Embodiment
The main contents that the present invention relates to comprise following 5 points: 1) based on the foundation of the distributed model of local Di Langnei triangulation; 2) sensing network kinematics; 3) distributed self-organizing; 4) composite formation and cooperation algorithm; 5) distributed mobile robot ad-hoc route and Data Transport Protocol.
Its purpose may be summarized to be following 5 points: 1) based on the distributed model of local Di Langnei triangulation, set up the method for sensor network being carried out general analysis; 2), make that the mobile robot can be constantly near desired locations to the kinematic research of sensor network; 3) distributed self organization ability can be so that network disperses, converges, oneself's identification and maximization monitoring range; 4) a kind of composite formation of exploitation and cooperation algorithm can make multirobot cooperate mutually so that reach common target; 5) distributed mobile robot ad-hoc route of exploitation and Data Transport Protocol.
1, based on the distributed mobile sensor network model of local Di Langnei triangulation
The mobile sensor network model comprises the mobile apparatus human model and based on the distributed model of graph theory.
In mobile sensor network, movable sensor is wireless connections, shown in Fig. 1 (a).Be expressed as R={R
1, R
2..., R
nN robot be configured in the smooth zone.R
iThe parameter that is provided with of individual robot is q
i(t)=[x
i, y
i, θ
i)
T, i=1,2 ..., n.X wherein
iAnd y
iBe R
iCoordinate in the individual robot place local coordinate system, θ
iIt is its deflection.Shown in Fig. 1 (b).Its dynamic model is described as
Wherein, u
iBe subsystem R
iControl input.
In this invention, relate to complete model and non-complete model based on the mobile robot.Complete model is described below:
Wherein, v
IxAnd v
IyBe in local coordinate system respectively along the translational speed of x and y direction, the control input u in the model
iBe defined as: u
i={ u
I1, u
I2}
T
Suppose that translational speed and rotary speed all are controlled, have the robot model f of nonholonomic constraint
iCan be described to:
Wherein, v
iBe mobile robot's radial velocity, ω
iBe angular speed.Mobile robot's nonholonomic constraint is described as :-x
iCos θ
i+ y
iSin θ
i=0.In local coordinate system, parameter is set in definition and control variables is unnecessary.For example: shown in Fig. 1 (b), at ∑
iIn defined q
i
In the distributed model based on graph theory, but mobile sensor network is a system with setting parameter and controlled variable, but its setting parameter and controlled variable are expressed as q={q respectively
1, q
2..., q
n}
TAnd u={u
1, u
2..., u
n}
TMulti-robot system can be modeled as a multidimensional interconnected systems, and whole system is expressed as:
Wherein f is the vector field of system dynamic model.In this invention, the distributed model based on graph theory has been proposed.It is made up of two parts: the relation between subsystem and the adjacent node 1) dynamic model 2 of each subsystem).
The definition of multi-robot system global object for convenience,
Be described as the R of robot in the unified rest frame
iThe position,
Be described as the position of all robots in the unified rest frame.And definition p
i={ x
i(t), y
i(t) }
TBe the local coordinate system ∑
iThe middle R of robot
iThe position.
Relation between adjacent machines people's node defines by Di Langnei triangulation and Wei Lang noy figure, as Fig. 2 (a).In smooth zone, tieing up bright noy figure and Di Langnei triangulation all is dual concerning the other side.Tie up bright noy figure and defined the overlay area of each movable sensor, the Di Langnei triangulation has then defined adjacent node on how much of the movable sensors and the relation between them.
The overlay area of sensor network is an important attribute to network.In this invention, tie up bright noy figure and be used for describing this attribute, shown in the dotted line among Fig. 2 (a).Consider in parameter
The colony of the robot of following configuration, each robot all should cover the part among the overall area Ω.In Ω, sensor node R
iThe subregion that is covered is defined as:
V={V
1, V
2..., V
n) form a kind of division of flat site Ω, wherein sensor node R
iOnly at subregion V
iThe generation effect.The R of robot
iBe also referred to as figure " generator ".V
iBe defined by the convex polygon zone, covered the R of robot that ties up among the bright noy figure
iTherefore, for convex polygon V
iCalculating, need a kind of location algorithm support distributed model.Communication range and monitoring range also as shown in Fig. 2 (b), are defined by two disks.In order to make the complete overlay area Ω of sensor network, the monitoring range of each transducer must cover its bright noy figure of dimension.
The inserted layout of Di Langnei (nested grid) is a kind of based on point set
The triangulation to flat site Ω, it is defined as: all intersect with there being one of limit two internodal any additional sides.In the drawings, direct and R
iThe node that links to each other becomes R
iAdjacent node, be defined as gathering K
iThe Di Langnei triangulation has defined the connection attribute between adjacent node, and this attribute is by boundary set ε={ e
I_j(t), i, j=1,2 ..., n, i ≠ j} defines.Multi-robot system is driven by connection matrix, and connection matrix A (t) is defined the connectivity of the arrangement that is used for indicating that Di Langnei is inserted.A (t) is by the uniquely defined square formation of Di Langnei triangulation.Wherein, if R
iAnd R
jBe adjacent node, A then
Ij=e
I_j, otherwise, A
Ij=0.For example, in Fig. 2 (a), the R of robot
2Five adjacent nodes are arranged, and Fig. 2 (b) is the subgraph of Fig. 2 (a) Di Langnei triangulation, has specifically shown R
2And the relation between its adjacent node.R
iBe defined as vectorial d (t)={ e with the connection attribute of its adjacent node
I_i1, e
I_i2..., e
I_ik}
T, wherein, R
I1, R
I2..., R
IkBe R
iAdjacent node.e
I_j=e
I_j∠ α
I_jNot only defined the spacing e of two robots
I_j, also defined R in robot
iLocal coordinate system under, the R of robot
jDirection α
I_jAngle α
I_jShown in Fig. 2 (c).
For a collaborative task, R
2Move and only to be influenced by its adjacent node.Distributed robot's model description of considering adjacent node is as follows:
Wherein, s
iBe to wait a moment the motion of introduction with reference to amount.p
I1, p
I2..., p
IkBe at the R of robot
iThe local coordinate system ∑
iIn adjacent node
The position.
In addition, one of advantage of mobile sensor network is that movable sensor can produce high-quality monitoring information and relayed information for query node by cooperation.Fig. 3 (a) has showed this mode.
Mobile robot R
8Detect an object, and the adjacent machines people R that in the Di Langnei triangulation, defines of request
6R
9And R
10Surround this object.By four information that transducer obtained to, query node R need sent
12Merged before.Suppose that a more than transducer can monitoring object, information is transferred to a sensor node and carries out information fusion.Graph model provides a kind of perfect scheme that addresses this problem.If object in the bright noy of the dimension of sensor node zone, so this object apart from the distance of this transducer than shorter apart from other movable sensor.This node will be asked for information, and with it fusion.Data after the fusion are put the transmission path of destination node from the seedbed, obtain by the ad-hoc Routing Protocol.This invention comprises: 1) statistical model of combined sensor node is considered the sensor fusion in the network; 2) for from monitoring node to the query node transmission of Information, attribute e connects
I_jStatistical model; 3) the mobile sensor network of describing in order to simple form is realized and testing algorithm.
Three-dimensional mobile sensor network is than the better prospect of having of two-dimensional space.Therefore, the Cooperation And Coordination method that is distributed in three-dimensional transducer is discussed.Three-dimensional mobile sensor network comprises one group of ground dolly and a drone, or is made up of one group of ground dolly and some climbing robots.Fig. 3 (b) is a three-dimensional wireless sensor network schematic diagram.
2, sensor network kinematics analysis
In setting up in the process of mobile sensor network kinematic relation, the parameter q of robot
iIn the local coordinate system shown in Fig. 1 (b), be defined.Relation between the local coordinate system of adjacent node is to share the key of useful information between them.For each robot, local coordinate system all is different, as the R of robot among Fig. 1 (b)
iAnd R
jShown in.The sensing of each robot and data fusion are all about local coordinate system.In order to share useful information in the different machines human world, need a transformation matrix.For example, the R of robot
jDetect an object, its parameter is used the local coordinate system ∑
jIn ξ
s j(t) represent.For with the R of robot
iThis information of sharing needs to determine two coordinate system ∑s
iAnd ∑
jBetween relation.
The R of robot
iAnd R
jBetween relation can pass through transformation matrix T
i jAnd T
j iDefine, be specifically by between two robots apart from e
I_jWith their relative bearing θ
I_jRepresent.Can communicate by letter between adjacent node, so that calculate relative bearing between them.Fig. 2 (c) example of having passed the imperial examinations at the provincial level.Based on the R of robot
iAnd R
jBetween communication, α
1_2And α
2_1For two robots all is as can be known.Wherein, α
1_2Be at R
1R in the local coordinate system
2The orientation; α
2_1Be at R
2R in the local coordinate system
1The orientation.The calculating formula of relative bearing is between the two: θ
1_2=α
1_2-α
2_1+ π.By sharing information, two robots can both the computational transformation matrix T
i jAnd T
j i, wherein
α
J_iBe at R
iR in the local coordinate system
jThe orientation.In sensor network, communication is considered to a kind of perception.Based on the location between adjacent node, resulting transformation matrix T
i jIt is the kinematic foundation stone of overall sensor network.In each local coordinate system, transformation matrix has defined the relation between the adjacent node, so that share Useful Information.In real system, α
I_j, θ
I_jAnd e
I_jBe respectively to have variances sigma
α i_j, σ
θ i_jAnd σ
Ei_jStatistical variable.Thereby set up the mobile sensor network kinematic relation.
3, distributed self-organizing technique
A mobile sensor network must be a multi-robot system.In this multi-robot system, robot can cooperate to finish distributed task scheduling, such as: distributed monitoring, collaborative process.In this system, one of target of robot cooperation is: the overlay area of maximization network.Therefore, method of addition and potential field method have been introduced.In this invention, a kind of distributed self-organized algorithm based on potential field method and Di Langnei triangulation has been proposed, comprise the self-organizing of no constraint space, based on following self-organizing and the dispersion of motion planning, barrier and the environmental constraints of biology excitation, converge and fault-tolerance approach and non-holonomic constraint tracking.
In the sensor network, the coverage of each sensor node is to define according to their communication range or monitoring range, shown in Fig. 2 (b).The definition R of robot
iCommunication range be c
i, monitoring range is s
i, and c
iCompare s
iGreatly, shown in Fig. 2 (b).
For some sensor nodes, it is not closed tieing up bright noy zone, as the R among Fig. 4
1Can be with a special line segment
Or arc
Seal this curve, as shown in Figure 4.If there is an object to appear in the zone, so Qu Yu definition will be considered the shape of object.For example: among Fig. 4, line segment
With
Be used for closed area V
iThe bright noy of the dimension of gained zone V
iThe normal areas that can be considered to robot cooperated system.
In the self-organizing of no constraint space, for the mobile robot R in the multi-robot system
i, performance index (candidate's liapunov function) are defined as follows:
Wherein,
And d
I_jBe respectively reality and the desired distance between two robots.p
IjBe at R
iIn the coordinate system, from the R of robot
iTo the R of robot
jVector.m
iBe R
iThe sum of adjacent node.k
iAnd k
IvIt is the parameter of virtual potential energy and robot dynamics's kinetic energy.The R of robot then
iControl be input as:
F
i={ F
Ix, F
IyBe by adjacent node R
iThe virtual potential field power that produces.If except R
iOther robot all is static, proves that so easily controller discussed above is global convergence.Consider that multi-robot system is an integral body, the performance index of setting up according to the potential energy of the dynamics kinetic energy of overall system and virtual potential field can be described as follows:
Based on total energy function V, can obtain the control input vector u of multi-robot system.When considering whole system, the R of robot
iVirtual potential field power be defined as:
Here we suppose, k
i=k
jWhen considering whole system, R
iControl be input as:
Because k
iBe controller gain, so the R of robot that is indicated by governing equation (6) and (7)
iDistributed director can think identical.Therefore, we can prove, are global convergences based on the control of the multi-robot system of virtual potential field and Di Langnei triangulation method.Based on the R of robot
iThe distributed control of adjacent node causes the whole system global convergence.
In the process of adjusting certainly, the change of Di Langnei triangulation shape is unworthy.Some nodes have increased adjacent node, and the adjacent node of some nodes has reduced simultaneously.The discontinuous variation that has caused virtual potential field power in the equation (6) of topology incident, this can cause movement locus unsmooth of node.Therefore, introduced the motion planning that encourages based on biology.
In a dynamic Di Langnei triangulation, the position of aforementioned " generator " changes in time.Up to the generation of topological incident, along with the motion of robot, the shape of triangulation is a continually varying.For a robot, topological event definition is: " generator " lost some nodes or increased some new nodes.According to the definition of virtual potential field function, for a robot, V
iContinuity depend on the quantity of its adjacent node.When it was lost or increases some adjacent nodes, violent variation had just taken place.Therefore the continuity of virtual potential field power also just has been subjected to the influence of topological incident.At first, virtual potential field power is very large, might exceed mobile robot's speed limit.In order to make movement locus level and smooth, be subjected to biological inspiration, the shunting model is introduced into.
Originally, this shunting model is based on thin-skin model development and comes.It is introduced in Control of Nonlinear Systems and the mobile robot's tracking Control.It can describe the behavior to complicated and dynamic environment real-time adaptive.In this invention, one based on virtual potential field power F
i={ F
Ix, F
Iy}
TNew reference r
i={ r
Ix, r
Iy}
TBe defined.New reference r
iAnd F
iThe shunting model be defined as follows:
Parameter A, B and D are respectively the upper and lower bounds of passive attenuation rate, nervous system behavior.Function h
+() and h
-() represented the excitation input respectively and suppressed input, and they are defined as: h
+(x)=and max (0, x), h
-(x)=max (0 ,-x).For any excitation and inhibition input F
i, the new virtual potential field power based on the shunting model all has been limited among bounded interval-[D, the B].Based on this new potential field power, the control law of equation (6) representative is changed to:
u
i=-r
i-k
ivV
i (9)
In addition, carrying out the distributed of network based on particle group optimizing method can be described as from deployment and self-organizing:
The self-organizing of network arranges that the optimization problem that needs to solve is: cluster is at the transducer of certain regional random distribution, how to control them in the short period of time, carry out self-organizing and arrange, make the coverage maximization of sensor network, and each transducer reaches even distribution.
Also have a constraints in addition, that is: guarantee after network carries out self-organizing that each node has at least one one hop node to be communicated with it, guarantees internodal communication.
Following specific description adopts particle cluster algorithm to carry out the process of self-organization of network:
When node more after a little while, for the cluster transducer, adopt following method to carry out the self-organizing of node.
1. to the hypothesis of particulate: owing to be self-organizing at the enterprising line sensor network of two dimensional surface, so establish x
i=(x
I1, x
I2..., x
Im), y
i=y
I1, y
I2..., y
Im) i=1,2 ..., n is the position vector of i particulate among the particulate group.Wherein, m represents the quantity of node in the cluster transducer; N represents particulate group's scale, that is: n particulate arranged among the particulate group; x
iAnd y
iAbscissa and the ordinate of representing i particulate position respectively.
Establish v again
Xi=(v
Xi1, v
Xi2..., v
Xim) v
Yi=(v
Yi1, v
Yi2..., v
Yim) be respectively that particulate i is along the velocity vector on x and the y direction; p
Xi=(p
Xi1, p
Xi2..., p
Xim), p
Yi=(p
Yi1, p
Yi2..., p
Yim) be particulate i in optimizing process the horizontal ordinate of position of process with best adaptive value; p
Xg=(p
Xg1, p
Xg2..., p
Xgm), p
Yg=(p
Yg1, p
Yg2..., p
Ygm) be the horizontal ordinate of the optimal location that searches of whole population.Wherein, the implication of m, n is with top identical.
2. the initialization of particulate: arrange because node generally carries out self-organizing around bunch head.Again because the needs of inter-node communication, so, when the particulate initialization, being the center of circle, be node initializing in the circle of radius with 1 with bunch head.Such initialization, originally the transducer in the expression self-organizing network is to center on a bunch random arrangement.
3. the calculating of fitness: establish each euclidean distance between node pair and be:
In the self-organization of network model of being discussed, the cluster transducer carries out the self-organizing configuration in certain zone, be to lead this bunch mobile node to enter this zone by a bunch headband earlier, in this zone, carry out the self-organizing configuration around bunch head then, so be positioned at the center of network as bunch first, other transducer is as its hop node.Therefore, the employed fitness function of this paper is:
Wherein, d
IjBe the distance between node i and the node j, D
kBe k node except that bunch head to the distance of bunch head, the implication of m is still the same.Through after changing, fitness function is represented be the node distance that arrives other node except that bunch head with, get fitness function like this, reduced optimization aim, can make the self-organizing of network rapider.
4. the improvement evolution equation of speed and position:
For the j of i particulate dimension (j transducer in the cluster transducer), evolve to the position in t+1 generation and speed with following evolution equation calculating from t generation:
x
ij(t+1)=x
ij(t)+η(t)v
xij(t+1)
y
ij(t+1)=y
ij(t)+η(t)v
yij(t+1)(10)
v
xij(t+1)=ω(t)v
xij(t)+c
1r
1(p
xij(t)-x
ij(t))
+c
2r
2(p
xgj(t)-x
ij(t))
v
yij(t+1)=ω(t)v
yij(t)+c
1r
1(p
yij(t)-y
ij(t))
+c
2r
2(p
ygj(t)-y
ij(t))(11)
Wherein, i=1,2 ..., n, j=1,2 ..., m, ω are inertia weight, c
1, c
2Be acceleration constant, r
1, r
2Be equally distributed random number in [0,1].ω, c
2With η is dynamically to adjust, and its adjustment law is respectively:
ω(t)=ω
max-(ω
min×t/g
max)(12)
In the formula, ω max, ω min are maximum and the minimum values that ω changes; η
Max, η
MinBe maximum and the minimum value that η changes, generally get η
Max∈ [1.0,1.8], η
Min∈ [0.4,0.8]; c
2max, c
2minBe c
2The maximum and the minimum value that change are generally got c
2max∈ [1.6,2.0], c
2min∈ [0.6,1.0]; T is an optimizing algebraically, g
MaxBe maximum iteration time; λ is a constant, generally gets λ ∈ [0.005,0.015].
5. the processing of constraints: after passing through the calculating of Position And Velocity evolution equation, the position of node may surpass the communication range of bunch head in the sensor network, and tackle it and reinitialize according to the initialized method in front this moment.
Utilize above-mentioned self organizing network model, and, guaranteeing under the prerequisite that to communicate by letter between node, must make the coverage maximization of network, and each transducer can evenly distribute the self-organizing of realization sensor network in conjunction with particle cluster algorithm.
When considering than the self-organizing problem of multinode in certain zone, if fitness function also according to the internodal distance that adopts above and, increase along with node, internodal connection meeting is with exponential increase, the target of optimizing also can increase, and algorithm may cause self-organization of network not prompt enough.At this moment need introduce auxilliary bunch head.
Auxilliary bunch head is to carry out some nodes after the self-organizing according to the method in the last trifle.It is intended for new center, instructs other transducer that does not dispose to carry out self-organizing.Concrete grammar is as follows.
At first, use said method, count N, calculate the main bunch head optimal location of node on every side according to given node around main bunch head.And all the sensors is divided into N submanifold of quantity approximate equality;
Secondly, in N submanifold, select a node respectively as auxilliary bunch head, it is configured to the optimal location place of being calculated by main bunch head, in this course, the submanifold at its place is followed it and is moved near the optimal location;
Once more, each auxilliary bunch head calculates their optimal locations (will consider that wherein the node that has configured is no longer to move) on every side again; And according to the optimal location sensors configured;
So carry out recursion, make whole network constantly launch, finish up to all the sensors self-organizing.
To the self-organizing under barrier and the environmental constraints, can be based on the Di Langnei triangulation of virtual potential field method with environmental constraints and dynamic barrier combined factors.For environment as shown in Figure 5, based on the heat transfer agent of robot, additional connection is added into.Additional connection is used to define the R of robot
iVirtual energy function.Have the energy of the system of barrier to be described as:
Wherein, k
IlBe a parameter, m
oBe the R of robot
iThe total quantity of barrier in the monitoring range, ‖ p
I_l‖ and d
I_lBe respectively actual range and the desired distance between robot and barrier.The R of robot
iOnly consider its adjacent node and the barrier in its monitoring range.
Subsequently, we can access node R
iVirtual potential field power:
Based on new potential field power, the robot controller under the environment that is tied can be devised.
For distributed mobile sensor network, disperseing and converging is two important operations.With the controller that equation (6) and equation (9) are described, can be by adjusting the desired distance d between two nodes
I_jSolve the problem of disperseing and converging.If the mobile robot is that algorithm can make that robot converges, so that cover certain zone in the sparse space that is configured in an opening.
For a multi-robot system, one of most important characteristic is its redundancy and fault-tolerance.
If one or more robots had lost efficacy, other robot can reconfigure to their position according to the needs of task.
For the multi-robot system of forming by the non-holonomic constraint mobile robot, have to revise above-mentioned algorithm, so that satisfy the non-holonomic constraint of robot.Based on the transducer of having equipped tracking control unit,, can realize by the combination of Di Langnei triangulation and virtual potential field method to non-holonomic constraint mobile robot's tracking Control.For a non-holonomic constraint robot shown in Fig. 1 (b), along the X of robot
i' by a bit of reference Point C that has apart from d be specified.The center and the kinematic relation between the C of trailing wheel can be described below:
Imperfect ROBOT CONTROL input based on virtual potential field power is described as:
Virtual potential field power r
iX along robot
i' axle and Y
i' axle is broken down on two vertical directions, shown in Fig. 1 (b).A generation is along X
iThe lateral displacement of ' axle, another produces angular displacement.In design, along X
iForward direction that ' direction allowed and back are unnecessary to moving.
4, the formation control of sensor node and cooperation technology
The target of controlling of forming into columns is to coordinate one group of robot with a kind of method.This method is: with respect to other robot, they will keep a kind of given formation.One group of robot can be used for carrying out a task very difficult for a robot.For search and rescue under the hazardous environment and military aspect, formation is important, can cooperate and finish the task of a complexity.Do not consider the robot diversified application background of forming into columns, the target of multi-robot coordination algorithm is to accomplish a task with being used for cooperation.In this invention, we propose a kind of distributed algorithm of formation control exploitation for large scale system.Distributed director is based on model q
i=f
i(q
i, u
i), u
i=h
i(s
i, q
i, p
I1, p
I2..., p
Ik).Wherein, s
iBe called motion reference, it is the task of forming into columns.Each mobile robot's controller all is the linear combination of task object and its local virtual potential field power.This invention has related to the method for two kinds of controls of forming into columns---the asynchronous formation control and the control of forming into columns synchronously.And, composite formation's control method has been proposed based on this.
In robot formed into columns, iff the desired motion that is certain the robot understand network in the sensor network, and it did not inform other robot by communication, and we claim that this situation is asynchronous formation control so.There is the robot of whole formation desired motion information to be called as in " leader robot ".If " leader robot " according to the position of the exercise program change of forming into columns oneself, its adjacent machines people also can this motion of perception by the change of relative distance between them so.
Although the form of forming into columns in asynchronous formation control changes sometimes slightly, the robot in the sensor network can both follow the motion of " leader robot ".Under the local coordinate system of each robot, all motions of controlling them are unworthy.For global coordinate system less than the approval.In asynchronous formation system, the motion of " leader robot " is to be subjected to the restricted number of robot in the response time of monitoring system and the sensor network.If the speed of " leader robot " is too big, other robot may not catch up with its motion, and formation can be damaged.
The control of forming into columns synchronously is such: sensor network is the network of a synchronous communication, if a robot receives the formation moving projection, the network agreement that floods just is used to give other robot in the network formation planned transmission.For different transducers, identical plan is explained it is unnecessary respectively.S is expressed as the expected path of formation, then s
iBe that it is at the R of robot
iExplanation under the coordinate system.Adjacent node R
iAnd R
jBetween the shared of information be by transformation matrix T
i jFinish.
The control of forming into columns synchronously is not a kind of " leader---follower " system.If " leader robot " lost efficacy, other any one robot can both replace it as " leader robot ".
For model u
i=h
i(s
i, q
i, p
I1, p
I2..., p
Ik), if s
iConflict with virtual potential field power, virtual potential field power can be arranged motion, and this causes moving forward and backward of robot sometimes.In order to reduce the consumption of energy, designed a kind of hybrid formation control.If R
iThe motion and the plan s of formation inconsistent, another controller u so
iWill be called.Wherein the motion of each individual robot all is consistent with the exercise program of formation.Hybrid algorithm needs sensor node that the consistent of exercise program admitted.The purpose of composite formation's control algolithm is improve network intelligent.Composite formation's control algolithm of more complexity is designed.For example, when other robot was freely disposed, three adjacent machines people can form a kind of special geometric figure in the network.
Distributed formation control algolithm is a kind of expansion from setting algorithm.The utilization of formation control algolithm is admitted the consistent of task object s, and each mobile robot explains it according to own position in sensor network.The only conversion of existence formation has just been formed into columns by robot like this.Therefore, we propose mobile sensor network formed into columns and are defined as " visual human ", and this just means that formation can conversion and rotation in two-dimensional space so, forms into columns and also can disperse and converge.Combining, develop a kind of distributed algorithm and come " visual human " operated to the consistent algorithm of admitting of task object with from setting algorithm.Perhaps develop a kind of composite formation control algolithm.In mobile sensor network, the control of adjusting certainly and form into columns is important, but for the monitoring task, the cooperation of movable sensor is more importantly.Hybrid algorithm can be considered the demand and the overlay area of transducer cooperation simultaneously, and realizes the merging of forming into columns, the interception of barrier etc.
5, mobile robot ad-hoc route and Data Transport Protocol technology
The cooperation of multi-robot system is supported by the ad-hoc communication network.Monitoring Data is propagated in network also needs to pass through communication network.For the data query of suitable operation and mobile sensor network, be essential to the identification of network and control.In the mobile sensor network, each mobile robot is configured to router.For transmission of monitoring data and packet in network, to the improvement of required ad-hoc Routing Protocol.The transmission of data is different from general ad-hoc network in the mobile sensor network.For example: shown in Fig. 3 (a), for from source node R
8To destination node R
12The transmission of monitoring data need to seek a paths.In a router, the Monitoring Data of input need be transformed under the local coordinate system, has the required local coordinate information of next router and binds with the output Monitoring Data again.The Routing Protocol that is proposed be based on distributed Di Lang inner model and sensor network kinematic.
Based on the sensor network kinematics, the relative position of two adjacent nodes is very important as can be seen.The MAC layer and the application layer of wireless network have been proposed locating information is integrated into.As prototype, the existing IEEE802.11MAC layer of our plan modification is so that make it comprise locating information.This method can overcome the problem of hiding in the mobile ad-hoc wireless network with exposed node.Each node of network all has locating information, and when it attempts to obtain the access right of common share communication medium (CTS/RTS exchanges data), propagates this locating information.Node listens to this information, and calculates about the relative position between transmitter and receiver.From locating information, other node of wishing communication can both know its emission whether can disturb right in nodes in communication.If node determines that it is enough far away apart from other communication node, it just begins emission so.Because know all that about the locating information that all existing communications connect the energy of adjusting transmitter so dynamically is also just easy, this adjustment helps making unnecessary interference to drop to minimum.This will cause: 1) because more node can use communication media, so increased the message capacity of channel; 2) because reduced power consumption, so prolonged the life cycle of node.