• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Optimization of communication topology for persistent formation in case of communication faults

    2023-09-05 08:48:42GuoQiangWang王國強HeLuo羅賀XiaoXuanHu胡笑旋andJianWeiTai臺建瑋
    Chinese Physics B 2023年7期
    關(guān)鍵詞:王國強

    Guo-Qiang Wang(王國強), He Luo(羅賀),?, Xiao-Xuan Hu(胡笑旋), and Jian-Wei Tai(臺建瑋)

    1School of Management,Hefei University of Technology,Hefei 230009,China

    2Key Laboratory of Process Optimization&Intelligent Decision-making,Ministry of Education,Hefei 230009,China

    3Engineering Research Center for Intelligent Decision-making&Information Systems Technologies,Ministry of Education,Hefei 230009,China

    Keywords: persistent formation, communication topology, formation communication cost, communication fault

    1.Introduction

    Agent formation has been widely used in many fields, such as large-object transportation,[1]mobile target tracking,[2]and earth observing.[3]To maintain the formation shape, there are mainly the following three formation control methods: position-based,[4,5]displacement-based,[6,7]and distance-based[8,9]methods.In comparison with other methods, the distance-based method has the advantage of requiring no knowledge of the orientations of the agents in global coordinates.[10]Agent formation using distancebased method can be further divided into two categories:rigid formation[11,12]and persistent formation.[13,14]Compared with the rigid formation, the persistent formation only needs to use half of the communication links.[15]

    In a persistent formation, one agent usually acts as the leader and tracks a predetermined reference trajectory, and the other agents receive the information from some of their neighboring agents through communication links to keep the distances between them constant.All the available communication links and the communication links being used to maintain the formation shape constitute the network topology and communication topology of the persistent formation respectively.[16]To ensure that the persistent formation can maintain its formation shape, the communication topology must be a persistent graph of the network topology.Meanwhile,different communication links of each agent may have different communication costs,that is,they consume different energy of the agent, and the available energy of each agent is limited, so it is necessary to optimize the communication topology of a persistent formation to minimize the energy consumption of the agents.

    With the aim of minimizing the sum of the communication costs of all communication links in the communication topology, also called the formation communication cost, the concept of a min-weighted persistent formation[17]or an optimally persistent formation[18]has been proposed.To generate the communication topology of an optimally persistent formation, a generation algorithm based on optimally rigid graph and triangular configuration has been proposed,[17]which is suitable for a specific two-dimensional(2D)formation shape.Two generation algorithms based on optimally rigid graph and several oriented operations have been proposed,[19,20]which are suitable for any 2D and three-dimensional (3D) formation shape respectively.In addition, a generation algorithm based on optimally rigid graph with a top–down strategy has been proposed,[21]which can be applied to some 3D formation shapes.

    Furthermore,considering the leader constraint(only certain agents can act as the leader of the formation), two algorithms called LC-HMAS-OPFGA[16]and CTOA-3DPFLC,[22]based on optimally rigid graph,minimum cost arborescence (MCA), and path reversal operation, have been proposed,which are suitable for any 2D and 3D formation shapes respectively.Then, two corresponding improved algorithms with less calculation time,called FOA-CT-2DPF[23]and FGACT-3DOPF,[24]have also been proposed.

    In summary, the existing studies have mainly addressed the offline optimization problem of communication topology for persistent formation without communication faults.However, owing to hostile attacks, signal jamming or occlusion,mechanical or communication hardware failure and so on,a persistent formation may suffer communication faults during formation movement.These communication faults may cause some communication links in the current communication topology to no longer be usable or some agents to withdraw from the formation.[25]As a result,the formation shape cannot be maintained, and agent collisions may even occur in serious cases.Therefore, it is necessary to optimize the communication topology after such communication faults to ensure the safety of all agents and continue to maintain the formation shape while reducing the formation communication cost as much as possible.Therefore,studied in this work is the optimization problem of communication topology for persistent formation in case of communication faults, and the main novel contributions of this paper are as follows.

    (i) A two-stage optimization model is proposed, which decomposes this problem into two simpler subproblems, that is, the fast reconstruction subproblem of communication topology and the re-optimization subproblem of communication topology.

    (ii) For the first subproblem, a fast reconstruction algorithm of communication topology, based on optimally rigid graph, arc addition operation and path reversal operation, is proposed,which can ensure the safety of the agents in a timely manner and maintain the formation shape.

    (iii) For the second subproblem, a re-optimization algorithm of communication topology,based on agent position exchange,is proposed,which can ensure that the formation communication cost is as low as possible while still maintaining the formation shape.

    The remainder of this article is organized as follows.The problem analysis and modeling are discussed in Section 2.In Section 3,the proposed algorithms are presented in detail,and their time complexities are analyzed.In Section 4, the effectiveness of the proposed algorithms is validated by numerical experiments.Concluding remarks and expectations are presented in Section 5.

    2.Problem analysis and modeling

    First, some basic knowledge about persistent formation is presented.Second,the communication fault constraint of a persistent formation is analyzed.On this basis,a two-stage optimization model is proposed to model the optimization problem of communication topology for persistent formation in case of communication faults.

    2.1.Basic knowledge about persistent formation

    The formation shape, network topology, communication topology, leader constraint, and communication fault constraint of persistent formation are introduced respectively.

    2.1.1.Formation shape of persistent formation

    Thed-dimensional formation shape of a persistent formation composed ofnagents can be represented by ann×3 matrixS,[23,24]as shown below:

    Here, the positions in the formation shape are numbered as 1,2,...,n, and (xi,yi,zi) represent the relative coordinates of thei-th position in a coordinate system.In particular,whendis equal to 2,the value ofzi(1≤i ≤n)is 0.

    2.1.2.Network topology of persistent formation

    The network topology of a persistent formation includes all of the available communication links between agents, and can be represented by a weighted digraphD=(V,A,W,P)as follows.[23,24]

    i)V={vi},1≤i ≤n,is the set of vertices,wherevirepresents thei-th agent.

    ii)A={aij}?V×V,1≤i/=j ≤n, is the set of arcs,whereai j=(vi,vj)denotes that there is a communication link fromvitovjsuch thatvican send information tovj.

    iii)W={wij},aij ∈A, is the set of arc weights, wherewijdenotes the communication cost ofaij,which is equivalent to the communication distance fromvitovj.

    iv)P={pi},1≤i ≤n,is the set of specific positions of agents, referred to as the agent position configuration, wherepirepresents the number of the specific position ofviinS.

    2.1.3.Communication topology of persistent formation

    The communication topology of a persistent formation is a persistent graph in its network topology.It can be expressed asT=(V,A?,W?,P),A??A,W??W,and has the following attributes.[23,24]

    (I)Its underlying undirected graph is a rigid graph.

    (II)aij ∈A?indicates that the communication link fromvitovjis used during the formation movement.

    (IV)|A?| represents the total number of communication links in the communication topology,which must satisfy|A?|≥d×(|V|?3)+3.When|A?|=d×(|V|?3)+3, the communication topology is a minimally persistent graph,and the persistent formation is a minimally persistent formation.

    (V)w(T)=∑ai j∈A?wi jrepresents the formation communication cost.If the correspondingw(T) of the current communication topology is the smallest in all possible persistent graphs in the same network topology,then this communication topology is an optimally persistent graph, and this formation is an optimally persistent formation.

    2.1.4.Leader constraint of persistent formation

    To facilitate agent control,the leader-first-follower(LFF)or leader-remote-follower(LRF)control mode is usually used by a persistent formation to maintain a 2D formation shape.Furthermore, for a heterogeneous persistent formation, only specific agents can act as the leader in the LFF or LRF control mode, and these candidate leaders can be represented byVl ?V.Therefore,the leader constraint of a 2D persistent formation can be described as Eq.(2),[23]in whichviacts as the leader.

    Similarly,the leader-first follower-second follower(L-FF-SF)control mode is usually used by a persistent formation to maintain a 3D formation shape, and the leader constraint of a 3D persistent formation can be described as follows:[24]

    wherevi,vj,andvkact as the leader,the first follower,and the second follower,respectively.

    2.2.Communication fault constraint of persistent formation

    Four specific types of communication faults can be summarized in Table 1.[26]

    Table 1.Four types of communication faults.

    Meanwhile, for all agents to be informed of the same communication fault information in time, we assume that all agents have the same diagnosis and identification strategy of communication faults based on a broadcast communication channel(BC)as indicated in the studies of Giuliettiet al.,[27]Polliniet al.,[28]and Yanget al.[29]

    In summary,different types of communication faults will make certain communication links invalid or cause certain agents to withdraw from the formation.LetA?={axy},1≤x /=y ≤n, represent the set of invalid communication links, and letV?={az}, 1≤z ≤n, denote the set of agents that must be withdrawn from the formation.If the current communication topology contains one or more of these invalid communication links or if one or more of the agents need to be withdrawn from the formation,then the current communication topology can no longer be used to maintain the formation shape,and the communication topology must be adjusted.Therefore, the communication fault constraint of a persistent formation can be described as follows:

    That is, the adjusted communication topology cannot contain any invalid communication links,nor can it contain any agent that needs to withdraw from the formation.

    2.3.Problem modeling

    When a communication fault prevents the current communication topology from being used to maintain the formation shape, if the communication topology is not adjusted in time, this communication fault may lead to collision accidents between agents.[25]Therefore, after a communication fault occurs,the communication topology must be quickly reconstructed to ensure the safety of the agents and restore the formation shape.After the communication topology reconstruction, the security of the persistent formation is ensured.At this time,the communication topology can be reoptimized via agent position exchange(e.g.,exchanging the positions of two agents in the formation or having one agent filling in the space left by another agent exiting the formation) to achieve the minimal formation communication cost required to continue to maintain the formation shape.Based on the above analysis, the optimization problem of communication topology for persistent formation in case of communication faults can be decomposed into the following two subproblems.

    I) Fast reconstruction subproblem of communication topology for persistent formation

    First,the corresponding invalid arcs or invalid vertices are removed from the current network topologyD=(V,A,W,P)to obtain a reconstructed network topologyDr=(Vr,Ar,Wr,Pr)that satisfies the communication fault constraint; then, a suitable subgraphTrneeds to be quickly selected fromDras the reconstructed communication topology.

    The reconstructed communication topology must be a persistent graph inDr; however, there may be more than one graph like this inDr,of which,the optimally persistent graph has the lowest formation communication cost.Moreover, because only certain agents can serve as the leader, it may not be possible to use any arbitrary optimally persistent graph inDras the reconstructed communication topology.Therefore,this problem is modeled as an optimization problem of obtaining the optimally persistent graph inDrthat satisfies the leader constraint and the communication fault constraint.

    II)Re-optimization subproblem of communication topology for persistent formation

    Since different position exchanges will result in different agent position configurationsPmand there are multiple feasible communication topologies for eachPm, in this work this problem is modeled as a combinatorial optimization problem in which the leader constraint and the communication fault constraint must be satisfied: the objective is to select an optimal position configurationPofrom the set of all possible agent position configurations{Pm}such that the formation communication cost of the corresponding optimal communication topologyTois globally minimum.

    Specifically, the network topologyDmthat satisfies the communication fault constraint is constructed for eachPm,and then,the optimally persistent graphTmthat satisfies the leader constraint inDmis obtained.Finally,the graph with the lowest formation communication cost is selected as the reoptimized communication topologyTofrom among allTm, and the corresponding agent position configurationPois the reoptimized agent position configuration.

    The necessary and sufficient conditions for an optimally persistent graph are shown in Lemma 1 as follows.

    Lemma 1[16,22]A weighted digraphD=(V,A,W) is ad-dimensional optimally persistent graph if and only if it satisfies the following two conditions:

    (i)The in-degree of each vertex is no greater thand.

    (ii) The corresponding weighted undirected graph,G=(V,E,W),is ad-dimensional optimally rigid graph.

    The necessary and sufficient conditions for an optimally rigid graph, as mentioned in condition 2 of Lemma 1, are shown in Lemma 2 as follows.

    Lemma 2[16,22]A weighted undirected graphG=(V,E,W)is ad-dimensional optimally rigid graph if and only if it satisfies the following three conditions:

    i)The number of edges,|E|,isd×(|V|?3)+3.

    ii)The corresponding rigidity matrix,M ∈R|E|×(3×|V|),is full rank,that is,the rank ofMisd×(|V|?3)+3.

    iii)The sum of the weights of all edges inEis the smallest in all weighted undirected graphs that have the same vertices and satisfy conditions 1 and 2.

    3.Model-solving algorithms

    For solving the above two subproblems, a fast reconstruction algorithm of communication topology for persistent formation (FRA-CT-PF), based on optimally rigid graph, arc addition operation and path reversal operation, is proposed.Then,a re-optimization algorithm of communication topology for persistent formation (ROA-CT-PF), based on agent position exchange, is proposed.The details of these two algorithms are described in the following subsections.

    3.1.FRA-CT-PF

    The algorithm first deletes invalid arcs and vertexes from current network topology to obtain a reconstructed network topology, then generates an optimally rigid graph of this reconstructed network topology by considering the communication fault constraint,and transforms the optimally rigid graph into an optimally persistent graph satisfying the communication fault constraint through converting into a digraph, obtaining the intersection with current communication topology, arc addition and path reversal operations.Finally, the algorithm makes the optimally persistent graph satisfy both the leader constraint and the communication fault constraint through path reversal operation.The pseudocode of FRA-CTPF is shown in Table 2.

    Next, the correctness of FRA-CT-PF is proved theoretically.

    Theorem 1The reconstructed communication topologyTrobtained by FRA-CT-PF must be the optimally persistent graph in the reconstructed network topologyDrthat satisfies the leader constraint and the communication fault constraint.

    ProofBecauseTis ad-dimensional optimally persistent graph ofD,the in-degree of each vertex inTis no greater thandaccording to Lemma 1.Therefore,after the operation of line 09,the in-degree of each vertex inTris also no greater thand.The premise of the operation of line 15 is(vj)

    After the operation of line 08,DRris the digraph of optimally rigid graphRr.After the operation of line 09,Tris a subgraph ofDRrand no duplicate arcs exist inTr.In the operations of lines 11 to 27, all arcs added toTrmust belong toDRrand no duplicate arcs are added toTr.Meanwhile, after the operations of lines 11 to 27,m=d×(|Vr|?3)+3.In the operations of lines 28 to 36, only the direction of some arcs inTris changed.Therefore,the corresponding weighted undirected graph of the finalTrmust be the optimally rigid graphRr.That is,Trobtained by FRA-CT-PF must satisfy condition 2 of Lemma 1.

    Because all conditions of Lemma 1 are satisfied,Trobtained by FRA-CT-PF must be ad-dimensional optimally persistent graph inDr.

    After the operation of lines 28 to 36,d=2 and equation (2) are satisfied, ord=3 and equation (3) are satisfied.Therefore,Trobtained by FRA-CT-PF must satisfy the leader constraint.

    After the operation of line 03,A?andV?are deleted fromDr,which means thatDrsatisfies the communication fault constraint.After the operations of lines 08 to 09,DRris the subgraph ofDrandTris the subgraph ofDRr, soTrsatisfies the communication fault constraint.In the operations of lines 11 to 36,all arcs added toTrmust belong toDRrand all the reverse paths must exist inDRr.Therefore,Trobtained by FRA-CT-PF must satisfy the communication fault constraint.

    In summary,Trobtained by FRA-CT-PF must be the optimally persistent graph inDrthat satisfies the leader constraint and the communication fault constraint.

    The time complexity of FRA-CT-PF is analyzed as follows.In this algorithm,the main calculation time is in lines 07,19,22,31,33,and 34,respectively.Line 07 is an improvement on the original optimally rigid graph generation algorithm,[30]and its time complexity is approximatelyO(|Vr|4).Lines 19,22, 31, 33, and 34 use the classical Dijkstra shortest path algorithm, and the time complexity of each of them is approximatelyO(||+|Vr|×log|Vr|).Because|Vr|≤|V| and||≤|Ar|≤|V|×(|V|?1),the total time complexity of FRACT-PF is aboutO(|Vr|4+4×(||+|Vr|×log|Vr|))≈O(|V|4).

    3.2.ROA-CT-PF

    To obtain a reoptimized communication topology through agent position exchange,the following solution method can be adopted.First, for each agent position configurationPm, the network topologyDmthat satisfies the communication fault constraint is constructed.Then,the optimally persistent graphTmthat satisfies the leader constraint and the communication fault constraint in eachDm, is obtained.Finally, the graph with the lowest formation communication cost is selected from among allTmas the reoptimized communication topology,To.

    However, eachPmmust be a permutation of the selection of|Vr| positions from among the|V| positions in the formation shape, so the total number of all possiblePmis|V|!/(|V|?|Vr|)!.As the value of|V| continues to increase,the corresponding value of|V|!/(|V|?|Vr|)! becomes very large,indicating that it is difficult to obtain a solution in a limited time.Therefore, ROA-CT-PF is proposed in this work to achieve a suboptimal reoptimized communication topology through position exchange between only two agents.Nevertheless, it can still obtain a good reoptimized communication topology, as verified in the following numerical experiments.The pseudocode of this algorithm is shown in Table 3.

    Table 3.Pseudocode of ROA-CT-PF.

    Next, the correctness of ROA-CT-PF is proved theoretically.

    Theorem 2The reoptimized communication topologyToobtained by ROA-CT-PF must be feasible and its formation communication cost is no more than that ofTrobtained by FRA-CT-PF.

    ProofIn the operation of line 01,Tois initialized by the reconstructed communication topologyTrobtained by FRACT-PF.According to Theorem 1,Trmust be feasible, so the initialTois feasible.In the operation of line 08, FRA-CT-PF is used to obtain the corresponding optimal reconstructed communication topologyTm.According to Theorem 1,Tmmust be feasible.In the operations of lines 10 and 12,Tois replaced byTm.Therefore,Toobtained by ROA-CT-PF must be feasible.

    In the operation of line 01,Tois initialized by the reconstructed communication topologyTrobtained by FRA-CT-PF,so the formation communication cost of the initialTois equal to that ofTr.In the operations of lines 09 to 13, only whenw(Tm)≤w(To)Tocan be replaced byTm.Therefore,the formation communication cost ofToobtained by ROA-CT-PF is no more than that ofTrobtained by FRA-CT-PF.

    The time complexity of ROA-CT-PF is analyzed as follows.Line 08 is the core step of this algorithm and is the same as lines 06 to 36 of FRA-CT-PF;therefore,the time complexity of line 08 is approximatelyO(|V|4).In line 06 the number of possible combinations of two positions selected from among the|V|total positions is|V|×(|V|?1)/2; that is, the total number of all feasiblePmis|V|×(|V|?1)/2.Therefore,it can be seen from line 05 of this algorithm that line 08 will run|V|×(|V|?1)/2 times at most.Consequently, because|Vr|≤|V|, the time complexity of this algorithm is approximatelyO(|V|4×|V|×(|V|?1)/2)≈O(|V|6).

    4.Numerical experiments

    Next, several numerical experiments are carried out to verify the effectiveness of the proposed algorithms.

    4.1.Experimental settings

    Assuming that there are no communication faults at the beginning,all agents are able to communicate with each other,viis located at thei-th position inSand onlyv1,v2,andv4can serve as leaders,so the initial network topology is shown in Fig.2(a)and its corresponding Laplace matrix is

    The initial communication topology of this persistent formation,obtained by CTOA-PF-LC,[22]is shown in Fig.2(b).Its corresponding Laplace matrix is

    Fig.2.Initial network topology and communication topology of this persistent formation,showing(a)initial network topology and(b)initial communication topology.

    In this communication topology,v1,v2, andv5are the leader, the first follower, and the second follower, respectively,and the formation communication cost is 11163,whileP={1,2,3,4,5}is the corresponding agent position configuration.To easily distinguish the in-degree of each vertex in figures,arcs from small-to large-numbered vertices are drawn with blue solid lines, and other arcs are drawn with magenta solid lines.

    To verify the effectiveness of proposed algorithms under various types of communication faults in Table 1, eight communication fault scenarios as shown in Table 4, are designed and the corresponding numerical experiments are carried out.Among these scenarios,because the influence of a transmitter or transceiver failure on a leader in the communication topology is very different from that of a transmitter or transceiver failure on a non-leader,scenarios 2 and 3 are designed to represent transmitter failures on a leader and a non-leader, and scenarios 5 and 6 are designed to represent transceiver failures on a leader and a non-leader, respectively.In addition, scenarios 7 and 8 are designed to represent transceiver failures on two agents.

    The development environment of all algorithms is Visual Studio 2013, and the programming language is C++.All experiments are performed on a computer with an 8-core i7 4.0-GHz processor,16-GB RAM and a Windows 10 operating system.

    4.2.Experimental results of fast reconstruction of communication topology

    First, scenario 3 is used to illustrate the calculation process of FRA-CT-PF.It is assumed thatv5suffers a transmitter failure based on the current network topology shown in Fig.2(a) and the current communication topology shown in Fig.2(b).At the same time, onlyv1,v2, andv4can serve as leaders.Namely,the communication fault constraint is denoted byA?={a5y},1≤y ≤6 andV?= /0, and the leader constraint is represented byVl={v1,v2,v4}.Then,the main calculation process of this algorithm is as follows.

    In line 02, the invalid arcs and vertexes are deleted from current network topology,and the reconstructed network topologyDris shown in Fig.3(a).Its corresponding Laplace matrix is

    In line 07,the optimally rigid graphRrofGris generated as shown in Fig.3(b),which satisfies that the degree of each vertex inRris no more than the number of bidirectional arcs of this vertex inGrplus 3.This additional condition ensures thatRrcan be directed to an optimally persistent graph satisfying the communication fault constraint.

    In line 08,Rris converted into a digraphDRrsatisfying the communication fault constraint,and it is shown in Fig.3(c).Its corresponding Laplace matrix is

    In line 09,the intersection ofDRrandTis obtained and used to reinitializeTr,as shown in Fig.3(d).

    Fig.3.Calculation process of FRA-CT-PF under scenario 3,showing(a)constructed network topology,(b)optimally rigid graph Rr,(c)digraph corresponding to Rr,(d)intersection Tr of DRr and T,(e)Tr after adding a14,(f)Tr after adding a15,(g)Tr after adding a16,(h)Tr after adding a25,(i)Tr after reversing(1,4),(j)Tr after reversing(6,3),(k)Tr after reversing(6,2),and(l)reconstructed communication topology.

    In line 10,the number of its arcs,m,is 8.

    In line 11,becausem<3×|Vr|?6,the operations in lines 12 to 26 will be performed untilm=3×|Vr|?6.The specific process is as follows: whenl=1,thel-th edgeei jinRrise14,neithera14nora41exists inTr,d(v4)=0<3 anda14∈DRr,soa14is added toTrin line 15 as shown in Fig.3(e).Similarly, whenl=2,a15is added toTras shown in Fig.3(f).Whenl=3,a16is added toTras shown in Fig.3(g).Whenl=4,a25is added toTras shown in Fig.3(h).

    In line 28, because the leader constraint is not satisfied,jumps directly to line 31.

    In line 31,v4is found to be the leader because it satisfiesv4∈Vl.However, becaused(v4)=1 now, a path (1,4) is found inTrand its arcs are reversed inTras shown in Fig.3(i),where the reversed path is identified by dashed lines.After this path reverse operation,Trsatisfiesd(v4)=0 andTr?DRr.Therefore,v4acts as the leader.

    In line 33,v3is found because it satisfiesa43∈.However,becausedr(v3)=2 now,a path(6,3)is found inTrand its arcs are reversed inTras shown in Fig.3(j), where the reversed path is identified by dashed lines.After this path reverse operation,Trsatisfiesa43∈,d(v3)=1 andTr?DRr.Therefore,v3acts as the first follower.

    In line 34,v2is found because it satisfiesa42,a32∈.However, because(v2)=3 now, a path (6,2) is found inTrand its arcs are reversed inTras shown in Fig.3(k),where the reversed path is identified by dashed lines.After this path reverse operation,Trsatisfiesa42,a32∈,(v2)=2 andTr?DRr.Therefore,v2acts as the second follower.

    In line 37,Tris the reconstructed communication topology as shown in Fig.3(l), and the corresponding formation communication cost is 11903.1.Its corresponding Laplace matrix is

    Comparatively,if CTOA-3DPF-LC[22]is used to acquire the reconstructed communication topology in this case,the finalTrobtained is shown in Fig.4.However, the arcsa53,a54, anda56exist in the finalTr, which means that the finalTrdoes not satisfy the communication fault constraint.Therefore, the reconstructed communication topology obtained by CTOA-3DPF-LC is infeasible.

    Fig.4.Reconstructed communication topology obtained by CTOA-3DPF-LC.

    Similarly,if FGA-CT-3DOPF[24]is used to obtain the reconstructed communication topology in this case, the finalTrobtained is shown in Fig.5.However, it does not satisfy the condition 1 of Lemma 2.Therefore, the reconstructed communication topology obtained by FGA-CT-3DOPF is also infeasible.

    Fig.5.Reconstructed communication topology obtained by FGA-CT-3DOPF.

    Table 5.Reconstructed communication topologies and calculation time under eight scenarios.

    Scenario FRA-CT-PF CTOA-3DPF-LC FGA-CT-3DOPF

    Table 5.(Continued).

    Next, for the eight communication fault scenarios in Table 4, the reconstructed communication topologies obtained by FRA-CT-PF and existing algorithms, and the calculation time of these algorithms are shown in Table 5.

    In scenario 1, the reconstructed communication topologies obtained by FRA-CT-PF and FGA-CT-3DOPF are both feasible and have the same formation communication cost.However,the reconstructed communication topology obtained by CTOA-3DPF-LC is infeasible because it does not satisfy the communication fault constraint.In scenarios 2 and 4, the same conclusion can be drawn.

    In scenario 3, the reconstructed communication topology obtained by FRA-CT-PF is feasible.However,the reconstructed communication topology obtained by CTOA-3DPFLC is infeasible because it does not satisfy the communication fault constraint.Meanwhile,the reconstructed communication topology obtained by FGA-CT-3DOPF is infeasible because it does not satisfy the condition 1 of Lemma 2.

    In scenario 5, the reconstructed communication topologies obtained by FRA-CT-PF,CTOA-3DPF-LC and FGA-CT-3DOPF are both feasible and have the same formation communication cost.In scenarios 6,7,and 8,the same conclusion can be drawn.

    Meanwhile,in all scenarios,the average calculation time of FRA-CT-PF is only 3.48%of that of CTOA-3DPF-LC,but only 1.02%of that of FGA-CT-3DOPF.

    In summary, compared with existing algorithms, FRACT-PF proposed in this paper can always obtain feasible reconstructed communication topology in all communication fault scenarios,and its calculation time is much less.

    4.3.Experimental results of re-optimization of communication topology

    First,scenario 3 continues to be used to illustrate the calculation process of ROA-CT-PF.Based on the reconstructed communication topology as shown in Fig.3(l),the calculation process of ROA-CT-PF is as follows.

    In line 01,Tois equal toTrandPois equal toPr.

    In line 02,becausew(To)=11903.1 andw(T)=11163,the algorithm executes the operations of lines 05 to 14.

    Whenm= 2, the correspondingP2in line 06 is{2,1,3,4,5,6}and the correspondingT2in line 08 is shown in Fig.6(a),wherev1andv2exchange their positions in the formation shape andw(T2)=11903.1.Becausew(T2)=w(To)and the total agent movement distance for switching fromPrtoP2is 1562.05, which is more than that fromPrtoPo, there is no change inTonorPo.

    Whenm= 3,P3is{3,2,1,4,5,6}andT3is shown in Fig.6(b),wherev1andv3exchange their positions in the formation shape andw(T3)=11903.1.Becausew(T3)=w(To)and the total agent movement distance for switching fromPrtoP3is equal to that fromPrtoPo,there is no change inToandPo,either.

    Whenm= 4,P4is{4,2,3,1,5,6}andT4is shown in Fig.6(c),wherev1andv4exchange their positions in the formation shape andw(T4)=11903.1.Becausew(T4)=w(To)and the total agent movement distance for switching fromPrtoP4is 3072.46, which is more than that fromPrtoPo, there is also no change inTonorPo.

    Whenm= 5,P5is{5,2,3,4,1,6}andT5is shown in Fig.6(d),wherev1andv5exchange their positions in the formation shape andw(T5)=11163.Becausew(T5)

    Fromm=6 to the end,the conditions in line 09 or line 11 are not satisfied, so there is no change inTonorPo.Namely,T5is the reoptimized communication topology, wherev1andv5exchange their positions.

    Fig.6.Example of calculation process of ROA-CT-PF, showing: (a) reconstructed communication topology when m=2, (b) reconstructed communication topology when m=3,(c)reconstructed communication topology when m=4,and(d)reoptimized communication topology.

    In contrast, lines 05 and 06 of ROA-CT-PF are modified into“whilem<|V|!/(|V|?|Vr|)!do”and“Letm=m+1,and obtain the next permutationPmof the selection of|Vr|positions from among the|V|positions in the formation shape”,and this modified algorithm is called ROA-CT-PF-P.Then, ROA-CTPF-P is used to obtain the reoptimized communication topology in this case.TheToobtained by ROA-CT-PF-P is the same as that obtained by ROA-CT-PF, but the calculation time of ROA-CT-PF-P is much longer than that of ROA-CT-PF.

    Next, for the eight communication fault scenarios in Table 4,the reoptimized communication topologies obtained by FRA-CT-PF and existing algorithms,and the calculation time of these algorithms are shown in Table 6.

    In scenario 1,because the formation communication cost of reconstructed communication topology is more than that of current communication topology, the operations of agent position exchange in ROA-CT-PF and ROA-CT-PF-P will be executed.The reoptimized communication topologies obtained by ROA-CT-PF and ROA-CT-PF-P are the same.However,the calculation time of ROA-CT-PF is much less than that of ROA-CT-PF-P.Meanwhile, both CTOA-3DPF-LC and FGACT-3DOPF do not support the re-optimization of communication topology for persistent formation.In scenarios 3,5,6,and 7,the same conclusion can be drawn.

    In scenario 2,because the formation communication cost of reconstructed communication topology is equal to that of current communication topology, the operations of agent position exchange in ROA-CT-PF and ROA-CT-PF-P will not be executed.That is,the reoptimized communication topologies obtained by ROA-CT-PF and ROA-CT-PF-P are the same as the reconstructed communication topology, and the calculation time of ROA-CT-PF and ROA-CT-PF-P are both approximately equal to 0.In scenario 4,the same conclusion can be drawn.

    Table 6.Reoptimized communication topologies and calculation time under eight scenarios

    Table 6.(Continued).

    In scenario 8,because the formation communication cost of reconstructed communication topology is more than that of current communication topology, the operations of agent position exchange in ROA-CT-PF and ROA-CT-PF-P will be executed.The formation communication cost of the reoptimized communication topology obtained by ROA-CT-PF is more than that by ROA-CT-PF-P.This shows that ROA-CTPF cannot ensure obtaining the optimal reoptimized communication topology through position exchange between only two agents.However,the calculation time of ROA-CT-PF is much less than that of ROA-CT-PF-P.

    In summary,the following conclusions can be drawn from Table 6:

    (i) Compared with CTOA-3DPF-LC and FGA-CT-3DOPF,the ROA-CT-PF proposed in this paper can reoptimize the communication topology to further reduce the formation communication cost based on the reconstructed communication topology.

    (ii)Compared with ROA-CT-PF-P,the ROA-CT-PF proposed in this paper can almost obtain the same reoptimized communication topology in all communication fault scenarios,and the calculation time is much less.

    5.Conclusion and perspectives

    In this paper, the optimization problem of communication topology for persistent formation in case of communication faults is studied.This problem is decomposed into two subproblems, namely, the fast reconstruction subproblem of communication topology and the re-optimization subproblem of communication topology.

    To address the fast reconstruction subproblem of communication topology, an algorithm called FRA-CT-PF, based on optimally rigid graph,arc addition operation and path reversal operation, is proposed.Compared with existing algorithms,the algorithm can always obtain feasible reconstructed communication topology in all communication fault scenarios,and its calculation time is much less.

    To address the re-optimization subproblem of communication topology, an algorithm called ROA-CT-PF, based on agent position exchange,is proposed.Compared with existing algorithms,the algorithm can obtain a reoptimized communication topology to further reduce the formation communication cost in a shorter time.

    In the future,a distributed optimization algorithm of communication topology for persistent formation in case of communication faults will be studied.

    Acknowledgements

    Project supported by the National Natural Science Foundation of China(Grant Nos.71871079,72271076,71971075,and 71671059) and the Anhui Provincial Natural Science Foundation,China(Grant No.1808085MG213).

    猜你喜歡
    王國強
    王國強作品選
    外逃原市委書記:給自己定了個刑期
    新傳奇(2019年40期)2019-11-01 01:30:32
    回國投案的市委書記
    檢察風云(2017年9期)2017-05-18 07:42:39
    現(xiàn)實版“丁義珍”
    ——外逃是死路一條
    新傳奇(2017年17期)2017-01-30 07:13:58
    Some Thoughts on Proposing a Customer Satisfaction Strategy in Education
    美國驚魂,市委書記海外逃亡始末
    亚洲欧洲国产日韩| 岛国毛片在线播放| 欧美精品人与动牲交sv欧美| 亚洲综合色网址| 国产精品女同一区二区软件| 秋霞在线观看毛片| 亚洲国产精品999| 亚洲欧美一区二区三区黑人 | 美女国产视频在线观看| 国产免费视频播放在线视频| 18禁观看日本| 黄色一级大片看看| 哪个播放器可以免费观看大片| 欧美精品av麻豆av| 亚洲国产色片| 在线观看一区二区三区激情| 国产熟女欧美一区二区| 一本色道久久久久久精品综合| 国产免费一区二区三区四区乱码| 久久人人97超碰香蕉20202| 亚洲久久久国产精品| 亚洲一码二码三码区别大吗| 又粗又硬又长又爽又黄的视频| 性色avwww在线观看| 内地一区二区视频在线| 国产男女超爽视频在线观看| 视频区图区小说| 亚洲精品久久午夜乱码| 亚洲欧美日韩另类电影网站| 人妻 亚洲 视频| 精品熟女少妇av免费看| 一边亲一边摸免费视频| 熟女人妻精品中文字幕| 久久久欧美国产精品| 精品福利永久在线观看| av又黄又爽大尺度在线免费看| 男男h啪啪无遮挡| 秋霞在线观看毛片| 久久久久精品人妻al黑| 欧美日韩精品成人综合77777| 精品第一国产精品| 免费黄色在线免费观看| 色5月婷婷丁香| 最新的欧美精品一区二区| √禁漫天堂资源中文www| 亚洲四区av| av线在线观看网站| 一级毛片黄色毛片免费观看视频| 一级片'在线观看视频| 欧美国产精品va在线观看不卡| 欧美成人午夜免费资源| 最近中文字幕2019免费版| 精品国产一区二区久久| 国语对白做爰xxxⅹ性视频网站| 久久久久久人妻| 国产亚洲一区二区精品| 午夜福利在线观看免费完整高清在| 人人妻人人澡人人看| 免费观看性生交大片5| 成人二区视频| 亚洲久久久国产精品| 如日韩欧美国产精品一区二区三区| 人人澡人人妻人| 久久99热6这里只有精品| 国产又色又爽无遮挡免| 国产极品天堂在线| 免费观看性生交大片5| 国产精品蜜桃在线观看| 人人妻人人澡人人爽人人夜夜| 视频区图区小说| 大片免费播放器 马上看| 99久久人妻综合| 五月开心婷婷网| 日韩精品有码人妻一区| av卡一久久| 自线自在国产av| 大香蕉久久网| 十分钟在线观看高清视频www| 午夜福利影视在线免费观看| 国产日韩欧美在线精品| 午夜福利网站1000一区二区三区| 一级毛片电影观看| 久久99热6这里只有精品| 青青草视频在线视频观看| 美女福利国产在线| 婷婷成人精品国产| 国产成人精品在线电影| 99热网站在线观看| 黄色视频在线播放观看不卡| 少妇被粗大的猛进出69影院 | 亚洲,一卡二卡三卡| 久热久热在线精品观看| 成人亚洲精品一区在线观看| 日本vs欧美在线观看视频| 国产免费一区二区三区四区乱码| 少妇精品久久久久久久| 美女大奶头黄色视频| 国产亚洲午夜精品一区二区久久| 激情视频va一区二区三区| 中国美白少妇内射xxxbb| 日产精品乱码卡一卡2卡三| 成人影院久久| 蜜桃国产av成人99| 少妇高潮的动态图| 亚洲国产欧美在线一区| 欧美xxxx性猛交bbbb| 成人午夜精彩视频在线观看| 欧美精品一区二区免费开放| 黄片播放在线免费| 黄色视频在线播放观看不卡| 亚洲精品第二区| 青春草亚洲视频在线观看| 日本av手机在线免费观看| 少妇 在线观看| 国产日韩欧美亚洲二区| 亚洲丝袜综合中文字幕| www.av在线官网国产| 久久久亚洲精品成人影院| 我的女老师完整版在线观看| 国产精品久久久久久av不卡| 麻豆乱淫一区二区| 久久狼人影院| 肉色欧美久久久久久久蜜桃| 人人妻人人澡人人看| 啦啦啦中文免费视频观看日本| 国产无遮挡羞羞视频在线观看| 久久久欧美国产精品| 亚洲精品视频女| 97在线视频观看| 国产精品嫩草影院av在线观看| 亚洲国产精品国产精品| kizo精华| 精品熟女少妇av免费看| 十分钟在线观看高清视频www| 国产精品久久久久久av不卡| 婷婷色av中文字幕| 亚洲国产毛片av蜜桃av| 日韩电影二区| 国产精品一区二区在线观看99| 一级片免费观看大全| 亚洲欧洲国产日韩| 中文字幕最新亚洲高清| 国内精品宾馆在线| 晚上一个人看的免费电影| 少妇的逼水好多| 亚洲,欧美,日韩| 高清黄色对白视频在线免费看| 精品一区二区免费观看| 自拍欧美九色日韩亚洲蝌蚪91| 国产精品欧美亚洲77777| 免费观看性生交大片5| 亚洲欧美清纯卡通| 青春草国产在线视频| 男女午夜视频在线观看 | 2022亚洲国产成人精品| 欧美日韩精品成人综合77777| 亚洲精品乱久久久久久| 亚洲综合精品二区| 99久久人妻综合| 捣出白浆h1v1| 精品国产露脸久久av麻豆| 国产 一区精品| 自拍欧美九色日韩亚洲蝌蚪91| 好男人视频免费观看在线| 亚洲五月色婷婷综合| 欧美成人精品欧美一级黄| 男女午夜视频在线观看 | 黑人高潮一二区| 国产精品女同一区二区软件| 国产老妇伦熟女老妇高清| 999精品在线视频| 国产xxxxx性猛交| 欧美+日韩+精品| 男女下面插进去视频免费观看 | 一边摸一边做爽爽视频免费| 亚洲伊人色综图| 夜夜爽夜夜爽视频| 91精品伊人久久大香线蕉| 一二三四在线观看免费中文在 | 久久久欧美国产精品| 亚洲激情五月婷婷啪啪| 制服诱惑二区| 一区二区三区四区激情视频| av在线app专区| 国产男女超爽视频在线观看| 又大又黄又爽视频免费| 一级毛片我不卡| 国产无遮挡羞羞视频在线观看| 日韩一区二区视频免费看| 99国产综合亚洲精品| 天堂中文最新版在线下载| 大陆偷拍与自拍| 国产极品天堂在线| 久久久久精品久久久久真实原创| 男女国产视频网站| 老司机影院成人| 欧美成人午夜免费资源| 少妇被粗大的猛进出69影院 | 久久精品aⅴ一区二区三区四区 | 国产成人一区二区在线| 亚洲精品久久久久久婷婷小说| 亚洲人成77777在线视频| 欧美日韩综合久久久久久| 黄色一级大片看看| 日韩成人伦理影院| 五月开心婷婷网| 国产麻豆69| 自拍欧美九色日韩亚洲蝌蚪91| 毛片一级片免费看久久久久| 成年人午夜在线观看视频| 午夜福利影视在线免费观看| 最近的中文字幕免费完整| 亚洲精品国产av蜜桃| 国产高清三级在线| 高清视频免费观看一区二区| 秋霞伦理黄片| 久久久欧美国产精品| 亚洲精品日韩在线中文字幕| 国产精品三级大全| 人妻少妇偷人精品九色| 男女无遮挡免费网站观看| 久久久久久久久久久免费av| 精品久久久久久电影网| videossex国产| 一个人免费看片子| 日本黄大片高清| 中文精品一卡2卡3卡4更新| 亚洲国产色片| 午夜视频国产福利| 一级片免费观看大全| 一级毛片电影观看| 久热久热在线精品观看| 亚洲在久久综合| 亚洲激情五月婷婷啪啪| 2018国产大陆天天弄谢| av在线播放精品| 97人妻天天添夜夜摸| 老女人水多毛片| 80岁老熟妇乱子伦牲交| 在线看a的网站| 少妇的逼好多水| 国产成人aa在线观看| 另类亚洲欧美激情| 如日韩欧美国产精品一区二区三区| 91在线精品国自产拍蜜月| 欧美成人精品欧美一级黄| 精品国产一区二区三区久久久樱花| 七月丁香在线播放| 成人无遮挡网站| 美女中出高潮动态图| 超色免费av| 久久青草综合色| 中文字幕av电影在线播放| 免费黄网站久久成人精品| 亚洲成色77777| 国产精品国产三级国产av玫瑰| 久久久久网色| 国产午夜精品一二区理论片| 国产黄色免费在线视频| 9色porny在线观看| 女人被躁到高潮嗷嗷叫费观| 国产一区二区激情短视频 | 久久久久久久国产电影| 不卡视频在线观看欧美| 你懂的网址亚洲精品在线观看| 夫妻午夜视频| 高清欧美精品videossex| 国产一区二区在线观看日韩| 国产精品国产三级专区第一集| 一区在线观看完整版| 美女福利国产在线| av在线老鸭窝| 亚洲伊人久久精品综合| 九九爱精品视频在线观看| 汤姆久久久久久久影院中文字幕| 亚洲欧美中文字幕日韩二区| 亚洲第一区二区三区不卡| 日日爽夜夜爽网站| 欧美日本中文国产一区发布| 美女xxoo啪啪120秒动态图| 免费人成在线观看视频色| 亚洲美女视频黄频| 女人被躁到高潮嗷嗷叫费观| 男人操女人黄网站| 在线观看www视频免费| 免费女性裸体啪啪无遮挡网站| 久久97久久精品| 国产不卡av网站在线观看| 亚洲一级一片aⅴ在线观看| 免费黄网站久久成人精品| 欧美精品一区二区免费开放| 最近2019中文字幕mv第一页| 免费女性裸体啪啪无遮挡网站| 国产精品久久久久久av不卡| 国产精品.久久久| 国产高清国产精品国产三级| 日韩精品免费视频一区二区三区 | 女的被弄到高潮叫床怎么办| 日产精品乱码卡一卡2卡三| 在线 av 中文字幕| 免费观看在线日韩| 亚洲欧美清纯卡通| 中文字幕另类日韩欧美亚洲嫩草| 精品国产国语对白av| 国产一区二区三区综合在线观看 | 一本一本久久a久久精品综合妖精 国产伦在线观看视频一区 | 精品久久国产蜜桃| 久久精品熟女亚洲av麻豆精品| 色婷婷av一区二区三区视频| 亚洲av在线观看美女高潮| 国产一区二区三区av在线| 亚洲五月色婷婷综合| 国产免费一级a男人的天堂| 亚洲婷婷狠狠爱综合网| 久久女婷五月综合色啪小说| 啦啦啦中文免费视频观看日本| 日韩成人伦理影院| 美女xxoo啪啪120秒动态图| 国产av一区二区精品久久| 国产男人的电影天堂91| 亚洲三级黄色毛片| 欧美精品国产亚洲| 久久婷婷青草| 日韩av免费高清视频| 老女人水多毛片| 极品人妻少妇av视频| 中文字幕免费在线视频6| 看免费av毛片| 久久亚洲国产成人精品v| 成人国产av品久久久| 91精品国产国语对白视频| 精品人妻熟女毛片av久久网站| 午夜激情久久久久久久| 亚洲丝袜综合中文字幕| 免费av中文字幕在线| 国产又爽黄色视频| 日韩中文字幕视频在线看片| 精品亚洲成a人片在线观看| av免费观看日本| 国产无遮挡羞羞视频在线观看| 蜜臀久久99精品久久宅男| 亚洲人与动物交配视频| 中文天堂在线官网| 久久久久网色| 精品人妻偷拍中文字幕| 黄色毛片三级朝国网站| 日韩一区二区视频免费看| 男人操女人黄网站| 婷婷色综合www| 亚洲精品一区蜜桃| 天堂中文最新版在线下载| 久久久精品94久久精品| 日韩制服丝袜自拍偷拍| 久久99蜜桃精品久久| 欧美 亚洲 国产 日韩一| 日韩熟女老妇一区二区性免费视频| 午夜久久久在线观看| 人妻 亚洲 视频| 久久av网站| 亚洲精品色激情综合| 国产精品久久久久久精品古装| 成人亚洲精品一区在线观看| √禁漫天堂资源中文www| 内地一区二区视频在线| 国产黄色免费在线视频| 国语对白做爰xxxⅹ性视频网站| 久久久久精品人妻al黑| 桃花免费在线播放| 国产69精品久久久久777片| 日韩一本色道免费dvd| 最后的刺客免费高清国语| 亚洲性久久影院| 国产成人精品婷婷| 久久精品国产鲁丝片午夜精品| 午夜av观看不卡| 国产片特级美女逼逼视频| 免费看光身美女| 精品一区在线观看国产| 国产一区有黄有色的免费视频| 国产精品偷伦视频观看了| 久久综合国产亚洲精品| 热re99久久精品国产66热6| 九九在线视频观看精品| 黄色怎么调成土黄色| 久久久欧美国产精品| 国产国拍精品亚洲av在线观看| 国产乱人偷精品视频| 看非洲黑人一级黄片| 丰满饥渴人妻一区二区三| 欧美日韩综合久久久久久| 亚洲精品av麻豆狂野| 国产精品国产av在线观看| 免费不卡的大黄色大毛片视频在线观看| 晚上一个人看的免费电影| 亚洲av国产av综合av卡| 汤姆久久久久久久影院中文字幕| 亚洲内射少妇av| 伊人久久国产一区二区| av免费观看日本| 熟女人妻精品中文字幕| 波野结衣二区三区在线| 亚洲av电影在线观看一区二区三区| 男人添女人高潮全过程视频| av免费在线看不卡| 在线亚洲精品国产二区图片欧美| 丝袜人妻中文字幕| 熟女人妻精品中文字幕| 国产免费福利视频在线观看| av卡一久久| 亚洲久久久国产精品| 精品一区在线观看国产| 男女啪啪激烈高潮av片| 国产1区2区3区精品| 亚洲丝袜综合中文字幕| 啦啦啦在线观看免费高清www| 九九爱精品视频在线观看| 超色免费av| 午夜免费鲁丝| 免费在线观看完整版高清| 777米奇影视久久| a级毛色黄片| 久久99一区二区三区| 成人影院久久| 免费黄频网站在线观看国产| 视频区图区小说| 欧美 日韩 精品 国产| 18在线观看网站| 少妇人妻精品综合一区二区| 亚洲人成网站在线观看播放| 极品人妻少妇av视频| 亚洲精品第二区| 边亲边吃奶的免费视频| 美国免费a级毛片| 伦理电影大哥的女人| 两个人看的免费小视频| 日本91视频免费播放| www.色视频.com| 成年av动漫网址| 美女主播在线视频| 日韩一区二区三区影片| 日本免费在线观看一区| 亚洲情色 制服丝袜| 美女中出高潮动态图| 爱豆传媒免费全集在线观看| 少妇熟女欧美另类| 天天操日日干夜夜撸| 99久久精品国产国产毛片| 精品少妇黑人巨大在线播放| 成人国产av品久久久| 成年人午夜在线观看视频| 成年美女黄网站色视频大全免费| av国产久精品久网站免费入址| 波野结衣二区三区在线| 日本爱情动作片www.在线观看| 国产毛片在线视频| 国产成人av激情在线播放| 日韩成人伦理影院| 午夜免费鲁丝| 亚洲国产日韩一区二区| 亚洲欧美日韩卡通动漫| 视频区图区小说| 丁香六月天网| 久久久久久伊人网av| 男人舔女人的私密视频| 国产一区二区在线观看av| 精品人妻一区二区三区麻豆| 99热网站在线观看| 一个人免费看片子| 亚洲av福利一区| 久久这里有精品视频免费| 有码 亚洲区| 国产午夜精品一二区理论片| 欧美国产精品一级二级三级| 国产日韩欧美在线精品| 又粗又硬又长又爽又黄的视频| 国产亚洲精品第一综合不卡 | 亚洲美女搞黄在线观看| 日韩中字成人| 天天躁夜夜躁狠狠躁躁| 69精品国产乱码久久久| 妹子高潮喷水视频| 日韩av不卡免费在线播放| 午夜激情av网站| 中文天堂在线官网| 亚洲精品国产av成人精品| 国产精品久久久久久久电影| 亚洲国产最新在线播放| 99国产综合亚洲精品| 亚洲成人手机| 精品国产一区二区三区久久久樱花| 精品久久久精品久久久| 日本免费在线观看一区| 成人无遮挡网站| 大陆偷拍与自拍| 一区二区三区乱码不卡18| 色网站视频免费| 最近手机中文字幕大全| 精品国产一区二区三区久久久樱花| 2021少妇久久久久久久久久久| 大话2 男鬼变身卡| 99久国产av精品国产电影| 久久人妻熟女aⅴ| 国产精品嫩草影院av在线观看| 男女下面插进去视频免费观看 | 一级片免费观看大全| 97精品久久久久久久久久精品| 久久久久久久久久久久大奶| av女优亚洲男人天堂| 成人毛片a级毛片在线播放| 一区二区三区乱码不卡18| 精品国产露脸久久av麻豆| 午夜福利影视在线免费观看| av.在线天堂| 亚洲高清免费不卡视频| 国产有黄有色有爽视频| 春色校园在线视频观看| 亚洲第一区二区三区不卡| 高清欧美精品videossex| 在线天堂最新版资源| 日韩欧美精品免费久久| 成人亚洲精品一区在线观看| 青青草视频在线视频观看| 亚洲,欧美,日韩| 日韩成人av中文字幕在线观看| 尾随美女入室| 人人澡人人妻人| 亚洲,欧美,日韩| 久久99热这里只频精品6学生| 亚洲,欧美,日韩| 久久热在线av| 欧美老熟妇乱子伦牲交| 黑人欧美特级aaaaaa片| 亚洲成av片中文字幕在线观看 | 久久精品久久久久久噜噜老黄| 18禁观看日本| 日韩制服骚丝袜av| 在线观看人妻少妇| 极品人妻少妇av视频| 精品亚洲成国产av| 最近最新中文字幕免费大全7| 久久人人97超碰香蕉20202| 又大又黄又爽视频免费| av线在线观看网站| 少妇的丰满在线观看| 巨乳人妻的诱惑在线观看| 色哟哟·www| 日韩电影二区| 精品午夜福利在线看| 精品国产一区二区三区四区第35| 国产亚洲午夜精品一区二区久久| 亚洲丝袜综合中文字幕| 搡老乐熟女国产| 久久久欧美国产精品| 久久热在线av| 久久精品国产亚洲av涩爱| 人人妻人人添人人爽欧美一区卜| 国产午夜精品一二区理论片| 一边亲一边摸免费视频| 亚洲精品一二三| 日本爱情动作片www.在线观看| 亚洲成人av在线免费| 大香蕉97超碰在线| 永久免费av网站大全| 最近2019中文字幕mv第一页| 国产精品偷伦视频观看了| 日韩制服丝袜自拍偷拍| 啦啦啦在线观看免费高清www| 午夜福利视频在线观看免费| 汤姆久久久久久久影院中文字幕| 男人爽女人下面视频在线观看| 99国产综合亚洲精品| 中文字幕精品免费在线观看视频 | 国精品久久久久久国模美| 久久久a久久爽久久v久久| 亚洲国产精品专区欧美| 国产69精品久久久久777片| 亚洲精品久久久久久婷婷小说| 久久人人爽人人片av| 男女啪啪激烈高潮av片| 少妇人妻精品综合一区二区| freevideosex欧美| 久久人人爽人人片av| 国产成人精品无人区| 欧美日韩国产mv在线观看视频| 插逼视频在线观看| 春色校园在线视频观看| 全区人妻精品视频| 我要看黄色一级片免费的| 在线天堂最新版资源| 大片免费播放器 马上看| 欧美变态另类bdsm刘玥| 成人国产av品久久久| 国国产精品蜜臀av免费| 久热这里只有精品99| 国产精品欧美亚洲77777| 欧美变态另类bdsm刘玥| 在线观看国产h片| 熟女av电影| 在线观看美女被高潮喷水网站| av卡一久久| 国产av码专区亚洲av| av不卡在线播放| 成人毛片60女人毛片免费| 97人妻天天添夜夜摸| a级毛片在线看网站| 韩国av在线不卡| 大香蕉久久成人网| 999精品在线视频| 又大又黄又爽视频免费| 亚洲少妇的诱惑av| 自拍欧美九色日韩亚洲蝌蚪91| 一区二区日韩欧美中文字幕 | 在现免费观看毛片| 一级片免费观看大全| 亚洲五月色婷婷综合|