CN112070383B - Dynamic task-oriented multi-agent distributed task allocation method - Google Patents

Dynamic task-oriented multi-agent distributed task allocation method Download PDF

Info

Publication number
CN112070383B
CN112070383B CN202010898123.7A CN202010898123A CN112070383B CN 112070383 B CN112070383 B CN 112070383B CN 202010898123 A CN202010898123 A CN 202010898123A CN 112070383 B CN112070383 B CN 112070383B
Authority
CN
China
Prior art keywords
task
agent
value
tasks
manager
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
Application number
CN202010898123.7A
Other languages
Chinese (zh)
Other versions
CN112070383A (en
Inventor
辛斌
丁玉隆
陈杰
方浩
杜鑫
张昊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Institute of Technology BIT
Original Assignee
Beijing Institute of Technology BIT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Institute of Technology BIT filed Critical Beijing Institute of Technology BIT
Priority to CN202010898123.7A priority Critical patent/CN112070383B/en
Publication of CN112070383A publication Critical patent/CN112070383A/en
Application granted granted Critical
Publication of CN112070383B publication Critical patent/CN112070383B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06316—Sequencing of tasks or work
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/10—Office automation; Time management
    • G06Q10/101—Collaborative creation, e.g. joint development of products or services

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Educational Administration (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Computer And Data Communications (AREA)

Abstract

The invention provides a multi-agent distributed task allocation method for dynamic tasks, which takes the dynamic performance of the tasks into consideration, utilizes an auction-call algorithm to form a pre-allocation scheme, has strong practicability, can be used for completing tasks such as coordinated multipoint aggregation, coordinated multi-target reconnaissance, coordinated multi-target capture and the like, ensures that the decision among agents is free from conflict and the task completion efficiency is effectively ensured, and the task allocation rate reaches 100%; in addition, the behavior and the communication of the intelligent agent are based on a unified framework, the intelligent agent can be added or deleted based on the framework, the robustness of the whole system is improved, the intelligent agent is suitable for being damaged, and the scenes of the intelligent agent can be increased at any time.

Description

Dynamic task-oriented multi-agent distributed task allocation method
Technical Field
The invention belongs to the technical field of multi-agent task allocation, and particularly relates to a dynamic task-oriented multi-agent distributed task allocation method.
Background
With the development of theories such as unmanned systems and the like, the performance of intelligent agents is improved, and the dynamic tasks (such as forest fire fighting) which are highly dangerous, have dispersed task points, are time-sensitive and contain various uncertainties by using multiple intelligent agents become the future trend. In forest fire fighting missions, the fire (intensity of the mission) may change over time due to natural or man-made factors. In the initial stage of fire, one intelligent agent can extinguish the fire source, but the fire intensity is increased along with the increase of the fire, so that the requirement of complex fire fighting tasks cannot be met by a single intelligent agent, and the given fire fighting tasks need to be completed by utilizing a plurality of intelligent agents together. The cooperation of the multiple intelligent agents can effectively improve the rescue efficiency and reduce the required number and casualties of rescuers.
From the solution mechanism, the multi-agent task allocation can be divided into centralized type and distributed type. In the centralized distribution, one robot is used as a central planner, obtains the current task and the environmental state through communication with other robots, and then performs global task distribution. The method has good global characteristics, but relies on a single central planner. There is no central planner in the distributed task allocation. Each intelligent agent autonomously selects and decides tasks through self perception of the intelligent agents to the environmental information, and a plurality of intelligent agents form a task allocation scheme through mutual negotiation. Common distributed task allocation methods include a market mechanism-based method, an idle chain-based method, a threshold response method, and the like. The distributed decision is suitable for parallel computation, has better expansibility and robustness, and is suitable for a larger-scale system.
However, in the existing distributed task allocation methods, task modeling mostly assumes that the intensity of the task does not change, and does not consider the situation that the intensity of the task changes along with time.
Disclosure of Invention
In order to solve the problems, the invention provides a multi-agent distributed task allocation method for dynamic tasks, which can effectively and quickly allocate a plurality of target tasks and enable the dynamic tasks to be completed as soon as possible.
A multi-agent distributed task allocation method for dynamic tasks, which allocates distributed tasks for multi-agents based on an auction-levying algorithm, comprises the following steps:
s1: respectively taking each agent as a current agent to execute the following operations:
obtaining the bid value of the current agent i to each uncompleted task, and taking the uncompleted task corresponding to the maximum bid value as the task j which is best at the current agent i*And constructing a bidding vector A corresponding to the current agent i according to the bidding vectors A(i)＝[i,j*,fi(j*)]Then the bid vector A is added(i)Broadcast to other agents, wherein,fi(j*) Task j which is the best task for current agent i*A bid value of;
s2: each agent compares the bid value in the received bid vector with the bid value in the bid vector, and takes the bid vector corresponding to the minimum value as the final bid vector APMAnd will broadcast a bid vector APMThe intelligent agent is used as a task manager, and the task manager takes the most adept task as the next task j to be executed in the task list of the task managerPM；
S3: each agent determines whether the task manager can independently complete task jPMIf the task manager cannot independently complete task jPMEach agent obtains its own task jPMThe difference between the bid value of (a) and the bid value of the task that is best at the user's own right; after receiving the difference value transmitted by each agent, the task manager sends task alliance invitation to the first m agents with the maximum difference value, so that the first m agents send task jPMRespectively as the next task to be executed in its own task list, forming a task alliance, and then entering step S4, where m is the number of participating tasks jPMIs greater than task jPMThe minimum number of agents to be added to the intensity change rate of (c); if the task manager can independently complete the task jPMIf so, the task manager forms a task alliance separately and directly enters the step S4;
s4: will task jPMAnd (4) eliminating the incomplete tasks, and re-executing the steps S1-S3 until all the tasks are distributed.
Further, the task j is executedPMBefore removing the incomplete tasks, the following steps are executed:
s3 a: task manager to not have task jPMBroadcasting calling information as an agent of a next task to be executed in a task list, wherein the calling information comprises the number of each agent in a task alliance and a task jPMCompletion time of
Figure BDA0002659105280000031
S3 b: each intelligent agent receiving the call information acquires the intelligent agent and the task alliance to complete the task j according to the position information and the call information of the intelligent agentPMIs completed by time tccAnd will complete the time tccSending the data to a task manager;
s3 c: the task manager according to all the completion time tccMinimum value of
Figure BDA0002659105280000032
Calculating a time difference ratio TS:
Figure BDA0002659105280000033
s3 d: the task manager judges whether the time difference value ratio TS is larger than a preset threshold value or not, if so, the task manager sends task alliance invitation to the intelligent agent corresponding to the minimum value, and the intelligent agent enables the intelligent agent to send a task jPMThe next task to be executed in the task list is taken as the task, the task alliance is updated, and the step S3e is carried out; if not, go to step S4;
s3 e: and re-executing the steps S3 a-S3 d by adopting the updated task alliance until the time difference value ratio TS is not larger than the preset threshold value.
Further, the completion time tccThe calculation method comprises the following steps:
Figure BDA0002659105280000041
wherein epsilon is a set threshold value,
Figure BDA0002659105280000042
for the s-th in the current task federationiCapability value of the individual agent, andi＝1,2,…,m，
Figure BDA0002659105280000043
is the siIndividual agent arrival task jPMThe time at which the position is located,
Figure BDA0002659105280000044
reach task j for the ith agentPMTime of location, blFor the ability value of the first agent that received the summons message,
Figure BDA0002659105280000045
as task jPMThe rate of change of intensity state.
Further, after any agent completes all tasks in the task list, the agent obtains the tasks currently executed by other agents, obtains the completion time for completing each currently executed task according to the position information of the agent, the position information of each currently executed task, the task strength information and the corresponding task alliance, adds the currently executed task corresponding to the maximum completion time into the task list of the agent, and starts to execute the task.
Further, the method for acquiring the output value of each uncompleted task by the current agent i comprises the following steps:
Figure BDA0002659105280000046
Figure BDA0002659105280000047
wherein f isi(j) Bid value, λ, for current agent i for jth taskjFor the intensity state change rate of the jth task,
Figure BDA0002659105280000048
for the moment when agent i reaches the location of the jth task,
Figure BDA0002659105280000049
as the current time of day, the time of day,
Figure BDA00026591052800000410
for the position coordinates of the current agent i,
Figure BDA00026591052800000411
is the position coordinate of the jth task,
Figure BDA00026591052800000412
is the current agent i speed.
Further, the intelligent agent is an unmanned aerial vehicle, a robot, a reconnaissance plane or an intelligent attack weapon.
Has the advantages that:
1. the invention provides a multi-agent distributed task allocation method for dynamic tasks, which takes the dynamic performance of the tasks into consideration, utilizes an auction-call algorithm to form a pre-allocation scheme, has strong practicability, can be used for completing tasks such as coordinated multipoint aggregation, coordinated multi-target reconnaissance, coordinated multi-target capture and the like, ensures that the decision among agents is free from conflict and the task completion efficiency is effectively ensured, and the task allocation rate reaches 100%;
in addition, the behavior and the communication of the intelligent agent are based on a unified framework, the intelligent agent can be added or deleted based on the framework, the robustness of the whole system is improved, the intelligent agent is suitable for being damaged, and the scenes of the intelligent agent can be increased at any time.
2. The invention provides a dynamic task-oriented multi-agent distributed task allocation method, which is characterized in that after a task alliance for completing a certain task is obtained, other agents capable of greatly shortening task completion time are found according to the calling information of the current task alliance and the information of agents except the task alliance, and the agents are requested to join the task alliance, so that the task completion time can be further shortened on the basis of effectively ensuring the task allocation efficiency.
3. The invention provides a dynamic task-oriented multi-agent distributed task allocation method, after each agent finishes self pre-allocation tasks, the agent dynamically adjusts self-planning to assist other agents to finish tasks according to task execution conditions, and can accelerate the completion of the whole task.
Drawings
FIG. 1 is a flow chart of a multi-agent distributed task assignment method provided by the present invention;
FIG. 2 is a schematic diagram of a task state change when an agent is at a task point;
FIG. 3 is a schematic diagram of a pre-allocation scheme generation process;
FIG. 4 is a state change diagram of task point 1 in an embodiment of the present invention;
FIG. 5 is a state change diagram of task point 2 in an embodiment of the present invention;
fig. 6 is a state change diagram of task point 3 in the embodiment of the present invention.
Detailed Description
In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application.
The invention provides a multi-agent distributed task allocation method which is suitable for multi-agent distributed task allocation. An agent is an abstract concept, and the agent generally has the following functions: dynamic conditions and information in the environment can be perceived; performing an action affects an environmental condition; reasoning is performed to solve the problem. The entities it maps may be: fire extinguishing unmanned aerial vehicle in forest fire; a recourse robot in a natural disaster recourse task; reconnaissance aircraft, intelligent striking weapons and the like in military striking missions.
In all the above scenarios, there are several features:
the ability of a single agent to accomplish a task is limited and varied. Many situations require multiple agents to collaborate to perform a task at a target task point. For example, in disaster relief, fire extinguishing ability, search ability, and the like of the robot are limited, and the values are given in advance according to expert experience, and hereinafter, are abstracted as "intelligent ability values".
The degree of difficulty varies for each task. For example, in a forest fire fighting task, the difficulty of a small fire starting task is small, and a single agent can complete the task; some tasks are large in ignition intensity and large in task difficulty, and a plurality of intelligent agents can complete the fire extinguishing task. Hereinafter, the degree of difficulty of the task is represented by "task strength value".
The intensity value of the task dynamically evolves over time, for example in a forest fire fighting task, the fire is relatively small at the beginning and the intensity value of the task is correspondingly relatively small. The fire behavior is continuously enhanced along with the time, and the task intensity is gradually increased along with the time.
Because a plurality of agents are required to perform and complete a task together, the task allocation scheme requires one agent to negotiate and allocate. Hereinafter, the agent responsible for coordinating the task allocation scheme is denoted "manager".
The invention provides a multi-agent and multi-task distributed task allocation method. Aiming at the dynamic task of the dynamic evolution of the task intensity value along with the time, the method generates a fast and reasonable task allocation scheme through multi-agent distributed negotiation. The invention takes a two-stage task planning algorithm as a solving algorithm of distributed task allocation. The algorithm has high task allocation speed and can effectively prevent task allocation conflict.
As shown in FIG. 1, the present invention provides a multi-agent distributed task allocation method, which comprises the following steps:
step 1: initializing agent information, the agent information comprising: an agent number, an agent position coordinate, an agent speed, an agent physical strength value, and an agent status value.
Wherein the intelligent physical ability value is a quantized value of the intelligent physical ability; the intelligent agent state value is a state indication value of the intelligent agent; the initialized agent state value is 0, i.e. the agent is indicated as idle state.
In the embodiment of the invention, the intelligent agent information comprises an intelligent agent number i and an intelligent agent position coordinate
Figure BDA0002659105280000071
Speed of agent
Figure BDA0002659105280000072
Intelligent physical ability value biState value of agent
Figure BDA0002659105280000073
Figure BDA0002659105280000074
The initial value is 0. When agent i reaches jth task point, the growth of task state is slowed down or the state is reduced, and the model of task strength state changes to
Figure BDA0002659105280000075
Wherein
Figure BDA0002659105280000076
Is the time that agent i predicts to reach the location of task j.
Initializing each task information, wherein the task information comprises: the method comprises the following steps of task number, task position, task initial state, task intensity, intensity state change rate of a task, a task change model and a task threshold epsilon; wherein the task strength value is the agent capability required to complete the task. Wherein the task strength and the intelligent ability value are in the same quantitative way, e.g. for a forest fire fighting task, its intelligent ability value may be its fire fighting ability quantized.
In the embodiment of the invention, the task information comprises a task number j and a task position coordinate (x)j T、yj T) Task intensity xj. Task intensity xjThe value of the intelligent physical ability required to complete task j is a variable that varies with time t, and the intensity model (assuming exponential growth) of task j is
Figure BDA0002659105280000077
Wherein alpha isjIs the current period intensity state change rate for the jth task point. As shown in FIG. 2, when an agent is within the effective scope of a task point, the intensity of the task point is:
Figure BDA0002659105280000081
Figure BDA0002659105280000082
wherein λ isjIs the intensity state change rate, x, of the jth task between time k and time k +1j(k) Is the intensity state value of the jth task point at time k, mjThe number of agents kept at the j-th task point between the time k and the time k +1, and epsilon is a threshold value of the intensity value.
Step 2: the calculated bid value f for all outstanding tasks j for each agent (e.g., agent i)i(j) The task with the highest bid value is the task j which is the best in self*And will bid vector A(i)＝[i,j*,fi(j*)]Broadcast to other agents.
In the embodiment of the invention, the bid value of the agent i to the task j is as follows:
Figure BDA0002659105280000083
wherein the content of the first and second substances,
Figure BDA0002659105280000084
the moment when the intelligent agent i estimates the position of the arrival task j is calculated, and the specific calculation formula is as follows:
Figure BDA0002659105280000085
wherein the content of the first and second substances,
Figure BDA0002659105280000086
it is the current time and is initialized to 0.
And step 3: agent received a of other agents(i)Is marked as
Figure BDA0002659105280000087
Smart body contrast
Figure BDA0002659105280000088
With its own A(i)(ii) a If it is not
Figure BDA0002659105280000089
Is/are as follows
Figure BDA00026591052800000810
Strictly less than A(i)F of (a)iThen use
Figure BDA00026591052800000811
Replace own A(i)(ii) a Otherwise, the signal is kept unchanged. The above process continues until each agent has all its receptions
Figure BDA00026591052800000812
The comparison is completed.
That is, after each agent receives the most adept tasks and corresponding output values from other agents, the most adept tasks and the most adept tasks are compared with each other to determine the tasks to be distributed in the current round and a task manager; a of each agent by the above process(i)Will all be the same, at this point A(i)Is marked as
Figure BDA00026591052800000813
Agent iPMIs considered to be task jPMThe task manager.
And 4, step 4: and for each intelligent agent, judging whether the task manager can independently complete the task or not according to the capability value of the task manager. If the task manager can complete the task independently,
Figure BDA0002659105280000091
i.e. the task manager capacity value is greater than or equal to task jPMThe intensity state change rate of (a) indicates that the task manager can complete the corresponding task jPMIf yes, jumping to step 8, otherwise executing step 5;
and 5: for each agent, compute it completes task j with the task managerPMThe bid value of (1) is marked as V1. At the same time, calculate its bid value (denoted V) for completing its best task2). Will V1-V2And transmitting to the task manager.
Step 6: and after receiving the bid difference value transmitted by each agent, the task manager sorts the bid values from high to low. According to the value, m agents (the number of which is as follows) before the value are selected
Figure BDA0002659105280000092
) Adding their ability values to the ability value of the task manager to complete the task
Figure BDA0002659105280000093
And issues invitations to these agents to form a task federation.
And 7: after the intelligent agent receives the invitation of forming task alliance, the task j is processedPMAdding the state of the agent into a task list of the agent, modifying the state of the agent to be 1, and starting to execute the task.
And 8: to further speed up task completion time, the task manager does not execute task jPMThe agent sends calling information, the information includes agent number of current task alliance
Figure BDA0002659105280000094
With the current task jPMCompletion time of
Figure BDA0002659105280000095
And step 9: the intelligent agent (supposing intelligent agent l) receiving the calling information calculates the calling information and the position of the intelligent agent according to the calling information of the task manager and the position of the intelligent agent, and completes the task j together with the formed task alliancePMTask completion time tcc：
Figure BDA0002659105280000096
And will tccAnd sending the data to a task manager.
Step 10: renThe method comprises the following steps that a business manager selects a proper intelligent agent according to task completion time returned by the intelligent agent, and specifically comprises the following steps: after the task manager receives the task completion time sent by the intelligent agent to be called, the intelligent agent with the lowest task completion time is selected and recorded as the intelligent agent i*Its corresponding task completion time is recorded as
Figure BDA0002659105280000097
Computing
Figure BDA0002659105280000101
If TS is larger than the set threshold, the intelligent agent i is sent*And sending out a task alliance composition invitation.
In the embodiment of the present invention, the threshold is set to 0.25.
Step 11: after receiving the invitation of forming task alliance, the intelligent agent sends task kPMAdding the task into the task list of the user, modifying the state of the agent to be 1, and starting to execute the task.
Step 12: based on not executing task jPMUntil TS is less than the threshold, it will help to speed up task jPMThe agents that complete the time all join the task federation.
Step 13: will task jPMAnd (3) removing the incomplete tasks, and circularly executing the steps 2-13 by adopting the remaining incomplete tasks until all the tasks are distributed, wherein each intelligent agent has a pre-distributed task thereof at the moment, and the intelligent agents execute the tasks according to respective pre-distributed schemes.
Step 14: and after the intelligent agent finishes all pre-allocation tasks of the intelligent agent, modifying the state value of the intelligent agent to be 0.
Step 15: and the agent with the agent state of 0 dynamically adjusts self-planning to assist other agents to complete the tasks for all the tasks which are not completed at the moment, and the completion of the whole task is accelerated. Specifically, the method comprises the following steps: for all tasks that are not completed at this time, the task completion time to assist in completing the task is calculated. And selecting the task with the longest task completion time, adding the task into the task list of the user, starting to execute the task, and modifying the state of the agent to be 1.
Step 16: steps 14-16 are executed in a loop until all tasks have been performed.
The following is a specific example:
agent and task point distribution see figure 3. The parameters of the agent in this multi-agent collaborative task allocation scenario are shown in table 1:
TABLE 1
Speed (m/s) Position (m) Capacity value
Agent
1 2 (40,-30) 0.6
Agent 2 2 (10,-30) 0.3
Agent 3 2 (-35,30) 0.4
The parameters of the task points in the multi-agent cooperative task allocation scenario are shown in table 2:
TABLE 2
Position (m) Rate of change of state
Task point 1 (30,10) 0.4
Task point 2 (50,30) 0.3
Task point 3 (100,50) 0.2
The agent for task allocation obtained in this embodiment executes a task sequence, as shown in table 3:
TABLE 3
Order of execution of tasks
Agent
1 Task point 1 → task point 3
Agent 2 Task point 1 → task point 2 → task point 3
Agent 3 Task point 2 → task point 3
The task execution time obtained in this embodiment is 116.8 s. The state change of the task point in the task process is shown in fig. 4-6.
Therefore, the dynamic deduction of the tasks is considered, the algorithm forms a pre-allocation scheme by using an auction-levying algorithm, so that the decisions among the intelligent agents are free from conflict, and the task completion efficiency is effectively ensured. After each agent finishes the pre-distributed task, according to the task execution condition, the self-planning is dynamically adjusted to assist other agents to finish the task, and the completion of the whole task is accelerated. The algorithm is strong in practicability, can be used for completing tasks such as coordinated multi-point aggregation, coordinated multi-target reconnaissance and coordinated multi-target capture, does not generate a conflicting task allocation scheme, and the task allocation rate reaches 100%. The behavior and the communication of the intelligent agent are based on a unified framework, the intelligent agent can be added or deleted based on the framework, the robustness of the whole system is improved, the intelligent agent is suitable for being damaged, and the scenes of the intelligent agent can be increased at any time.
The present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof, and it will be understood by those skilled in the art that various changes and modifications may be made herein without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (6)

1. A multi-agent distributed task allocation method for dynamic tasks is characterized in that distributed tasks are allocated to multi-agents based on an auction-levying algorithm, and the method specifically comprises the following steps:
s1: respectively taking each agent as a current agent to execute the following operations:
obtaining the bid value of the current agent i to each uncompleted task, and taking the uncompleted task corresponding to the maximum bid value as the task j which is best at the current agent i*And constructing a bidding vector A corresponding to the current agent i according to the bidding vectors A(i)＝[i,j*,fi(j*)]Then the bid vector A is added(i)Broadcast to other agents, wherein fi(j*) Task j which is the best task for current agent i*A bid value of;
s2: each agent compares the bid value in the received bid vector with the bid value in the bid vector, and takes the bid vector corresponding to the minimum value as the final bid vector APMAnd will broadcast a bid vector APMThe intelligent agent is used as a task manager, and the task manager takes the most adept task as the next task j to be executed in the task list of the task managerPM；
S3: each agent determines whether the task manager can independently complete task jPMIf the task manager cannot independently complete task jPMEach agent obtains its own task jPMThe difference between the bid value of (a) and the bid value of the task that is best at the user's own right; after receiving the difference value transmitted by each agent, the task manager sends task alliance invitation to the first m agents with the maximum difference value, so that the first m agents send task jPMRespectively as the next task to be executed in its own task list, forming a task alliance, and then entering step S4, where m is the number of participating tasks jPMIs greater than task jPMThe minimum number of agents to be added to the intensity change rate of (c); if the task manager can independently complete the task jPMIf so, the task manager forms a task alliance separately and directly enters the step S4;
s4: will task jPMAnd (4) eliminating the incomplete tasks, and re-executing the steps S1-S3 until all the tasks are distributed.
2. A dynamic task oriented multi-agent distributed task assignment method as claimed in claim 1, wherein the task j is assignedPMBefore removing the incomplete tasks, the following steps are executed:
s3 a: task manager to not have task jPMBroadcasting calling information as an agent of a next task to be executed in a task list, wherein the calling information comprises the number of each agent in a task alliance and a task jPMCompletion time of
Figure FDA0002659105270000021
S3 b: each intelligent agent receiving the call information acquires the intelligent agent and the task alliance to complete the task j according to the position information and the call information of the intelligent agentPMIs completed by time tccAnd will complete the time tccSending the data to a task manager;
s3 c: the task manager according to all the completion time tccMinimum value of
Figure FDA0002659105270000022
Calculating a time difference ratio TS:
Figure FDA0002659105270000023
s3 d: the task manager judges whether the time difference value ratio TS is larger than a preset threshold value or not, if so, the task manager sends task alliance invitation to the intelligent agent corresponding to the minimum value, and the intelligent agent enables the intelligent agent to send a task jPMThe next task to be executed in the task list is taken as the task, the task alliance is updated, and the step S3e is carried out; if not, go to step S4;
s3 e: and re-executing the steps S3 a-S3 d by adopting the updated task alliance until the time difference value ratio TS is not larger than the preset threshold value.
3. A dynamic task oriented multi-agent distributed task allocation method as recited in claim 2, wherein said completion time t isccThe calculation method comprises the following steps:
Figure FDA0002659105270000024
wherein epsilon is a set threshold value,
Figure FDA0002659105270000025
for the s-th in the current task federationiCapability value of the individual agent, andi＝1,2,…,m，
Figure FDA0002659105270000031
is the siIndividual agent arrival task jPMThe time at which the position is located,
Figure FDA0002659105270000032
reach task j for the ith agentPMTime of location, blFor the ability value of the first agent that received the summons message,
Figure FDA0002659105270000033
as task jPMThe rate of change of intensity state.
4. The multi-agent distributed task allocation method for dynamic tasks as claimed in claim 1, wherein when any one agent completes all tasks in its task list, the agent obtains the tasks currently executed by other agents, obtains the completion time for itself to complete each currently executed task according to its location information, the location information of each currently executed task, the task strength information, and the corresponding task alliance, adds the currently executed task corresponding to the maximum completion time value to its own task list, and starts to execute the task.
5. A dynamic task oriented multi-agent distributed task allocation method as claimed in claim 1, wherein the method for obtaining the output value of each uncompleted task by the current agent i is:
Figure FDA0002659105270000034
Figure FDA0002659105270000035
wherein f isi(j) Bid value, λ, for current agent i for jth taskjFor the intensity state change rate of the jth task,
Figure FDA0002659105270000036
for the moment when agent i reaches the location of the jth task,
Figure FDA0002659105270000037
as the current time of day, the time of day,
Figure FDA0002659105270000038
for the position coordinates of the current agent i,
Figure FDA0002659105270000039
is the position coordinate of the jth task,
Figure FDA00026591052700000310
is the current agent i speed.
6. A dynamic task oriented multi-agent distributed task allocation method as recited in claim 1, wherein the agents are drones, robots, scouts or intelligent weapons of percussion.
CN202010898123.7A 2020-08-31 2020-08-31 Dynamic task-oriented multi-agent distributed task allocation method Active CN112070383B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010898123.7A CN112070383B (en) 2020-08-31 2020-08-31 Dynamic task-oriented multi-agent distributed task allocation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010898123.7A CN112070383B (en) 2020-08-31 2020-08-31 Dynamic task-oriented multi-agent distributed task allocation method

Publications (2)

Publication Number Publication Date
CN112070383A CN112070383A (en) 2020-12-11
CN112070383B true CN112070383B (en) 2022-04-12

Family

ID=73666164

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010898123.7A Active CN112070383B (en) 2020-08-31 2020-08-31 Dynamic task-oriented multi-agent distributed task allocation method

Country Status (1)

Country Link
CN (1) CN112070383B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113268893B (en) * 2021-07-19 2021-10-01 中国科学院自动化研究所 Group trapping method and device based on communication maintenance constraint
CN116260882B (en) * 2023-05-15 2023-07-28 中国人民解放军国防科技大学 Multi-agent scheduling asynchronous consistency method and device with low communication flow
CN117236670A (en) * 2023-11-15 2023-12-15 中国船舶集团有限公司第七〇七研究所 Multi-dimensional distributed job task allocation method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106875090A (en) * 2017-01-09 2017-06-20 中南大学 A kind of multirobot distributed task scheduling towards dynamic task distributes forming method
CN110456633A (en) * 2019-06-29 2019-11-15 西南电子技术研究所（中国电子科技集团公司第十研究所） Airborne multi-platform distributed task scheduling distribution method
CN111199359A (en) * 2020-01-08 2020-05-26 中国电子科技集团公司第五十四研究所 Multi-agent task allocation method under network resource constraint

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2645196B1 (en) * 2012-03-30 2018-12-12 The Boeing Company Network of unmanned vehicles
CN105975332B (en) * 2016-05-03 2018-02-16 北京理工大学 A kind of multiple agent coalition formation method
US10456912B2 (en) * 2017-05-11 2019-10-29 King Fahd University Of Petroleum And Minerals Dynamic multi-objective task allocation
US10235642B2 (en) * 2017-08-11 2019-03-19 Tata Consultancy Services Limited Method and system for optimally allocating warehouse procurement tasks to distributed robotic agents
CN108009012B (en) * 2017-12-14 2021-12-14 中南大学 Multi-agent dynamic task allocation method based on task model
CN111489049B (en) * 2020-03-03 2022-07-05 北京理工大学 Multi-agent distributed task allocation method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106875090A (en) * 2017-01-09 2017-06-20 中南大学 A kind of multirobot distributed task scheduling towards dynamic task distributes forming method
CN110456633A (en) * 2019-06-29 2019-11-15 西南电子技术研究所（中国电子科技集团公司第十研究所） Airborne multi-platform distributed task scheduling distribution method
CN111199359A (en) * 2020-01-08 2020-05-26 中国电子科技集团公司第五十四研究所 Multi-agent task allocation method under network resource constraint

Also Published As

Publication number Publication date
CN112070383A (en) 2020-12-11

Similar Documents

Publication Publication Date Title
CN112070383B (en) Dynamic task-oriented multi-agent distributed task allocation method
Whitten et al. Decentralized task allocation with coupled constraints in complex missions
US7236861B2 (en) Mission planning system with asynchronous request capability
US8776074B1 (en) Methods, devices and systems for coordinating and optimizing resources
CN111489049B (en) Multi-agent distributed task allocation method
CN113395676B (en) Unmanned aerial vehicle task cooperation method for forming game based on overlapping alliance
CN111199360A (en) Unmanned aerial vehicle task allocation planning method
CN109922137B (en) Unmanned aerial vehicle assisted calculation migration method
Peng et al. Review of dynamic task allocation methods for UAV swarms oriented to ground targets
CN111160511A (en) Group intelligent method for consensus active learning
Frasheri et al. Adaptive autonomy in a search and rescue scenario
Mazdin et al. Distributed and communication-aware coalition formation and task assignment in multi-robot systems
CN111784211A (en) Cluster-based group multitask allocation method and storage medium
CN114326827B (en) Unmanned aerial vehicle cluster multitasking dynamic allocation method and system
CN116893861A (en) Multi-agent cooperative dependency task unloading method based on space-ground cooperative edge calculation
CN115016537A (en) Heterogeneous unmanned aerial vehicle configuration and mission planning joint optimization method under SEDA scene
Wang et al. Digital twin-enabled computation offloading in UAV-assisted MEC emergency networks
CN116560409A (en) Unmanned aerial vehicle cluster path planning simulation method based on MADDPG-R
He et al. An operation planning generation and optimization method for the new intelligent combat SoS
Nath et al. A distributed approach for autonomous cooperative transportation in a dynamic multi-robot environment
CN113657718B (en) Multi-robot dynamic alliance task allocation method and related device
Han et al. Cooperative Multi-task Assignment of Unmanned Autonomous Helicopters Based on Hybrid Enhanced Learning ABC Algorithm
CN115665869A (en) Multi-user collaboration platform and method based on edge calculation and directed acyclic graph
Iijima et al. Analysis of task allocation based on social utility and incompatible individual preference
Zhu et al. A real-time decentralized algorithm for task scheduling in multi-agent system with continuous damage

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