbitration service module in the process of determining the task by the AGV;
The task arbitration service module is used for performing arbitration service when receiving objection of the reasoning decision layer, and deciding which AGV is used for executing the task based on a task arbitration mechanism;
The task arbitration mechanism includes:
After receiving objections of the reasoning decision layer, the task arbitration service module arbitrates according to the health degree and the operation data of each AGV, and decides the AGV executing the task or reinitiates auction;
the operation data includes: failure rate, task execution efficiency, blocking rate, operation duration, idle rate.
Further, the condition that the reasoning decision layer proposes objection is:
In the current state, the task cost of the AGV executing the task is superior to the task cost of the winning candidate AGV, or the lowest cost of the AGVs executing the task is the same, and objections are presented, otherwise, no objections are presented.
Further, each AGV contains a task list of the same active state in the current AGV group system, and the reasoning decision layer comprises: the benefit estimation module is used for calculating the task cost of the benefit estimation module according to the bidding notification sheet, and also is used for automatically calculating the task cost which is not in a binding state in the task list at regular intervals, and the task competition module is used for participating in task auction and carrying out task exchange or transfer based on a task exchange and transfer mechanism.
Further, the task exchange and transfer mechanism includes:
before the AGVs in the winning state bind with the tasks, the AGVs in the winning state exchange or transfer the tasks with other AGVs;
After the AGV in the winning state is bound with the task, the AGV and the task are not separable until the task is executed;
Task exchange occurs between two bid-winning AGVs, a first bid-winning AGV and a second bid-winning AGV, and the conditions are required to be satisfied:
C1>C12+Cswap
C2>C21+Cswap
Wherein, C 1 represents the cost of executing the bid-winning task by the first bid-winning AGV;
c 12 represents the cost of the first winning AGV to execute the second winning AGV task;
C 2 represents the cost of the second winning AGV to execute the winning task;
c 21 represents the cost of the second winning AGV to execute the first winning AGV task;
c swap represents the cost of task exchange;
Task transfer occurs between an AGV in a winning state and an AGV in an unbiased state, and the conditions are required to be satisfied:
C1>C31+Ctransfer
Wherein, C 31 represents the cost of the bid-winning AGV executing the bid-winning AGV task;
C 1 represents the cost of the winning AGV to execute the task;
C transfer represents the cost of task transfer.
Further, the inference decision layer further includes: the operation reasoning module and the path planning module;
The operation reasoning module is used for judging task types after the AGVs are marked, if the tasks are collaborative tasks, a task collaborative operation mechanism is adopted, the collaborative AGVs with the number required by the collaborative tasks are summoned, and a collaborative operation group is built;
the control layer includes: the system comprises an execution control module and a motion control module, wherein the execution control module is used for controlling the motion of an execution mechanism, and the motion control module is used for controlling the traveling of an AGV;
The collaborative operation mechanism comprises:
After the AGV is marked, the AGV is a traction AGV, and a collaborative work task group is built through task auction;
The cooperative AGVs autonomously complete path planning from the current point to the starting point by utilizing a path planning module according to an asynchronous instruction issued by the traction AGVs;
the motion control module cooperates with the AGV to move to a starting point according to the path control planned by the path planning module;
The execution control module controls the execution structure to synchronously execute loading actions according to the synchronous instruction issued by the head pulling AGV until the actions are completed;
The motion control module controls the AGV to move to a target point of the cooperative task according to a synchronous instruction issued by the head-pulling AGV;
After the cooperative work group synchronously reaches the target point, the head-pulling AGVs uniformly send synchronous instructions to the cooperative AGVs, and the execution control module controls the execution structure to synchronously execute unloading actions until the actions are completed. Further, the reasoning decision layer is further used for differentiating the synchronous instruction into a plurality of unit time instructions by taking time as a unit, and sequentially issuing the unit time instructions to the cooperative AGV;
The synchronous instruction comprises a synchronous moving instruction and a synchronous operation instruction;
The motion control module controls the AGV to move according to the synchronous movement instruction, and the execution control module controls the execution structure to act according to the synchronous operation instruction;
Each cooperative AGV reports the states of the AGVs and the executing mechanism to the traction AGVs in real time through the AGV communication layer, and the motion control module and the executing control module strictly control the AGVs to execute actions according to unit time instructions; if the state of a certain cooperative AGV is not synchronous with the state of other AGVs, the other cooperative AGVs enter a waiting state until the last AGV finishes the unit time instruction, and all AGVs start the next unit time instruction again.
Further, a path planning module of the head pulling AGV calculates the running track of each cooperative AGV based on a path planning algorithm, and generates a corresponding synchronous movement instruction based on the running track; the reasoning decision layer of the head-pulling AGV generates synchronous operation instructions of the executing mechanism on the corresponding AGV based on the cooperative task.
Compared with the prior art, the invention has the beneficial effects that:
(1) According to the AGV group system based on the multiple agents, the task cost is calculated by each AGV, and the load of the central agent is greatly reduced.
(2) The real-time task cost is still calculated after each AGV is bid, and the task can be transferred or exchanged to the AGV with lower cost before the task is unbound, so that the rationality of task allocation is ensured, and the task completion speed is improved.
(3) The AGV team formation and cooperation can be realized, and the large and heavy goods carrying capacity is obviously improved.
Drawings
FIG. 1 is a system diagram of an AGV cluster system based on multiple agents according to embodiment 1;
FIG. 2 is a system block diagram of the AGV of example 1;
FIG. 3 is a schematic workflow diagram of an AGV cluster system based on multiple agents according to example 1;
FIG. 4 is a schematic workflow diagram of an auction mode of the AGV cluster system based on multiple agents of embodiment 1;
FIG. 5 is a flow chart of the collaborative operation mechanism in embodiment 1;
FIG. 6 is a diagram showing a synchronous movement command of the collaborative job team collaborative movement in embodiment 1;
FIG. 7 is a schematic diagram of a synchronous operation instruction of the collaborative job team collaborative operation in embodiment 1;
FIG. 8 is a system block diagram of a central agent in example 2;
FIG. 9 is a system block diagram of an AGV of example 2.
Detailed Description
The invention will be described in further detail below with reference to the drawings by means of specific embodiments.
Example 1
1-2, The present invention provides a multi-agent based AGV cluster system comprising: the central intelligent body, a plurality of AGVs based on 5G communication technology intercommunication connection, every AGV all is connected with the central intelligent body. The central agent includes: the external system communication layer, the AGV communication service layer and the operation decision layer are connected through an internal bus. The operation decision layer comprises: the task management module is used for analyzing tasks issued by the external system, generating task bidding notification sheets, and issuing the bidding notification sheets to each AGV through the AGV communication service layer. The task arbitration service module is used for performing arbitration service when receiving objection of the reasoning decision layer, and deciding which AGV is used for executing the task based on a task arbitration mechanism.
The AGV includes: the reasoning decision layer, the network interaction layer and the control layer are connected through an internal bus; wherein, the network interaction layer includes: central authorities agent communication layer and AGV communication layer, central authorities agent communication layer and AGV communication service layer are connected to realize the communication connection of central authorities agent and AGV, two AGVs pass through AGV communication layer communication connection, in order to realize the mutual communication connection of AGV. The reasoning decision layer comprises: the system comprises an operation reasoning module, a benefit estimation module, a path planning module and a task competition module. The benefit estimation module is used for calculating the task cost of the benefit estimation module according to the bidding notification bill and is also used for automatically calculating the task cost which is not in a binding state in the task list at regular intervals. The path planning module is used for autonomously completing path planning from the current point to the starting point. The task competition module is used for participating in task auction and carrying out task exchange or transfer based on a task exchange and transfer mechanism. The control layer comprises: the system comprises an execution control module and a motion control module, wherein the execution control module is used for controlling the motion of an executing mechanism, and the motion control module is used for controlling the traveling of the AGV. The operation reasoning module is used for judging task types after the AGVs are marked, if the tasks are collaborative tasks, a task collaborative operation mechanism is adopted, the number of collaborative AGV intelligent agents required by the collaborative tasks are summoned, and a collaborative operation group is built.
Specifically, as shown in fig. 3-5, the flow of the system is as follows:
After the external system communication layer receives the task issued by the external system, the operation decision layer analyzes the task issued by the external system and issues the task to the AGV according to the task allocation mode. The task allocation mode includes: a designated mode and an auction mode, wherein the designated mode is that a central agent designates an AGV for a task to execute, and after the task is designated, the designated AGV judges the task category, and if the task is an independent task, the task is completed independently; if the tasks are collaborative tasks, the AGVs serve as head-pulling AGVs, a collaborative operation mechanism is adopted by an inference decision-making layer, the number of collaborative AGVs required by the collaborative tasks are summoned, a collaborative operation group is formed together, and the tasks are executed together.
In the auction mode, the operation decision layer generates a task bidding notice and issues the bidding notice to each AGV through the AGV communication service layer. The bid notification sheet includes: task number, task attribute, task priority, number of AGVs required, task start time, operation list; the operation list includes a plurality of operations, each operation including: operating station, opcode attribute; the opcode attributes include: the type of the operation code, the lifting height and the lifting speed of the executing mechanism; the operation code types include: asynchronous operation and synchronous operation.
Each AGV participates in the auction through the task competition module. The benefit estimation module analyzes the bidding notification ticket, judges the legitimacy of the order, calculates the task cost of executing the task under the current state, and broadcasts the task cost to a central intelligent agent and other AGVs through a network interaction layer, namely the AGVs bid in a mode of disclosing the base price.
The task cost of the AGV is derived from a combination of static and dynamic costs. Specifically, the static costs include the theoretical time cost required by the AGV to travel the path after the AGV completes the path planning, the predictable acceleration and deceleration time cost of the AGV during travel, and the time cost spent performing the operation. Wherein, the theoretical time cost is:
Where n is the total number of all path segments that need to be traversed, L i is the distance of each path segment, V i is the speed of each path segment, A calibration factor for each path segment.
The task dynamic cost comprises unpredictable time cost caused by obstacles, traffic jams, AGV faults, human intervention and the like in the running process of the AGVs, the task dynamic cost is mostly unpredictable cost, and the task dynamic cost prediction is accumulated by referring to the average time spent by each AGV at a certain node. Such as: traffic congestion takes an average of two minutes at a node, and the task's dynamic cost will increase by two minutes when the task's AGV encounters a traffic congestion at that node. Human intervention takes three minutes on average in a certain path segment, and when the AGV executing the task encounters human intervention in that path segment, the task's dynamic costs will accumulate for three minutes. And the task dynamic cost is predicted by accumulation in the same way.
Under the condition of small interference of external conditions, the task cost is a decisive factor by the static cost, the dynamic cost is used as an auxiliary factor, more dynamic cost is referred in the movement process, and the task cost is predicted according to the situations of obstacles, traffic jams, AGV faults, human intervention and the like reported by the AGVs and is judged in real time, and the task execution cost is used as the auction cost.
And after the operation decision layer receives the quotation of each AGV, the AGV with the lowest cost is selected to become a winning bid candidate AGV, and the winning bid candidate AGV is marked as a pre-winning bid state. AGVs that become winning candidate AGVs broadcast a pre-winning notification to other AGVs and to the central agent at the same time. In this process, the status of the AGV is updated in real time, and the AGV that receives the bid-in-progress notification in principle agrees to the bid-in-progress task. However, when the task cost of executing the task in the current state of the AGV calculated by the benefit estimation module is better than the task cost of the winning candidate AGV or the lowest cost of executing the task by a plurality of AGVs, the inference decision layer presents objections to the task arbitration service module, and simultaneously notifies the central agent and the winning candidate AGV, and the task arbitration service module arbitrates according to the health degree and the operation data (failure rate, task execution efficiency, blocking rate, operation duration and idle rate) of each AGV after receiving the objections of the inference decision layer, so as to determine the AGV executing the task or reinitiate the secondary auction of the task. If no objection exists, the execution authority of the bid competition notice is given to the bid-winning candidate AGV by the task management module, and the bid-winning state is modified to be the bid-winning state.
Each AGV comprises a task order list with the same active state in the current AGV group system, and the benefit estimation module of each AGV calculates the task cost of the unbound task in the task list at regular intervals.
Before the AGVs in the winning state bind with the corresponding tasks, if the cost of executing the tasks by any AGV is lower, the task transfer is carried out if the following conditions are met:
C1>C31+Ctransfer
Wherein, C 31 represents the cost of the bid-winning AGV executing the bid-winning AGV task; c 1 represents the cost of the winning AGV to execute the task; c transfer represents the cost of task transfer.
In addition, before the task is bound, if two AGVs in the winning state exchange the winning task, the saved time cost is greater than the time cost loss caused by task exchange, namely, the exchange is performed under the following conditions:
C1>C12+Cswap
C2>C21+Cswap
Wherein, C 1 represents the cost of executing the bid-winning task by the first bid-winning AGV; c 12 represents the cost of the first winning AGV to execute the second winning AGV task; c 2 represents the cost of the second winning AGV to execute the winning task; c 21 represents the cost of the second winning AGV to execute the first winning AGV task; c swap represents the cost of task exchange.
For example, if the AGV1 wins task 1 and AGV2 wins task 4, the task list at this time is shown in table 1. The method comprises the following steps of marking a task 1 in an AGV1, marking a task 2 in advance, marking one task at most and marking one task in advance by the AGV, marking a task 3 by the AGV3, marking a task 4 by the AGV2, and marking a task 5 in advance.
TABLE 1
After a period of time, task 1 is completed after execution by AGV1, and pre-bid-winning task 2 of AGV1 is converted to a bid-winning task, as shown in Table 2 below.
TABLE 2
Before the AGV1 and the task 2 are bound, that is, before the AGV reaches the operation point, the task 3 is not bound with the AGV3, at this time, as shown in tables 3 and 4, since the cost 13 of the AGV1 for executing the task 3 is lower than the cost 16 of the AGV3 for executing the task 3, the cost of the AGV3 for executing the task 2 is 8, and is lower than the cost 15 of the AGV1 for executing the task 2, and the difference of the two costs is greater than the time loss caused by task exchange, at this time, any of the tasks of the AGV1 and the AGV3 can be directly exchanged, and the updated task table is shown in Table 5.
TABLE 3 Table 3
TABLE 4 Table 4
TABLE 5
After the AGV in the winning bid state is bound with the corresponding task, the operation reasoning module judges the task category, and if the AGV is an independent task, the task is completed independently. If the task is a cooperative task, the AGVs in the winning state act as head-pulling AGVs, the reasoning decision-making layer adopts a cooperative operation mechanism to summon the cooperative AGVs in the number required by the cooperative task, and form a cooperative operation group together to execute the task together.
Specifically, the collaborative job mechanism includes:
The reasoning decision layer of the head-pulling AGVs generates N-1 tasks according to the number N of the AGVs required in the collaborative tasks; and the required number of cooperative AGVs are summoned in a task broadcast bidding mode, the idle AGVs participate in bidding through a task competition module, each task performs bidding once, and finally summoned to N-1 cooperative AGVs, and the cooperative AGVs and the head-pulling AGVs form a cooperative operation group with the number of N.
After the collaborative operation group is formed, the traction AGVs issue asynchronous instructions to all collaborative AGVs through the AGV communication service layer, the asynchronous instructions comprise starting point positions, each collaborative AGV path planning module performs path planning according to the starting point positions and the positions of the current AGVs, and the motion control module controls the collaborative AGVs to move to the starting point according to the paths planned by the path planning module.
After all members of the collaborative operation team are in place, the self state is reported to the traction AGVs through the AGV communication layer, and after the traction AGVs receive the in-place states of all collaborative AGVs, the traction AGVs uniformly issue synchronous operation instructions to the collaborative AGVs, and each collaborative AGVs execute the control module to control the executing mechanism to synchronously execute loading actions until the actions are completed.
And then the traction AGVs issue synchronous movement instructions, and the movement control modules of the cooperative AGVs control the AGVs to move to the target points of the cooperative tasks. After reaching the target point, each cooperative AGV execution control module controls the execution mechanism to synchronously execute unloading actions until the actions are completed.
And the synchronous operation instruction is generated by an inference decision layer of the head AGV according to the collaborative task. And the synchronous movement instruction is calculated by a path planning module of the head-pulling AGV based on a path planning algorithm, and a corresponding synchronous movement instruction is generated based on the path planning algorithm. The path planning algorithm adopts the current mainstream path planning algorithm, such as Dijkstra algorithm, a-x algorithm, optimal priority searching algorithm and the like. The synchronous movement instruction comprises an AGV walking distance, a walking speed and a course angle and is used for controlling the AGV to travel. The synchronous operation instructions comprise an executing mechanism running speed and a lifting height and are used for controlling the executing mechanism on the AGV to work (namely lifting cargoes).
Specifically, the operation reasoning module of the head-pulling AGV differentiates the synchronous instruction into a plurality of unit time instructions by taking time as a unit, and issues the unit time instructions to the cooperative AGV through the AGV communication layer, namely, the head-pulling AGV issues the unit time instructions to the cooperative AGV in sequence according to the instruction type.
6-7, Each unit time instruction execution time is one cycle, and each cooperative AGV reports the states of itself and the executing mechanism to the head AGV in real time and executes actions strictly according to the unit time instruction. After all members complete the respective unit time instruction in one period, the head-pulling AGV issues the next unit time instruction. If the state of a certain cooperative AGV is not synchronous with the state of other AGVs and exceeds a set threshold value, the other cooperative AGVs enter a waiting state, the head-pulling AGVs issue a next unit time instruction until the last AGV finishes the unit time instruction, the head-pulling AGVs report to a central intelligent agent and the other AGVs in cooperation, the cooperative operation team is broken, and each AGV can be used as an independent individual auction task again.
Example 2
8-9, The AGV further includes, based on embodiment 1: the I/O layer is used for controlling the input and output of AGV digital signals or analog signals. The perception layer is used for communication connection and data interaction between AGV and the sensor. The sensor comprises a laser radar, a vision sensor, an electromagnetic sensor, a magnetic tape sensor, a photoelectric sensor and a safety protection sensor, and is used for collecting external environment data and realizing external environment perception; the sensor also comprises an encoder, a gyroscope and an inclination angle touch sensor, and is used for sensing the state change of the sensor.
The reasoning decision layer also comprises a traffic cooperation module and a private knowledge base, and the operation reasoning module is also used for learning the environment, constructing a map, estimating the health state and managing the private knowledge base, and reporting the public knowledge to the central agent based on a mechanism of converting the private knowledge into the public knowledge to convert the private knowledge into the public knowledge. The traffic cooperation module is used for calculating and distributing idle path resources of the paths and avoiding traffic with other AGVs. Specifically, after the path planning module autonomously completes path planning from the current point to the target point, the traffic coordination module calculates idle path resources according to the path planning and the running state data of other AGVs, wherein the path resources comprise path points and path segments; the walkable path points, segments are then added to the occupied point, segment list, marked as occupied path resources, and broadcast to other AGVs. The traffic collaboration module releases the path resources after walking by itself, so that the occupied path resources are converted into idle path resources. The running state data includes: position, attitude, travel speed, target point, occupied point, segment list. The private knowledge base is used for storing an environment map, a path map, control parameters, installation parameters and operation data of the AGV intelligent agent.
The central intelligent agent also comprises a knowledge management layer which is respectively connected with an external system communication layer, an AGV communication service layer and an operation decision layer through an internal bus. The knowledge management layer is used for sharing knowledge management, AGV operation data statistical analysis and deciding the conversion from the private knowledge to the public knowledge based on the conversion mechanism from the private knowledge to the public knowledge. The AGV operation data includes: fault rate of AGV, task execution efficiency, blocking rate, operation duration, and idle rate. The operation decision layer also comprises an AGV management module, wherein the AGV management module is used for registering, deregistering and executing health degree management on the AG and the tasks. Specifically, the knowledge management layer includes: the system comprises a learning reasoning module, a knowledge management module and a shared knowledge base; the learning reasoning module is an AI module for self-learning and self-optimizing of the central agent and is responsible for reasoning, generalizing and summarizing knowledge. The knowledge management module is used for managing the shared knowledge base, reporting a certain private knowledge of the AGVs to the AGVs, and applying the private knowledge as the auditing and approval of the common knowledge of all the AGVs; the shared knowledge base includes: knowledge of all AGVs and knowledge of the central agent.
Specifically, the private knowledge is converted into public knowledge based on a mechanism for converting the private knowledge into the public knowledge, and the mechanism for converting the private knowledge into the public knowledge comprises the following steps:
the AGV applies for a central intelligent agent to convert a certain private knowledge into public knowledge;
Then the knowledge management layer of the central intelligent agent evaluates the private knowledge of the application and sends the private knowledge to other AGVs for evaluation;
Other AGVs verify the assessment knowledge issued by the knowledge management layer and report the verification result to the central agent; the verification comprises the steps of modifying or adding self-related data according to knowledge to be evaluated, and performing actual operation verification, wherein if no abnormality exists, the verification is considered to pass;
and after all AGVs pass verification, converting the AGVs into public knowledge.
The self-adaptive Automatic Guided Vehicle (AGV) intelligent agent comprises an AGV intelligent agent body, and further comprises AGV continuous learning of a new environment and generation of a navigation environment map and the like, wherein the self-adaptive automatic guided vehicle intelligent agent body comprises sensor installation parameters, driving wheel installation parameters, driving PID control parameters, steering PID control parameters and other AGV basic parameters. Public knowledge includes: the environment map is used for all AGVs and central intelligent bodies, and has universal applicability.
The operational decision layer further comprises: the AGV management module is used for registering, logging out and executing health degree management on tasks of the AGV.
The control layer also comprises a navigation module, and the navigation module is used for identifying the environment, constructing a map and updating the map based on the environment data collected by the perception layer. In addition, the position and the posture of the AGV in the map are calculated, and the positioning of the AGV in the map and the navigation and the guidance of the AGV are realized.
Claims (9)
1. An AGV group system based on multiple agents comprises at least one central agent and a plurality of AGVs in communication connection with each other, wherein the central agent is in communication connection with the AGVs; it is characterized in that the method comprises the steps of,
The central agent comprises: the system comprises an external system communication layer, an AGV communication service layer and an operation decision layer, wherein the external system communication layer, the AGV communication service layer and the operation decision layer are connected through an internal bus;
The AGV includes: the system comprises an inference decision layer, a network interaction layer and a control layer, wherein the inference decision layer, the network interaction layer and the control layer are connected through an internal bus;
the network interaction layer comprises: the system comprises a central agent communication layer for communicating with a central agent and an AGV communication layer for communicating with other AGVs, wherein the central agent communication layer is connected with the AGV communication service layer;
The external system communication layer is used for receiving tasks issued by an external system;
The operation decision layer is used for analyzing tasks issued by the external system, generating task bidding notification sheets, and issuing the bidding notification sheets to each AGV;
the reasoning decision layer is used for calculating the task cost of the self according to the bidding notification bill;
the network interaction layer is used for notifying the task cost of the network interaction layer to a central intelligent agent and other AGVs;
The operation decision layer is also used for selecting the AGV with the lowest task cost as the bid-winning candidate AGV, marking the AGV as the bid-winning state, giving the execution authority of the bid-winning notification bill to the bid-winning candidate AGV according to the bid result, and modifying the bid-winning state into the bid-winning state;
the reasoning decision layer is also used for judging task types after the AGVs are bound with the winning-bid task, if the tasks are collaborative tasks, a collaborative operation mechanism is adopted, and the number of collaborative AGVs required by the collaborative tasks are summoned to form a collaborative operation group together;
the control layer is used for controlling the motion of the AGV, controlling an executing mechanism and serving navigation;
The task cost includes: static cost and dynamic cost; the static cost includes: theoretical time cost required by an AGV driving path, predictable acceleration and deceleration time cost of the AGV in the driving process and other time cost for executing operation; the task dynamic costs include unpredictable time costs caused by obstacles, traffic jams, AGV failure, and human intervention during the travel of the AGV.
2. The system of claim 1, wherein the theoretical time cost C theorypath is:
Where n is the total number of all path segments that need to be traversed, L i is the distance of each path segment, V i is the speed of each path segment, A calibration factor for each path segment.
3. The system of claim 1, wherein the operational decision layer comprises: the system comprises a task management module and a task arbitration service module, wherein the task management module is used for analyzing tasks issued by an external system, generating task bidding notification sheets, and issuing the bidding notification sheets to each AGV;
The reasoning decision layer is also used for proposing objections to the task arbitration service module in the process of determining the task by the AGV;
The task arbitration service module is used for performing arbitration service when receiving objection of the reasoning decision layer, and deciding which AGV is used for executing the task based on a task arbitration mechanism;
The task arbitration mechanism includes:
After receiving objections of the reasoning decision layer, the task arbitration service module arbitrates according to the health degree and the operation data of each AGV, and decides the AGV executing the task or reinitiates auction;
the operation data includes: failure rate, task execution efficiency, blocking rate, operation duration, idle rate.
4. The system of claim 3, wherein the inference decision layer proposes objection to the condition that:
In the current state, the task cost of the AGV executing the task is superior to the task cost of the winning candidate AGV, or the lowest cost of the AGVs executing the task is the same, and objections are presented, otherwise, no objections are presented.
5. The system of claim 1 wherein each AGV includes a task list of the same current AGV cluster system activity status, the inference decision layer comprising: the benefit estimation module is used for calculating the task cost of the benefit estimation module according to the bidding notification bill and also is used for automatically calculating the task cost which is not in a binding state in the task list at regular intervals;
The task competition module is used for participating in task auction and carrying out task exchange or transfer based on a task exchange and transfer mechanism.
6. The system of claim 5, wherein the task exchange and transfer mechanism comprises:
before the AGVs in the winning state bind with the tasks, the AGVs in the winning state exchange or transfer the tasks with other AGVs;
After the AGV in the winning state is bound with the task, the AGV and the task are not separable until the task is executed;
Task exchange occurs between two bid-winning AGVs, a first bid-winning AGV and a second bid-winning AGV, and the conditions are required to be satisfied:
C1>C12+Cswap
C2>C21+Cswap
Wherein, C 1 represents the cost of executing the bid-winning task by the first bid-winning AGV;
c 12 represents the cost of the first winning AGV to execute the second winning AGV task;
C 2 represents the cost of the second winning AGV to execute the winning task;
c 21 represents the cost of the second winning AGV to execute the first winning AGV task;
c swap represents the cost of task exchange;
Task transfer occurs between an AGV in a winning state and an AGV in an unbiased state, and the conditions are required to be satisfied:
C1>C31+Ctransfer
Wherein, C 31 represents the cost of the bid-winning AGV executing the bid-winning AGV task;
C 1 represents the cost of the winning AGV to execute the task;
C transfer represents the cost of task transfer.
7. The system of claim 5, wherein the inference decision layer further comprises: the operation reasoning module and the path planning module;
The operation reasoning module is used for judging task types after the AGVs are marked, if the tasks are collaborative tasks, a task collaborative operation mechanism is adopted, the collaborative AGVs with the number required by the collaborative tasks are summoned, and a collaborative operation group is built;
the control layer includes: the system comprises an execution control module and a motion control module, wherein the execution control module is used for controlling the motion of an execution mechanism, and the motion control module is used for controlling the traveling of an AGV;
The collaborative operation mechanism comprises:
After the AGV is marked, the AGV is a traction AGV, and a collaborative work task group is built through task auction;
The cooperative AGVs autonomously complete path planning from the current point to the starting point by utilizing a path planning module according to an asynchronous instruction issued by the traction AGVs;
The motion control module cooperates with the AGV to move to a task starting point according to the path control planned by the path planning module;
The execution control module controls the execution structure to synchronously execute loading actions according to the synchronous instruction issued by the head pulling AGV until the actions are completed;
The motion control module controls the AGV to move to a target point of the cooperative task according to a synchronous instruction issued by the head-pulling AGV;
After the cooperative work group synchronously reaches the target point, the head-pulling AGVs uniformly send synchronous instructions to the cooperative AGVs, and the execution control module controls the execution structure to synchronously execute unloading actions until the actions are completed.
8. The system of claim 7 wherein said inference decision layer is further configured to differentiate said synchronization command in units of time into a plurality of unit time commands and to issue unit time commands to a cooperating AGV in sequence;
The synchronous instruction comprises a synchronous moving instruction and a synchronous operation instruction;
The motion control module controls the AGV to move according to the synchronous movement instruction;
The execution control module controls the execution structure to act according to the synchronous operation instruction;
Each cooperative AGV reports the states of the AGVs and the executing mechanism to the traction AGVs in real time through the AGV communication layer, and the motion control module and the executing control module strictly control the AGVs to execute actions according to unit time instructions; if the state of a certain cooperative AGV is not synchronous with the state of other AGVs, the other cooperative AGVs enter a waiting state until the last AGV finishes the unit time instruction, and all AGVs start the next unit time instruction again.
9. The system of claim 8 wherein the path planning module of the lead AGV calculates travel trajectories for each coordinated AGV based on a path planning algorithm and generates corresponding synchronous movement instructions based on the travel trajectories; the reasoning decision layer of the head-pulling AGV generates synchronous operation instructions of the executing mechanism on the corresponding AGV based on the cooperative task.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111438028.XA CN114326608B (en) | 2021-11-30 | 2021-11-30 | AGV group system based on multiple agents |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111438028.XA CN114326608B (en) | 2021-11-30 | 2021-11-30 | AGV group system based on multiple agents |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114326608A CN114326608A (en) | 2022-04-12 |
| CN114326608B true CN114326608B (en) | 2024-05-31 |
Family
ID=81046814
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111438028.XA Active CN114326608B (en) | 2021-11-30 | 2021-11-30 | AGV group system based on multiple agents |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114326608B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115167381A (en) * | 2022-05-06 | 2022-10-11 | 中国科学院沈阳自动化研究所 | AGV cooperative work group implementation method based on temporary multi-group small wireless network |
| CN115049346B (en) * | 2022-08-16 | 2022-11-18 | 成都秦川物联网科技股份有限公司 | Industrial Internet of things system based on material transportation obstacle recognition and control method |
| CN116501063B (en) * | 2022-11-22 | 2024-09-20 | 山东卓越精工集团有限公司 | AGV dolly control system based on concrete transportation platform |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5280431A (en) * | 1985-08-30 | 1994-01-18 | Texas Instruments Incorporated | Method for controlling the movements of a mobile robot in a multiple node factory |
| CN105740979A (en) * | 2016-01-29 | 2016-07-06 | 上海海事大学 | Intelligent dispatching system and method for multi-AGV (Automatic Guided Vehicle) of automatic container terminal |
| CN106249738A (en) * | 2016-07-19 | 2016-12-21 | 南京航空航天大学 | A kind of AGV Contents in brief Intelligent Dynamic Scheduling method under workshop condition |
| CN107111307A (en) * | 2014-11-11 | 2017-08-29 | X开发有限责任公司 | Dynamically maintaining a map of a fleet of robotic devices in an environment to facilitate robotic actions |
| CN107657364A (en) * | 2017-09-06 | 2018-02-02 | 中南大学 | A kind of overloading AGV tasks towards tobacco plant material transportation distribute forming method |
| CN108910630A (en) * | 2018-07-06 | 2018-11-30 | 永大电梯设备（中国）有限公司 | Multiple lift control system and method based on multiple agent competitive way |
| CN110264120A (en) * | 2019-05-06 | 2019-09-20 | 盐城品迅智能科技服务有限公司 | A kind of intelligent storage route planning system and method based on more AGV |
| CN110989570A (en) * | 2019-10-15 | 2020-04-10 | 浙江工业大学 | Multi-AGV anti-collision collaborative path planning method |
| CN111596658A (en) * | 2020-05-11 | 2020-08-28 | 东莞理工学院 | Multi-AGV collision-free operation path planning method and scheduling system |
| CN112835333A (en) * | 2020-12-31 | 2021-05-25 | 北京工商大学 | Multi-AGV obstacle avoidance and path planning method and system based on deep reinforcement learning |
| CN114330831A (en) * | 2021-11-30 | 2022-04-12 | 云南昆船智能装备有限公司 | AGV scheduling method based on task bidding mechanism and storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100461869B1 (en) * | 2002-07-04 | 2004-12-14 | 삼성전자주식회사 | Control method of conveying system |
| GB201712256D0 (en) * | 2017-07-31 | 2017-09-13 | Robocoaster Ltd | Automated guided vehicles |
-
2021
- 2021-11-30 CN CN202111438028.XA patent/CN114326608B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5280431A (en) * | 1985-08-30 | 1994-01-18 | Texas Instruments Incorporated | Method for controlling the movements of a mobile robot in a multiple node factory |
| CN107111307A (en) * | 2014-11-11 | 2017-08-29 | X开发有限责任公司 | Dynamically maintaining a map of a fleet of robotic devices in an environment to facilitate robotic actions |
| CN105740979A (en) * | 2016-01-29 | 2016-07-06 | 上海海事大学 | Intelligent dispatching system and method for multi-AGV (Automatic Guided Vehicle) of automatic container terminal |
| CN106249738A (en) * | 2016-07-19 | 2016-12-21 | 南京航空航天大学 | A kind of AGV Contents in brief Intelligent Dynamic Scheduling method under workshop condition |
| CN107657364A (en) * | 2017-09-06 | 2018-02-02 | 中南大学 | A kind of overloading AGV tasks towards tobacco plant material transportation distribute forming method |
| CN108910630A (en) * | 2018-07-06 | 2018-11-30 | 永大电梯设备（中国）有限公司 | Multiple lift control system and method based on multiple agent competitive way |
| CN110264120A (en) * | 2019-05-06 | 2019-09-20 | 盐城品迅智能科技服务有限公司 | A kind of intelligent storage route planning system and method based on more AGV |
| CN110989570A (en) * | 2019-10-15 | 2020-04-10 | 浙江工业大学 | Multi-AGV anti-collision collaborative path planning method |
| CN111596658A (en) * | 2020-05-11 | 2020-08-28 | 东莞理工学院 | Multi-AGV collision-free operation path planning method and scheduling system |
| CN112835333A (en) * | 2020-12-31 | 2021-05-25 | 北京工商大学 | Multi-AGV obstacle avoidance and path planning method and system based on deep reinforcement learning |
| CN114330831A (en) * | 2021-11-30 | 2022-04-12 | 云南昆船智能装备有限公司 | AGV scheduling method based on task bidding mechanism and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114326608A (en) | 2022-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114326608B (en) | AGV group system based on multiple agents | |
| CN106251016B (en) | A kind of parking system paths planning method based on dynamic time windows | |
| WO2021248607A1 (en) | Deep reinforcement learning-based taxi dispatching method and system | |
| CN109532960B (en) | Automatic dispatching system for marshalling station | |
| CN105740979B (en) | Intelligent dispatching system and method for multiple automatic guided vehicles of automatic wharf | |
| RU2589869C2 (en) | Method and system for efficient scheduling for plurality of automated nonholonomic vehicles using scheduler of coordinated routes | |
| CN111638717B (en) | Design method of traffic coordination mechanism of distributed autonomous robot | |
| CN111105141B (en) | Demand response type bus scheduling method | |
| CN104021667A (en) | Co-hiring scheduling system and scheduling method integrating reservation service and real-time taxi-hiring | |
| CN103130050B (en) | A kind of dispatching method of multiple lift control system | |
| CN114723125A (en) | Intercity vehicle order allocation method combining deep learning and multi-task optimization | |
| Sun et al. | AGV-based vehicle transportation in automated container terminals: A survey | |
| CN114911255A (en) | Heterogeneous multi-unmanned aerial vehicle collaborative track planning method for communication relay guarantee | |
| Yoshihara et al. | Non-cooperative optimization of charging scheduling of electric vehicle via stackelberg game | |
| CN114240002A (en) | Bus departure timetable dynamic optimization algorithm based on deep reinforcement learning | |
| CN116400651A (en) | Multi-AGV cooperative scheduling method and device for intelligent factory digital twin platform | |
| Gong et al. | Toward safe and efficient human–swarm collaboration: A hierarchical multi-agent pickup and delivery framework | |
| CN116993255A (en) | AGVS dynamic collision-free path planning method based on multi-agent simulation | |
| Xu et al. | Dynamic spare point application based coordination strategy for multi-AGV systems in a WIP warehouse environment | |
| He et al. | A microscopic agent-based simulation for real-time dispatching problem of a busy railway passenger station | |
| Jégou et al. | A contract net based intelligent agent system for solving the reactive hoist scheduling problem | |
| CN115220461B (en) | Robot single system and multi-robot interaction cooperation method in indoor complex environment | |
| Žužek et al. | Simulation-Based Approach for Automatic Roadmap Design in Multi-AGV Systems | |
| CN116719282A (en) | Workshop integrated dynamic scheduling method considering path planning and disturbance response | |
| CN114936452B (en) | Modeling and solving method based on multi-layer AGV parking garage parking and taking system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |