CN104063541B - Multi-Robotics Cooperation Method based on hierarchical decision making mechanism - Google Patents
Multi-Robotics Cooperation Method based on hierarchical decision making mechanism Download PDFInfo
- Publication number
- CN104063541B CN104063541B CN201410274560.6A CN201410274560A CN104063541B CN 104063541 B CN104063541 B CN 104063541B CN 201410274560 A CN201410274560 A CN 201410274560A CN 104063541 B CN104063541 B CN 104063541B
- Authority
- CN
- China
- Prior art keywords
- role
- ball
- opponent
- holding person
- robotics
- 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
Landscapes
- Toys (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of multi-Robotics Cooperation Method based on hierarchical decision making mechanism, and sportsman carries out formation selection according to the position judgment of ball and goes reply to compete；Then all sportsmen vote in the holding person forward holding person for oneself thinking now optimal, then carry out other role's distribution；Determine whether forward holding person, if forward holding person, then run at ball, dribbling walking, carry out mathematical modeling to opponent's speed using ideal behavior forecast model and be used for forward holding person walking to play football decision-making module；If not forward holding person, then after carrying out other role's distribution, location point is run to, carries out formation selection.The present invention realizes the selection of forward holding person and the distribution of other all footballer characters successively, simultaneously DOBMP models are established for forward holding person dribbling decision-making module, the problem of high dimension amount of calculation caused by role's function is finally optimized using dynamic programming algorithm, ensure the fluency based on the role rotation under football position constantly change.
Description
Technical field
The present invention relates to a kind of multi-Robotics Cooperation Method based on hierarchical decision making mechanism.
Background technology
FIRA (Federation of International Robot-soccer with strongest influence power in the world now
Association, FIRA) and the big machine world cup people's football matches of RoboCup two, both maximum areas
It is not that FIRA is to allow a team to use traditional centralized control, equivalent to all teammates in a team by same
The control of individual brain.And RoboCup then necessarily requires to use distributed control mode, there is the big of oneself equivalent to each team member
Brain, thus be one independent " main body ".This just needs in depth to study MAS, allows multiple intelligent body planning to cooperate
Go to complete certain goal task with the mode of competition, using evolution algorithmic and group intelligence to reach the breakthrough of entirety
Performance-based objective.
RoboCup3D emulation match in, want to win a football match, depend merely on profile be it is impossible, must
There must be the mutual cooperation of whole team members with cooperating, and RoboCup3D emulation match mainly embodies multiple agent complicated dynamic
Efficiently cooperation how is realized in the environment of state and is tenaciously resisted.The number of player of RoboCup3D simulated environment was from 2010
6 intelligent bodies be changed to 9 people of 2011 to 11 intelligent bodies so far, this cooperation for multiple agent proposes more
High requirement.
On the coordination mechanism problem of multirobot, recent years all has started to different degrees of probe into both at home and abroad.Such as
Portuguese FC Portugal are directed to footballer character assignment problem, using repeated optimum allocation (IOA, Iterated
Optimal Assignment) method, it is based on seeking limited optimal value under famous greedy algorithm, and combines role swap
Mechanism；Observe the football of the mankind, it is thus proposed that it is desirable that by establishing learning by imitation mechanism, system and mankind's complex behavior and machine
Device human action, but in view of the non-intellectual of the basic framework of learning by imitation, interactive interface also are difficult to obtain；U.S. UT Austin
Villa troops application subtask collection optimization method completes the design of target framework, is coordinated using dynamic role assignment algorithm overall
The occupy-place of troop coordinates；BoldHearts troops of Britain use alliance's algorithm, it is intended to build a powerful team of alliance and meet
The requirement of external environment, it can be calculated according to its action parameter of algorithm optimization, while using the Infotaxis decision searches without gradient
Method, the rate value of local maxima information gain；The Robocanes teams in the U.S. are using space-time model matching process is based on, to build
Related motion model and its internal state are found, is calculated referring concurrently to the walking engine mechanism of German B-Human troops, and with heredity
Method and SARSA learning algorithms optimize different behavior act parameter configurations.
The above method is required for certain Optimization Mechanism and learning method, and for role's assignment problem, its is computationally intensive, more
New speed is slow.Above mentioned problem is should to pay attention to and solve the problems, such as during multi-robot Cooperation.
The content of the invention
It is an object of the invention to provide a kind of multi-Robotics Cooperation Method based on hierarchical decision making mechanism, whole multimachine is realized
The effective cooperation of Qi Ren team, the selection of forward holding person and the distribution of other all footballer characters are realized successively, is directed to simultaneously
Forward holding person dribbling decision-making module establishes DOBMP models, finally using caused by dynamic programming algorithm optimization role's function
The problem of high dimension amount of calculation, ensure the fluency based on the role rotation under football position constantly change.
The present invention technical solution be：
A kind of multi-Robotics Cooperation Method based on hierarchical decision making mechanism,
Sportsman carries out formation selection according to the position judgment of ball and goes reply to compete；
Then all sportsmen vote in the holding person forward holding person for oneself thinking now optimal, then carry out other roles
Distribution；
Determine whether forward holding person, if forward holding person, then run at ball, dribbling walking, use ideal
Behavior prediction model carries out mathematical modeling to opponent's speed and is used for forward holding person walking playing football decision-making module, be by ball kick to
Target point, which is still walked, dribbles to target point；
If not forward holding person, then after carrying out other role's distribution, location point is run to, carries out formation selection.
Further, mathematical modeling is carried out to opponent's speed using ideal behavior forecast model to walk for forward holding person
Play football decision-making module, be specially：
Average speed and its position for being currently located by opponent, calculate opponent reach spend required for ball position when
Between T；Know that our sportsman performs the time that striking action is spent simultaneously, given threshold is with our robot success of prediction
Ball is kicked to target point；
Assuming that opponent can prevent us from playing football within the t times, when T-t values are smaller, we successfully completes the task of playing football
Possibility is bigger；
When T-t value be less than setting threshold value when, being considered as the task of playing football can successfully complete, now take by ball kick to
Target point.
Further, opponent can still prevent us from playing football after making a policy, and change the instantaneous velocity of the opponent of foundation
Table, if it is, we, which will fail the completion task of playing football, to set penalty value p to speedometer：
Wherein, VerrIt is the true velocity of opponent and the difference of average speed, n is the number of the instantaneous velocity of sampling.
Further, amount of calculation is reduced using Dynamic Programming function optimization algorithm：
The distance value that each intelligent body reaches first character location is calculated first, is then counted using role's partition function yr
The distance value that each intelligent body arrives separately at all possibilities combination of first and second position is calculated, and preserves each pair intelligent body and arrives
Up to the lowest positioned cost combination of the two positions；
It is to reach { p based on k-1 intelligent body to establish new positioning for k-th of intelligent body1…pk-1Position, i.e., it is sharp
Each intelligent body, which is calculated, with role's partition function yr arrives separately at { p1…pk-1Position all possibilities combination distance value,
And preserve each pair intelligent body and reach { p1…pk-1Position lowest positioned cost combination；
Then distribute each intelligent body and reach pthkThe distance value of individual position simultaneously calculates all intelligent bodies and reaches these three
The lowest positioned cost combination of diverse location.
Further, when calculating lowest positioned cost combination：Lower positioning cost in any subset be present, then include
The cost of the whole positioning method of the positioning is inevitable lower.
Further, voted using the ballot system containing different weights.
Further, in ballot system, the distribution condition of communication information byte is：
Further, the dynamically distributes of footballer character, the role's partition function yr used is to realize optimal occupy-place：
By dictionary sort in the way of select, each intelligent body in all possible occupy-place mode, all intelligent bodies
The sum that walks is most short path；
In shortest path, when two sportsmen have intersection point on path, that is, the situation of collision, role's partition function occurs
Yr obtains lower cost according to triangle inequality by exchanging the target location of two sportsmen.
The beneficial effects of the invention are as follows：This method realizes forward holding person CF's successively under the support of ballot communication system
Selection and the distribution of other footballer characters, and all footballer characters of synchronized update；DOBMP moulds are used for CF judgment mechanisms of playing football
Type analysis decision-making；For the amount of calculation problem of update of role, amount of calculation greatly reducing using Dynamic Programming function, this for
The speed of update of role is very helpful, and ensure that the fluency based on the role rotation in the case of football change in location.
Brief description of the drawings
Fig. 1 is hierarchical decision making mechanism optimization process schematic.
Fig. 2 is the schematic diagram of formation selection.
Fig. 3 is overall occupy-place formation figure.
Fig. 4 is that minimum cost occupy-place illustrates schematic diagram.
Fig. 5 is to carry out mathematical modeling to opponent's speed using DOBPM and be used for CF walkings playing football the decision process of decision-making module
Figure.
Fig. 6 is the formation occupy-place under different match modes.
Fig. 7 is role rotation attack schematic diagram.
Embodiment
The preferred embodiment that the invention will now be described in detail with reference to the accompanying drawings.
Based on RoboCup3D emulation platforms, embodiment devises a kind of multi-robot Cooperation Method based on hierarchical decision making,
Realize the effective cooperation of whole multirobot team.It is tactful mainly to include based role partition function, ballot communication system and reason
Think the hierarchical decision making mechanism three under behavior prediction model (Desired of Behavior Prediction Model, DOBMP)
Individual aspect, the selection of forward holding person (CenterForward, CF) and the distribution of other all footballer characters is realized successively, together
When for CF dribbling decision-making modules establish DOBMP models, finally using high caused by dynamic programming algorithm optimization role's function
The problem of dimension amount of calculation, ensure the fluency based on the role rotation under football position constantly change.
Embodiment
In Apollo3D strategy, using hierarchical decision making mechanism (Hierarchical Decision Making, abbreviation
HDM), as shown in figure 1, the position judgment that so-called hierarchical decision making is exactly sportsman's foundation ball first should currently be gone using what formation
Reply match, then all sportsmen vote in the holding person CF for oneself thinking now optimal, because a football match key
It is CF, it is that the pass or oneself dribbling are advanced, and this is all the key of overall team's policy selection.In whole decision process
It is not to be always maintained at changeless between role and sportsman, current time robot A receives most convenient, and it may be exactly forward
CF, subsequent time due to opponent intercept, A can not realize oneself dribbling break through just passes the ball to teammate, after pass A will according to work as
The occupy-place conversion role at preceding moment.Now the role of other sportsmen and position are in accordance with depending on the position at itself current time
, the communication between all sportsmen and the synchronized update of role are finally realized using a kind of coordination system, that is, passes through communication system
System sends the position of sportsman's self-position and ball, and so each sportsman's can knows the optimal occupy-place of its teammate, allows for ball
Member reaches an agreement, the cooperation being just more beneficial between sportsman.Wherein CF selection must foundation：Whether the sportsman fall down,
Can see ball, ball in his front or rear, the distance apart from football, whether be goalkeeper, also have his upper decision-making
Whether it is CF in cycle, the weight shared by above-mentioned each case is all different.
Formation selects
As mankind's football match, RoboCup3D emulation match reply different situations are also provided with corresponding match mode, such as
Kick off (Kick-Off), goal kick (Goal_Kick), sideline ball (Throw_In), corner-kick (Corner_Kick) etc..From foot
From the perspective of ball match, team's whole strategy can be divided into two big system of attack and defense, and the action selection of sportsman is actually
That exactly according to ball-handling is we or other side, and we, which control ball, is put into attack state, and other side's ball-handling is put into defence state.No
Same formation is as shown in Figure 2.
For usually said team's formation is the position according to ball, as shown in figure 3, the ball when ball is located at court center
The overall erect-position of team, whole formation, which can be divided into, advances and guards two parts, advances the character location of part according to asking
Coordinate position adds what certain offset obtained again, including CF, WFL, WFR, SFL, SFR, CAM and FF.It is unique it is special just
It is this role of CF, its sportsman always nearest apart from ball, the position of ball is set to its coordinate position.By goal center and ball
Position connect into a line, guard type sportsman CDM, CBL and CBR position is on this line, and according to court bottom line again
Add certain offset.And goalkeeper GK position is substantially what is do not influenceed by its teammate, this is to ensure certainly
Family is not lost at goal, if the GK moment is CF excellent person, will have another sportsman to be assigned as GK role and stand in ball
At door center.
Role's partition function yr
Key is exactly the dynamically distributes of footballer character after overall formation determines, the role's partition function y usedrTo realize most
Good occupy-place, when inputting ambient conditions information, function can calculate current time sportsman and role's best match situation.Discussing
Three preconditions must are fulfilled for before the function：
(1) the nearest position of selected distance：Each intelligent body takes out from them respectively in all possible occupy-place mode
The position of (nearer) recently, to ensure that the sum that walks of all intelligent bodies is most short, this just needs the side to be sorted according to dictionary
Formula is selected.
(2) avoidance：Sportsman should avoid colliding with other sportsmen as far as possible when being moved to their set positions.
(3) dynamic is consistent：As long as a series of target location is given, if yrIn moment T output occupy-place modes m, then
Sportsman f in target location processes are moved to by output is m.
If n sportsman, it will there is n！Middle occupy-place.Give extraneous status information, the position of especially n sportsman with
N target location.Cost and descending arrangement successively with a kind of occupy-place mode of n element group representations.It can so be obtained according to cost
To n！The feasible occupy-place of kind, compare these costs according to lexicographic sequence, as shown in Figure 4 and Table 1.
The cost of various occupy-place modes sorts according to lexicographic：
The occupy-place cost of table 1 sorts
This minimum value for being easy to draw the requirement of attribute 1 according to lexicographic sequence.If two sportsmen have on path
The situation of collision, function y occurs in intersection pointrIt is able to can be obtained more according to triangle inequality by exchanging their target location
Low cost.
Ballot communication system
In order to allow all sportsmen of team accurately to reach respective target location, all sportsmen are just necessarily required
Can be harmonious and undoubted for being carrying out role's occupy-place.If sportsman can know ball and its teammate on court
Accurate location, then just without harmonious between sportsman, because calculating that each sportsman can be independent needs to use
Optimal occupy-place.But problem is that sportsman itself has 120 ° of visual angle limitation, and the perception information received is all folder
Miscellaneous noise jamming, so the object seen all has error in distance and angle, thus accurate position can not be obtained
Information.Fortunately intelligent body is allowed to carry out intercommunication in Simspark, i.e., every an emulation cycle between sportsman
(40ms) can be in communication with each other, but the bandwidth of this communications conduit is conditional, can only have a sportsman to send information every time
And the content of information is limited in 20 bytes.
3D simulated environment provides a so-called audio system so that each robot can each two cycle (40ms) it is wide
The information that oneself ' will say ' is broadcast, it is ' listening ' that other robots can receive the information in next emulation cycle, but can not
Know that the information received comes from that intelligent body, it is therefore necessary to sportsman number is added in the information of transmission.All sportsmen
The information sent and received is all to be limited in the ASCII character of 20 bytes, and has part ASCII character not allow to use.
Court is divided into the grid of 5000*5000 sizes, uses 83 between ' * ' to '~' by Apollo3D for amount of compressed data
Individual character is encoded, then can transmit 8320 bit informations.
The specific distribution condition of information byte is as shown in table 2 below, wherein noticeable 14-18 byte conducts
The basis for the layering decision-making system that Apollo3D is used.In addition, ' saying ' that Apollo3D is sent for each sportsman and receiving
' listening ' information uses encryption and decryption strategy, to ensure the safety of our information communication and increase certain antijamming capability.
The distribution condition of the communication information byte of table 2
It must be emphasized that it is very unadvisable that the occupy-place information that communication finally receives, which is only used only, because heat
In noisy interference, sportsman occur falling down or when self poisoning error has accumulation, and the or information sent from server
Occasionally there is the situation lost or be delayed to occur, the information that sportsman receives is just more inaccurate.Contain different weights so using
Ballot system, the situation for the wrong data that the information that even occasionally has is lost or sportsman sends occurs, and can also cause whole team
Use unified occupy-place.
Using the ballot system containing different weights, it is specifically, in play, the task of dribbler is most heavy, can be referred to as
CF.Wherein CF selection gist is：Can whether the sportsman fall down, see ball, ball at his front or rear, apart from football
Distance, whether be goalkeeper, also have the sportsman whether be CF in a upper decision-making period.Power shared by above-mentioned each case
Weight is all different, but is represented with the probability between (0,1).
Preferable behavior prediction model
In MAS, the behavior prediction for other intelligent bodies is the research for having much challenge.It is single in theory
Intelligent body can directly observe the behavior of other intelligent bodies, so as to establish fixed behavior model, but only when intelligent body it
Between there are many repeated information exchanges to establish model., can not be by simply observing in RoboCup3D emulation matches
Behavior with regard to predicting opponent, and enough interbehaviors also are difficult to establish useful mould during real-time change of competing
Type.
Embodiment devises a kind of preferably behavior prediction model DOBPM, to predict single intelligent body under prescribed conditions
Optimal behavior.DOBPM is not based on theory analysis to assume what other intelligent bodies will do, but analyzes their optimal row
It is error to describe its anticipatory behavior.DOBPM models can be used for determining when shoot, pass and optimal held ball moment etc.
Deng.Embodiment using DOBPM carries out mathematical modeling to opponent's speed and is used for CF walkings playing football decision-making module, is to kick ball to mesh
Punctuate, which is still walked, dribbles to target point.The flow chart of whole decision is as shown in Figure 5.
During the games, speed of travel value of the opponent within several cycles is sampled first, and calculate its instantaneous velocity
Vi：
Wherein (xb,yb) be sample last moment opponent positional value, (xc,yc) be current time opponent positional value.
In order to obtain the average speed of opponent, the method for harmonic-mean can be used：
By the average speed of opponent and its position being currently located, it can calculate required for opponent reaches ball position and spend
Time T.Simultaneously it is also known that our sportsman performs the time that striking action is spent, it is possible to which given threshold is with our machine of prediction
Device people success kicks ball to target point.Assuming that opponent can prevent us from playing football within the t times, when T-t values are smaller, we
The possibility for successfully completing the task of playing football is bigger.When T-t value is less than the threshold value of setting, being considered as the task of playing football can succeed
Complete, now take and kick ball to target point.If opponent can still prevent us from playing football after making a policy, illustrate to opponent
The predicted value of average speed be inaccurate, the instantaneous velocity table for the opponent that should now change foundation.If that is,
We, which will fail the completion task of playing football, to set penalty value p to speedometer：
Wherein, VerrIt is the true velocity of opponent and the difference of average speed, n is the number of the instantaneous velocity of sampling.And its is true
Real speed, which is opponent, is run to the spacing of time that final position (position of ball) spent and two positions by initial position
From obtaining.
Dynamic Programming optimization method based on hierarchical decision making
The four module of hierarchical decision making mechanism is elaborated respectively above, based on ballot communication mechanism and ideal behavior model
Establish, so as to show that the process of footballer character distribution distributes to 11 machines with 11 different roles of a troop of playing football
People, but goalkeeper always acts as the role for guarding goal, CF sportsman always nearest from ball, remaining nine character location is all
By Dynamic Programming function yrDraw.If goalkeeper is by chance again when being the sportsman nearest from ball, i.e., when GK is CF, now yrNeed
Want 10！Different targeting scheme in=3,628,800, then calculate their cost respectively and select optimal cost by dictionary sequence
Value, all these calculating must all be completed in 0.02s emulation cycle, and this just needs to consider to use Dynamic Programming function
(Dynamic Planning Function) optimized algorithm reduces amount of calculation.
Wherein A, P represent the set of n intelligent body and its position, positioning method m respectively:=yr(A, P), if any
Lower positioning cost in subset be present, then the cost of the whole positioning method comprising the positioning is inevitable lower.For k-th
It is to reach { p based on k-1 intelligent body that intelligent body, which establishes new positioning,1…pk-1Position.Such as the dynamic rule of three robots
Journey is streaked, as shown in table 3, calculates the distance value that three intelligent bodies reach first character location first, then utilizes role point
The distance value of all possibilities combination of the one or two diverse location is arrived separately at function yr three intelligent bodies of calculating, and is preserved
Each pair intelligent body reaches the lowest positioned cost combination of the two positions.Then distribute each intelligent body and reach the 3rd position
Distance value simultaneously calculates the lowest positioned cost combination that all intelligent bodies reach these three diverse locations.
The occupy-place allocative decision of 3 three robots of table
N intelligent body iterates to calculate by n Dynamic Programming, every time equivalent to the binomial meter to most high order for n-1
Calculate：
When then 11 intelligent bodies participate in match, goalkeeper's totally 10 intelligent body participation role distribution are removed, are advised using dynamic
Amount of calculation is n2 after drawing optimized algorithmn-1=10 × 29=5120, but be without role's allocation algorithm amount of calculation of optimization
10！=3,628,800, it will be apparent that reduce amount of calculation, while decrease role switching time cost.
Experimental verification
All experiments be all use DrawAnnotation functions in Roboviz by our sportsman current time role
Place is shown title overhead, and the implication of each role is illustrated in Figure 5, and our Apollo3D is the machine of blueness
People, red robot are opponents.
Experiment one：Formation occupy-place under different match modes
This experiment is primarily directed to the formation occupancy under different match modes in RoboCup3D emulation matches, such as group
Shown in Fig. 6, Fig. 6 is the formation occupy-place under different match modes, wherein, both sides' occupy-place figure before (a) kicks off；(b) our left corner
Ball occupy-place figure；(c) our croquet occupy-place figure；(d) our forbidden zone corner-kick occupy-place figure.CF, WFL, WFR, SFL, SFR, CAM and FF are
It is responsible for the role of attack in whole troop, wherein WFL, WFR, SFL, SFR and CAM is to follow CF closely to form rectangle behind, point
Do not stand at rectangular four angles and center, and itself is as far as possible towards ball, so can be in the case where ensureing formation, often
Individual sportsman is nearest apart from ball position.FF stands before opponents' goal forbidden zone all the time, shows by many experiments：CF shooting may
It can be intercepted by other side or shooting angle has deviation, this is that FF can occupy vantage point as soon as possible, is switched to the CF of subsequent time
Remedy shooting excellent；CDM, CBL, CBR and GK role are the defence tasks for undertaking oneself half-court, if we is in attack
During state, CDM can occupy field position, and this is to prevent opponent strikes back or scooped out our sportsman from forming our counterattack.
Experiment two：Role rotation and the checking of DOBMP models
Group Fig. 7 describes occupy-place and the role switching of attack part, and No. 2 sportsmen are forward CF in a figures, due to by right
The stop of square sportsman is fallen down, and the role of No. 2 switches to rapidly CAM, and now No. 7 sportsmen are towards ball and nearest apart from ball, its angle
Color switches to rapidly CF, as shown in figure b, c；When No. 7 sportsmen are also intercepted by opponent, using DOBMP models judge by ball kick to
Target point is No. 3 positions of Player, while No. 3 wheels are changed to CF role, and No. 2 and No. 7 simultaneous wheels are changed to SFL and CAM, as shown in figure d；
Due to Simspark match platform regulations：When having more than 2 Tongfang sportsmen in 1 meter of circle of radius centered on ball, it can automatically spring open
All sportsmen apart from ball farther out, so automatically springed open when No. 2 sportsmen are close to ball by platform, while No. 5 wheels are changed to CF, No. 3 wheels
SFL is changed to, as shown in figure e.
Hierarchical decision making mechanism under based role distribution is exactly to realize that forward holds successively under the support of ballot communication system
Ball person CF selection and the distribution of other footballer characters, and all footballer characters of synchronized update；Adopted for CF judgment mechanisms of playing football
With DOBMP model analysis decision-makings；For the amount of calculation problem of update of role, calculating greatly reducing using Dynamic Programming function
Amount, this is very helpful for the speed of update of role, ensure that based on the role rotation in the case of football change in location
Fluency.
Claims (7)
- A kind of 1. multi-Robotics Cooperation Method based on hierarchical decision making mechanism, it is characterised in that：Sportsman carries out formation selection according to the position judgment of ball and goes reply to compete；Then all sportsmen vote in the holding person forward holding person for oneself thinking now optimal, then carry out other roles point Match somebody with somebody；Determine whether forward holding person, if forward holding person, then run at ball, dribbling walking, use ideal behavior Forecast model carries out mathematical modeling to opponent's speed and is used for forward holding person walking playing football decision-making module, is to kick ball to target Point or walking are dribbled to target point；Specially：By the average speed of opponent and its position being currently located, calculate opponent and reach the time T spent required for ball position； Know that our sportsman performs the time that striking action is spent simultaneously, given threshold is with our robot success of prediction by ball Kick to target point；Assuming that opponent can prevent us from playing football within the t times, when T-t values are smaller, we successfully completes the possibility for the task of playing football Property is bigger；When T-t value is less than the threshold value of setting, being considered as the task of playing football can successfully complete, and now take and kick ball to target Point；If not forward holding person, then after carrying out other role's distribution, location point is run to, carries out formation selection.
- 2. the multi-Robotics Cooperation Method as claimed in claim 1 based on hierarchical decision making mechanism, it is characterised in that making certainly Opponent can still prevent us from playing football after plan, the instantaneous velocity table for the opponent for changing foundation, if it is, we fails to complete The task of playing football will set penalty value p to speedometer：Wherein, VerrIt is the true velocity of opponent and the difference of average speed, n is the number of the instantaneous velocity of sampling.
- 3. the multi-Robotics Cooperation Method as claimed in claim 1 based on hierarchical decision making mechanism, it is characterised in that use dynamic Function optimization algorithm is planned to reduce amount of calculation：The distance value that each intelligent body reaches first character location is calculated first, is then calculated using role's partition function yr every Individual intelligent body arrives separately at the distance value of all possibilities combination of first and second position, and preserves each pair intelligent body and reach this The lowest positioned cost combination of two positions；It is to reach { p based on k-1 intelligent body to establish new positioning for k-th of intelligent body1…pk-1Position, that is, utilize angle Color partition function yr calculates each intelligent body and arrives separately at { p1…pk-1Position the combination of all possibilities distance value, and protect Deposit each pair intelligent body and reach { p1…pk-1Position lowest positioned cost combination；Then distribute each intelligent body and reach pthkThe distance value of individual position simultaneously calculates all intelligent bodies and reaches these three different positions The lowest positioned cost combination put.
- 4. the multi-Robotics Cooperation Method as claimed in claim 3 based on hierarchical decision making mechanism, it is characterised in that calculating most During low positioning cost combination：The generation of lower positioning cost, the then whole positioning method comprising the positioning in any subset be present Valency is inevitable lower.
- 5. the multi-Robotics Cooperation Method as claimed in claim 4 based on hierarchical decision making mechanism, it is characterised in that using containing not Ballot system with weight is voted.
- 6. the multi-Robotics Cooperation Method as claimed in claim 5 based on hierarchical decision making mechanism, it is characterised in that ballot system In, the distribution condition of communication information byte is：
- 7. the multi-Robotics Cooperation Method as claimed in claim 6 based on hierarchical decision making mechanism, it is characterised in that footballer character Dynamically distributes, the role's partition function yr used is to realize optimal occupy-place：Selected in the way of dictionary sorts, in all possible occupy-place mode, all intelligent bodies walk each intelligent body Sum is most short path；In shortest path, when two sportsmen have intersection point on path, that is, the situation of collision, role's partition function yr roots occurs According to triangle inequality lower cost is obtained by exchanging the target location of two sportsmen.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410274560.6A CN104063541B (en) | 2014-06-18 | 2014-06-18 | Multi-Robotics Cooperation Method based on hierarchical decision making mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410274560.6A CN104063541B (en) | 2014-06-18 | 2014-06-18 | Multi-Robotics Cooperation Method based on hierarchical decision making mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104063541A CN104063541A (en) | 2014-09-24 |
| CN104063541B true CN104063541B (en) | 2017-12-01 |
Family
ID=51551254
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410274560.6A Active CN104063541B (en) | 2014-06-18 | 2014-06-18 | Multi-Robotics Cooperation Method based on hierarchical decision making mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104063541B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110520813B (en) * | 2017-05-05 | 2022-06-24 | 赫尔实验室有限公司 | System, computer-implemented method, and storage medium for predicting multi-agent movement |
| EP3740293A4 (en) * | 2018-01-21 | 2022-07-06 | Stats Llc | Method and system for interactive, interpretable, and improved match and player performance predictions in team sports |
| CN109254584B (en) * | 2018-09-13 | 2021-08-17 | 鲁东大学 | Role distribution method and device based on multiple intelligent agents, computer equipment and storage medium |
| CN109794937B (en) * | 2019-01-29 | 2021-10-01 | 南京邮电大学 | Football robot cooperation method based on reinforcement learning |
| CN110377048B (en) * | 2019-06-26 | 2023-02-03 | 沈阳航空航天大学 | Unmanned aerial vehicle group defense deployment method based on genetic algorithm |
| CN112221160B (en) * | 2020-10-22 | 2022-05-17 | 厦门渊亭信息科技有限公司 | Role distribution system based on random game |
| CN113001545B (en) * | 2021-03-01 | 2022-04-29 | 北方工业大学 | Robot control method and device and robot |
| CN113799143B (en) * | 2021-11-18 | 2022-04-19 | 广东隆崎机器人有限公司 | Safe cooperation method and device of multiple robots in working area |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3994950B2 (en) * | 2003-09-19 | 2007-10-24 | ソニー株式会社 | Environment recognition apparatus and method, path planning apparatus and method, and robot apparatus |
| CN103292804B (en) * | 2013-05-27 | 2015-07-15 | 浙江大学 | Monocular natural vision landmark assisted mobile robot positioning method |
-
2014
- 2014-06-18 CN CN201410274560.6A patent/CN104063541B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN104063541A (en) | 2014-09-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104063541B (en) | Multi-Robotics Cooperation Method based on hierarchical decision making mechanism | |
| Zhang et al. | Efficient communication in multi-agent reinforcement learning via variance based control | |
| CN104942807B (en) | Method for capturing targets by aid of multiple robots on basis of extensive cooperative games | |
| Riley et al. | Planning for Distributed Execution through Use of Probabilistic Opponent Models. | |
| CN110114735A (en) | The image of match is captured by nobody the autonomous vehicles | |
| MacAlpine et al. | Positioning to win: A dynamic role assignment and formation positioning system | |
| CN103301630B (en) | Coordination and cooperation control method for football robots and system thereof | |
| Akiyama et al. | Online cooperative behavior planning using a tree search method in the robocup soccer simulation | |
| Lau et al. | Multi-robot team coordination through roles, positionings and coordinated procedures | |
| Chen et al. | Research on the approach of task decomposition in soccer robot system | |
| CN110134139A (en) | The tactical decision method and apparatus that unmanned plane is formed into columns under a kind of Antagonistic Environment | |
| CN104090573A (en) | Robot soccer dynamic decision-making device and method based on ant colony algorithm | |
| Lau et al. | Robot team coordination using dynamic role and positioning assignment and role based setplays | |
| Yasui et al. | Analyzing and learning an opponent’s strategies in the RoboCup small size league | |
| Shi et al. | Research on self-adaptive decision-making mechanism for competition strategies in robot soccer | |
| Kober et al. | Learning elementary movements jointly with a higher level task | |
| Zhang et al. | Multi-robot cooperation strategy in game environment using deep reinforcement learning | |
| Haobin et al. | Robot soccer confrontation decision-making technology based on MOGM: Multi-objective game model | |
| Sari et al. | Decision system for robosoccer agent based on ooda loop | |
| Wang et al. | Research on multi-robots self-organizing cooperative pursuit algorithm based on Voronoi graph | |
| He et al. | Formation optimization of RoboCup3D soccer robots using delaunay triangulation network | |
| Wei et al. | [Retracted] Application of RoboCup 3D and Intelligent Technology in Football Simulation League | |
| Atkinson et al. | On-the-fly generation of multi-robot team formation strategies based on game conditions | |
| Burris | A trajectory planning algorithm for quantifying space ownership in professional football | |
| Bamal | Collision-free path finding for dynamic gaming and real time robot navigation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |