Bio-inspired large-scale bee colony networking and formation control method

Abstract

The invention relates to a bio-inspired large-scale bee colony networking and formation control method, which comprises the following steps of: constructing a hierarchical clustering network by using a method based on maximum connectivity, and constructing routes in clusters and among the clusters; a hierarchical clustering network and grouping formation control are fused, and a message format is designed; the method comprises the following steps: constructing formation groups by network clustering; realizing pigeon flock formation simulation control in the groups; updating pigeon flock social levels by utilizing cluster heads; realizing starling formation simulation control among the groups; on the grouping level, a starling group mode, an obstacle avoidance mode and a following mode are realized. Compared with the prior art, the hierarchical clustering network architecture effectively reduces the routing overhead of the large-scale unmanned aerial vehicle cluster, and meanwhile, the control complexity and the control communication overhead of the large-scale unmanned aerial vehicle cluster are effectively reduced through the pigeon flock formation simulation control method and the starling formation simulation control method and the distributed architecture thereof.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) cluster control, and in particular to a biologically inspired large-scale swarm networking and formation control method. Background Art

[0002] As an active routing protocol, OLSR (Optimized link state routing) effectively improves the battlefield adaptability of UAV self-organizing networks by actively maintaining the network topology. However, when the scale of the UAV cluster increases further, the flooding overhead of TC (Topology control) messages will increase exponentially with the increase of the cluster size, thereby exhausting the available bandwidth. Existing related studies can reduce the topology flooding overhead of OLSR to varying degrees, but most of the studies only optimize the routing topology flooding process of OLSR in specific scenarios. When the scale of the UAV cluster increases to a super-large scale, the routing overhead is still huge, and the computational overhead, storage overhead and communication hop count problems have not been solved, so the optimized scalability performance is relatively limited. HOLSR (Hierarchical optimized link state routing) limits the transmission of TC messages through a hierarchical clustering method, changes the shortcomings of insufficient scalability of flat routing, can truly solve the scalability problem of the OLSR routing protocol, and shows obvious potential in the design of large-scale UAV cluster network architecture.

[0003] Distributed control methods point the way for the control of large-scale UAV swarms, while biomimetic methods provide design ideas for distributed control. Since bird flock flight has many similarities with UAV formation control, bird flight provides many biomimetic algorithms for UAV formation control, among which pigeon flock formation and starling formation have attracted the attention of researchers. The pigeon flock control algorithm achieves the generation and maintenance of tight formations by dividing the UAV swarm into social classes and adopting the pilot method. However, when the size of the formation increases, its following errors will accumulate layer by layer, and the accumulated errors will greatly reduce the topological maintenance and following effects of the formation. At the same time, the formation structure of the pigeon flock formation is relatively fixed, which is too simple for large-scale UAV swarm formations and is not suitable for complex and changeable battlefield environments. The starling-like formation control algorithm adds phase change control to the UAV, imitates the speed following mechanism of the starling, and improves the cluster's maneuverability such as obstacle avoidance and turning. The starling-like formation control algorithm only focuses on the changes in the movement of the individual itself. Therefore, the connection relationship between the individuals and the topological structure of the cluster will change with the change of the relative position between the individuals, and a relatively stable topological connection relationship cannot be formed, which leads to the generation of invalid routes and increases the cost of route maintenance. At the same time, the fully distributed control mode of the starling formation is difficult to adapt to complex task planning.

[0004] Therefore, the existing methods have the problems of insufficient adaptability of the HOLSR protocol and formation control method, and insufficient scalability of the pigeon formation control method and the starling formation control method. Summary of the invention

[0005] The purpose of the present invention is to provide a bio-inspired large-scale bee swarm networking and formation control method, which reduces routing overhead and network maintenance overhead through a group clustering architecture; by combining the network clustering architecture with the drone group formation, the conflict between the network group architecture and the drone group formation control is avoided, and the stability of the network architecture is effectively improved; the communication overhead of the formation control is effectively reduced through a distributed control architecture; at the same time, considering the impact of the network communication architecture on the formation control, the pigeon-like formation control is realized within the group, and the starling-like formation control is realized between groups, which effectively improves the formation control stability and maneuverability of the large-scale drone cluster.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A biologically inspired large-scale bee swarm networking and formation control method, which uses a maximum connectivity-based method to build a hierarchical clustered network, provides a network maintenance method, and builds routes within and between clusters; integrates the hierarchical clustered network with group formation control, and comprehensively designs the message format; constructs formation groups with network clusters, implements pigeon-like formation control within the group, uses cluster heads to update the social hierarchy of pigeons, builds distributed interaction rules, implements starling-like formation control between groups, builds a closed circle model, uses cluster heads to control the distance between groups, uses leaders to coordinate the speed between groups, and implements starling clustering mode, obstacle avoidance mode, and following mode at the group level.

[0008] The method comprises the following steps:

[0009] Step 1) jointly design the message frame structure based on the information required for network architecture, routing maintenance, and formation control;

[0010] Step 2) Through the interaction of neighbor messages, the connectivity information, location information and cluster information of the neighbors within the two-hop range of the drone are obtained, and the drone with the largest connectivity is elected as the cluster head. The cluster head floods the cluster management message to form a cluster. After multiple rounds of elections, the clustering of all drones is completed;

[0011] Step 3) Construct a network maintenance mechanism for different network architecture failure situations, and adjust the network architecture to restore it to a stable hierarchical and clustered state;

[0012] Step 4) Active routing is constructed within and between clusters to limit the routing topology control message within the cluster. The cluster head drone floods the routing topology control message in the second layer of the network and broadcasts the member information of the cluster.

[0013] Step 5) Using network clustering as the basis for formation grouping, a multi-formation grouping control architecture is constructed to realize the formation control of UAVs, where network clustering and formation grouping are consistent:

[0014] Step 51) constructing a pigeon-like flock formation control mechanism within the group, using cluster heads to periodically manage social levels, establishing a pigeon-like flock neighbor discovery model and a neighbor interaction model, and achieving close following flight;

[0015] Step 52) Use the closed circle model to make each drone group equivalent, and define the equivalent position, equivalent radius and equivalent speed; based on the closed circle model, build a starling-like formation control mechanism between groups, obtain the position information of neighbor groups through information interaction between cluster heads, calculate and flood messages through cluster heads to control the relative position of the entire group and other groups, and the navigator adjusts its own speed by listening to the speed information of neighbor groups; map the starling's clustering mode, obstacle avoidance mode and following mode to the group formation control to achieve collision avoidance and obstacle avoidance between drone groups.

[0016] The neighbor message includes the neighbor's location, speed, and cluster head number information. Each drone exchanges its own location, speed, cluster head number, and connectivity information with its neighbor drones, and exchanges the location, cluster head number, and connectivity information of one-hop neighbors.

[0017] The step 3) is specifically as follows: designing network maintenance mechanisms for failure of member drones, failure of cluster head drones, joining of new drones, partial cluster overlap, and cluster merging, and determining the behavior of cluster heads and members under different failure scenarios, wherein:

[0018] In the scenario of member drone failure, the cluster head queries the two-hop neighbor table and records the failed member drone, and the member takes no action;

[0019] For the cluster head drone failure scenario, the cluster head does not take any action, and the members wait for a preset time before switching to the unclustered state;

[0020] For the scenario where a new drone joins, the cluster head queries the two-hop neighbor table and records the newly joined member drone. The member replaces the cluster head ID of the drone with the ID of the new cluster head.

[0021] For the scenario where some clusters overlap, the cluster head does not take any action, and the members join the new cluster with fewer hops. If there is no new cluster with fewer hops, the members remain unchanged.

[0022] In the cluster merging scenario, the cluster head with smaller connectivity or the same connectivity but smaller ID gives up the cluster head status, and the members do not take any action.

[0023] In step 51), the pigeon-like neighbor discovery model is specifically as follows: the drone is based on the exclusion radius R Rep , Alignment radius R Al And the attraction radius R Att The positions of neighbors are distinguished to determine the neighbors within a preset radius with the drone as the center. UAVs in the blind spot of the drone’s field of view are not selected as neighbors, and the neighbor discovery model has a maximum number limit.

[0024] In the step 51), the pigeon-like neighbor interaction model is specifically:

[0025] If node i and node j are in the same cluster, and node j is in the perception set N of node i i In, N i There is a maximum limit on the number of nodes in the , and there is attraction or repulsion between the nodes, as shown below:

[0026]

[0027] Among them, the node is the drone, r ij is the distance between node i and node j, is the unit vector from node i to node j, R Rep is the repulsion radius, R Al is the alignment radius, R Att is the attraction radius, is the force between nodes. The superscript indicates that the force is used to control the relative position between UAVs. i is the social level of node i, A is a constant, A>0; when 0≤r ij <R Rep When R Al ≤r ij <R Att When , there is attraction between nodes;

[0028] If N i Node j and node i do not belong to the same cluster. In order to avoid collision between nodes, neighbor nodes in different clusters do not attract each other but only repel each other. The force function is as follows:

[0029]

[0030] In order to ensure the tightness of the formation, the nearest superior pigeon generates additional following force to make up for the problem of insufficient attraction when the number of superior individuals is small. The following force is shown in the following formula:

[0031]

[0032] in, represents the position vector of node i, is the position vector of neighbor node j, h i is the social level of node i, node j is the set N i The node closest to node i;

[0033] The drone sets its speed to the average of individuals higher up in the social hierarchy, as shown below:

[0034]

[0035] in, is the velocity vector of UAV i, H i is the set of neighboring drones with a higher social rank than the drone within the sensing range of drone i, |H i | is the number of drones in the set.

[0036] In step 52), the closed circle model treats the drone group as a whole, so that the drone group is controlled as a single starling in formation, wherein the center of the closed circle model is the average position of the drones in the group, the radius of the closed circle is the distance from the drone farthest from the center to the center, and the speed of the closed circle is the speed of the leader.

[0037] In the step 52), a starling-like neighbor discovery model is established between the groups. The center, radius, and speed information of the equivalent closed circle of the neighbor groups exchanged between cluster heads are used. The starling-like neighbor discovery model is specifically as follows: each drone group has 5 fan-shaped sensing areas, each sensing area corresponds to a sensing angle of 45°, and the remaining area is a blind area. There is only one neighbor group in each sensing area, and the radial difference between drone group i and drone group j is D ij , according to the size of the radial difference, different areas are divided. If the radial difference is less than the repulsion distance D Rep , then it is in the exclusion area. If the radial difference is less than the alignment distance D Al and greater than the repulsion distance D Rep , then it is in the alignment area. If the radial difference is less than the attraction distance D Att And greater than the alignment distance D Al , it is located in the attraction area.

[0038] In the step 52), the action follow sequence number information is added to the routing message and the cluster management message. When each drone group informs its neighbors of its own speed, the action follow sequence number is added to identify its own changes. The action follow sequence number increases with the increase of interactive messages. The navigator saves the action follow sequence number of the received drone group and determines whether it has followed the action of the drone group based on the size of the action follow sequence number.

[0039] In step 52), the starlings' clustering mode, obstacle avoidance mode and following mode are mapped to group formation control, specifically:

[0040] For the cluster mode, the action mode in the repulsive area and the attractive area is as follows:

[0041]

[0042] The group number is i, and the neighbor group number is j. is the unit vector from group i to group j, Dij represents the radial difference between group i and group j, N i is the neighbor group set of group i, when D ij <D Rep When D Al ≤D ij <D Att When the groups have an attractive force, the radial difference will eventually converge to D Rep With D Al between;

[0043] For neighbors in the alignment area, the leader adjusts its speed according to the speed of the neighbor grouping, as shown in the following formula:

[0044]

[0045] in, is the speed of group i, is the speed of group j, N Al Indicates the neighbors of group i in the alignment area, v0 is the cruising speed, that is, the speed of group i is the average speed of the neighbors in the alignment area;

[0046] For obstacle avoidance mode, the navigator changes the flight direction on its own to avoid obstacles;

[0047] In the following mode, the leader follows the neighbor with the largest speed change to avoid obstacles in advance. Specifically, first determine whether to follow the group based on the action following sequence number of the neighbor group. If following, make the following judgment:

[0048]

[0049] Among them, A j (n) represents the speed change index of neighbor group j. If the speed of the neighbor group is a speed that has not been imitated before, that is, the current speed of the neighbor group is not equal to the speed of group i at some time m ago The speed change of the neighbor is measured by comparing it with the speed change at the previous moment. Otherwise, it is considered that the neighbor cannot be followed and A j (n) is set to 0;

[0050] In the following mode, only one neighbor group is followed by this group, so the node with the largest speed change in the neighbor group is selected. Assuming that the index of this group is I(n), its expression is as follows:

[0051]

[0052] Establish a speed switching threshold to evaluate the size of the packet speed change, the speed switching threshold Ti It is expressed as:

[0053]

[0054] Among them, δ is the coefficient, N i is the set of neighbors, is the number of neighbors that can be followed in the neighbor set, γ i Reflects the degree of change in the average speed of the neighbors that can be followed at present;

[0055] Let I(n) correspond to group number k. If I(n)>T i , the leader enters the following mode, and group i imitates the speed of group k; if I(n)≤T i , the leader enters the cluster mode, group i does not imitate the speed of group k, and maintains the speed consistency between the group and the neighboring group, as shown in the following formula:

[0056]

[0057] Among them, v0 is the cruising speed of the drone.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] (1) The present invention unifies the clustering of the network and the grouping of the formation, designs a fusion message structure, avoids the conflict between the network structure and the formation structure, and reduces the additional communication overhead.

[0060] (2) The present invention designs a formation control method based on the existing network architecture and routing information, and implements part of the control function in the leader and the cluster head respectively, thereby avoiding the occurrence of deadlock and improving the control effect.

[0061] (3) The present invention realizes the pilot flight of a flock of pigeons within the group, designs a dynamic maintenance mechanism for the social hierarchy of the pigeon group, and realizes pilot intensive flight in a high-maneuverability environment.

[0062] (4) The present invention integrates the starling formation control method with large-scale formation control, effectively improving the obstacle avoidance and collision avoidance of large-scale UAV clusters.

[0063] (5) The present invention is designed for large-scale UAV clusters, taking into account the mutual influence of network communication and formation control, and has strong engineering value. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is a flow chart of the method of the present invention;

[0065] Figure 2 It is a schematic diagram of the key points of the design of the present invention;

[0066] Figure 3 A diagram of a hierarchical clustering network architecture in an embodiment of the present invention;

[0067] Figure 4 A schematic diagram of the social hierarchy structure of a pigeon flock in one embodiment of the present invention;

[0068] Figure 5 A schematic diagram of a pigeon-like flock neighbor discovery model in an embodiment of the present invention;

[0069] Figure 6 A schematic diagram of an equivalent circle model in an embodiment of the present invention;

[0070] Figure 7 A schematic diagram of a starling-like neighbor discovery model in an embodiment of the present invention;

[0071] Figure 8 A comparison diagram of routing costs in an embodiment of the present invention;

[0072] Fig. 9 A diagram comparing communication overhead in an embodiment of the present invention;

[0073] Fig.10 This is a diagram showing the obstacle avoidance effect of a BINC without a following mechanism in an embodiment of the present invention;

[0074] Fig.11 This is a diagram showing the obstacle avoidance effect of a BINC with a following mechanism in one embodiment of the present invention;

[0075] Fig.12 A comparison diagram of turning radius in an embodiment of the present invention;

[0076] Fig.13 A comparison chart of the number of cluster switching of different models in an embodiment of the present invention;

[0077] Fig.14 A comparison diagram of the number of cluster switching in a straight-line navigation scenario in an embodiment of the present invention;

[0078] Fig.15 A comparison diagram of the number of cluster switching in a turning obstacle avoidance scenario in an embodiment of the present invention; DETAILED DESCRIPTION

[0079] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0080] Example 1

[0081] The present embodiment provides a bio-inspired large-scale bee swarm networking and formation control method, which has low routing overhead, formation control overhead and network maintenance overhead. It uses a maximum connectivity-based method to construct a hierarchical cluster network, and provides a network maintenance method to construct routes within and between clusters; integrates the hierarchical cluster network with group formation control, and comprehensively designs the message format; constructs formation groups with network clusters, implements pigeon-like formation control within the group, uses cluster heads to update the social hierarchy of pigeons, constructs distributed interaction rules, implements starling-like formation control between groups, constructs a closed circle model, uses cluster heads to control the distance between groups, uses leaders to coordinate the speed between groups, and implements starling clustering mode, obstacle avoidance mode, and following mode at the group level.

[0082] Specifically, Figure 1 and Figure 2 As shown, the method comprises the following steps:

[0083] Step 1) Design the message frame structure together based on the information required for network architecture, routing maintenance, and formation control.

[0084] Step 2) Through the interaction of neighbor messages, the connectivity information, location information and cluster information of the neighbors within the two-hop range of the drone are obtained, and the drone with the largest connectivity is elected as the cluster head. The cluster head floods the cluster management message to form a cluster. After multiple rounds of elections, the clustering of all drones is completed.

[0085] The neighbor message includes the neighbor's location, speed, and cluster head number information. Through the interaction of HELLO messages, each drone exchanges the location, speed, cluster head number, and connectivity information of its own drone with the neighbor drone, and exchanges the location, cluster head number, and connectivity information of the one-hop neighbor, so that each drone can obtain the connectivity, cluster head number, and location information of the neighbor within two hops. In this embodiment, for a single drone node in the cluster, its connectivity refers to the number of neighbor nodes directly connected to the node. That is, a drone node can directly communicate with how many other drone nodes through a wireless communication link, and the number of these neighbor nodes that can directly communicate is the connectivity of the node.

[0086] According to the connectivity information of neighbors within two hops, the clustering structure is constructed using the maximum connectivity criterion. The clustering network structure is as follows: Figure 3 shown.

[0087] In this embodiment, the drone has two communication channels, one for long-distance inter-cluster communication and the other for short-distance intra-cluster communication. The network management message of the cluster head incorporates the position change information, social level, and cluster head sequence number information required for formation motion control.

[0088] Step 3) Build a network maintenance mechanism for different network architecture failure situations and adjust the network architecture to restore it to a stable hierarchical and clustered state.

[0089] The network maintenance mechanisms are designed for the failure of member drones, failure of cluster head drones, joining of new drones, overlap of some clusters, and merging of clusters, and the behaviors of cluster heads and members under different failure scenarios are determined, as shown in Table 1.

[0090] Table 1

[0091]

[0092] Data packets of the same cluster arrive via direct forwarding, and data packets of different clusters are forwarded via the cluster head by default.

[0093] Step 4) Active routing is constructed within and between clusters to limit the routing topology control messages within the cluster. The cluster head drone floods the routing topology control messages in the second layer of the network and broadcasts the member information of the cluster.

[0094] exist Figure 3 In the network architecture, the routing topology messages are restricted within the clusters, and the addresses of the member drones are flooded between cluster heads. When member drone 1 sends a data packet to member drone 4, the addressing process of the data packet is 1→A1→3→4. When drone 1 sends a data packet to drone 14, the addressing process of the data packet is 1→A1→A2→B2→C2→D2→D1→14.

[0095] Step 5) Using network clustering as the basis for formation grouping, a multi-formation grouping control architecture is constructed to realize the formation control of UAVs, in which network clustering and formation grouping are consistent.

[0096] The formation control grouping is based on network clustering. The member nodes generate attraction according to the position of the cluster head and aggregate to form groups.

[0097] Step 51) construct a pigeon-like flock formation control mechanism within the group, use the cluster head to periodically manage the social hierarchy, establish a pigeon-like flock neighbor discovery model and a neighbor interaction model, and achieve close following flight.

[0098] A pigeon-like group formation control mechanism is constructed within the group, including the periodic allocation of social levels, the neighbor discovery model, and the pigeon-like group interaction rules. The specific design is as follows:

[0099] (1) The cluster head periodically calculates the social rank based on the positions of the member nodes and makes allocations.

[0100] Specifically, the cluster head determines the social level according to the advancing direction of the drone group. The higher the social level is, the lower the level number is. The number of drones at each level is shown in the following formula:

[0101]

[0102] Where n is the social class number, K(n) is the number of drones corresponding to the social class number, k is a constant, and when n=0, the drone is selected as the leader. Figure 4 shown.

[0103] (2) Imitation pigeon flock neighbor discovery model Figure 5 As shown in the figure, each drone has a limited number of neighbors and there is a blind spot. The neighboring drones are identified to which cluster they belong. The relative distance to other drones is r, and the exclusion radius R Rep , Alignment radius R Al , attraction radius R Att Produce different interactive behaviors to surrounding drones.

[0104] The drone is based on the repulsion radius R Rep , Alignment radius R Al And the attraction radius R Att The positions of neighbors are distinguished to determine neighbors within a preset radius with the drone as the center, wherein drones located in the blind spot of the drone are not selected as neighbors, and the neighbor discovery model has a maximum number limit. In this embodiment, the maximum number of neighbors of the drone is 7.

[0105] (3) The neighbor interaction rules of pigeon flocks include repulsive behavior, speed alignment behavior, attractive behavior, and close following behavior, as shown in the following table:

[0106] Table 2

[0107] The relative distance between the drone and its neighbors Drone Behavior <![CDATA[0≤r<R Rep ]]> Drones are repelled <![CDATA[R Rep ≤r<R Al ]]> Drone alignment speed <![CDATA[R Al ≤r<R Att ]]> Drones are attracted none Drones are attracted by the recent high-level drones

[0108] If node i and node j are in the same cluster, and node j is in the perception set N of node i i In, N i The number of nodes in is no more than 7, and there is attraction or repulsion between the nodes, as shown below:

[0109]

[0110] Among them, the node is the drone, r ij is the distance between node i and node j, is the unit vector from node i to node j, R Rep is the repulsion radius, R Al is the alignment radius, R Attis the attraction radius, is the force between nodes. The superscript indicates that the force is used to control the relative position between UAVs. i is the social level of node i, A is a constant, A>0; when 0≤r ij <R Rep When R Al ≤r ij <R Att When , there is attraction between nodes;

[0111] If N i Node j and node i do not belong to the same cluster. In order to avoid collision between nodes, neighbor nodes in different clusters do not attract each other but only repel each other. The force function is as follows:

[0112]

[0113] In order to ensure the tightness of the formation, the nearest superior pigeon generates additional following force to make up for the problem of insufficient attraction when the number of superior individuals is small. The following force is shown in the following formula:

[0114]

[0115] in, represents the position vector of node i, is the position vector of neighbor node j, h i is the social level of node i, node j is the set N i The node closest to node i.

[0116] The drone sets its own speed to the average of individuals with higher social rank, so that the entire cluster can achieve consistency in trajectory planning, as shown in the following formula:

[0117]

[0118] in, is the velocity vector of UAV i, H i is the set of neighboring drones with a higher social rank than the drone within the sensing range of drone i, |H i | is the number of drones in the set.

[0119] The drone follows the movement information of the nearest high-level drone.

[0120] Step 52) Use the closed circle model to make each drone group equivalent, and define the equivalent position, equivalent radius and equivalent speed; based on the closed circle model, build a starling-like formation control mechanism between groups, obtain the position information of neighbor groups through information interaction between cluster heads, calculate and flood messages through cluster heads to control the relative position of the entire group and other groups, and the navigator adjusts its own speed by listening to the speed information of neighbor groups; map the starling's clustering mode, obstacle avoidance mode and following mode to the group formation control to achieve collision avoidance and obstacle avoidance between drone groups.

[0121] Specifically, a starling-like formation flight mechanism is constructed within the cluster, including a closed circle equivalent model, a starling-like neighbor discovery model, and interaction rules under different modes.

[0122] (1) The center of the closed circle is the geometric center of the UAV group, the radius is the distance of the UAV farthest from the geometric center, and the speed is the speed of the leader.

[0123] like Figure 6 As shown in the figure, the closed circle model treats the UAV group as a whole, so that the UAV group is controlled as a single starling in formation, where the center of the closed circle model is the average position of the UAVs in the group, the radius R of the closed circle is the distance from the UAV farthest from the center to the center, and the speed of the closed circle is the speed of the leader.

[0124] (2) Establish a starling-like neighbor discovery model between groups, such as Figure 7 As shown in the figure, the equivalent closed circle center, radius, and speed information of the neighbor groupings exchanged between cluster heads. Among them, each drone group has 5 fan-shaped sensing areas, each sensing area corresponds to a sensing angle of 45°, and the remaining area is a blind area. There is only one neighbor group in each sensing area, and the radial difference between drone group i and drone group j is D ij , according to the size of the radial difference, different areas are divided. If the radial difference is less than the repulsion distance D Rep , then it is in the exclusion area. If the radial difference is less than the alignment distance D Al and greater than the repulsion distance D Rep , then it is in the alignment area. If the radial difference is less than the attraction distance D Att And greater than the alignment distance D Al , then it is located in the attraction area. Only one group can be selected in each perception area, such as Figure 7 In the above example, among group m and group n, the group m with a closer distance is selected as the neighbor of the area.

[0125] This step realizes the separation of formation control. The leader controls the speed and direction of trajectory planning, and the cluster head controls the relative position of neighbor groups. Inter-group formation control can maintain a relatively stable distance between groups, thereby avoiding network structure conflicts between clusters. The leader can autonomously plan the trajectory to achieve obstacle avoidance.

[0126] (3) The neighbor interaction rules of starlings are determined by clustering mode, obstacle avoidance mode, and following mode, as shown in Table 3.

[0127] Table 3

[0128]

[0129] The action follow sequence number information is added to the routing message and cluster management message. When each drone group informs its neighbors of its own speed, the action follow sequence number is added to identify its own changes. The action follow sequence number increases with the increase of interactive messages. The navigator saves the action follow sequence number of the received drone group and determines whether it has followed the action of the drone group based on the size of the action follow sequence number.

[0130] For the cluster mode, the action mode in the repulsive area and the attractive area is as follows:

[0131]

[0132] The group number is i, and the neighbor group number is j. is the unit vector from group i to group j, D ij represents the radial difference between group i and group j, N i is the neighbor group set of group i, when D ij <D Rep When D Al ≤D ij <D Att When the groups have an attractive force, the radial difference will eventually converge to D Rep With D Al between;

[0133] For neighbors in the alignment area, the leader adjusts its speed according to the speed of the neighbor grouping, as shown in the following formula:

[0134]

[0135] in, is the speed of group i, is the speed of group j, N Al It indicates the neighbors of group i in the alignment area, and v0 is the cruising speed, that is, the speed of group i is the average speed of the neighbors in the alignment area.

[0136] For obstacle avoidance mode, the navigator changes the flight direction on its own to avoid obstacles.

[0137] In the following mode, the leader follows the neighbor with the largest speed change to avoid obstacles in advance. First, determine whether to follow the group based on the action following sequence number of the neighbor group. If following, make the following judgment:

[0138]

[0139] Among them, A j (n) represents the speed change index of neighbor group j. If the speed of the neighbor group is a speed that has not been imitated before, that is, the current speed of the neighbor group is not equal to the speed of group i at some time m ago The speed change of the neighbor is measured by comparing it with the speed change at the previous moment. Otherwise, it is considered that the neighbor cannot be followed and A j (n) is set to 0. Where D ij It is the radial difference value of the neighbor group, which is used to normalize the index size.

[0140] In the following mode, only one neighbor group is followed by this group, so the node with the largest speed change in the neighbor group is selected. Assuming that the index of this group is I(n), its expression is as follows:

[0141]

[0142] A speed switching threshold is established to evaluate the size of the group speed change. When the speed consistency between multiple clusters is poor, random speed following will make it difficult for the cluster speed to converge. A higher threshold should be set to put the cluster in replication mode and produce consistent speed. When the speed consistency between multiple clusters is good, a lower threshold is set. In this embodiment, the speed switching threshold T i It is expressed as:

[0143]

[0144] Among them, δ is the coefficient, N i is the set of neighbors, is the number of neighbors that can be followed in the neighbor set, γ i Reflects the average speed change of the current neighbors that can be followed.

[0145] Let I(n) correspond to group number k. If I(n)>T i , the leader enters the following mode, and group i imitates the speed of group k; if I(n)≤T i, the leader enters the cluster mode, group i does not imitate the speed of group k, and maintains the speed consistency between the group and the neighboring group, as shown in the following formula:

[0146]

[0147] Among them, v0 is the cruising speed of the drone.

[0148] Example 2

[0149] This embodiment is based on the method of Embodiment 1, and specifically provides a bio-inspired large-scale UAV swarm network architecture and formation control integrated design method, referred to as BINC, including the following steps:

[0150] S1, at 30000×30000m 2 Within the range, 40 aircraft / km 2 The density of the simulation is 520, 640, 760, 880, and 1000 nodes. Each drone has two channels, one for long-distance inter-group communication and one for short-distance intra-group communication. The large-scale drone cluster is randomly distributed in a certain area at the initial moment.

[0151] S2, add the node's connectivity, position, cluster head number, speed, and follow-up number information to the HELLO message, so as to obtain the connectivity, position, cluster head number, speed, and follow-up number information of the one-hop neighbor; add the connectivity, position, and cluster head number to the one-hop neighbor table, so as to obtain the connectivity, position, and cluster head number information of the two-hop neighbor.

[0152] S3, large-scale drone clusters complete clustering through multiple rounds of elections by exchanging connectivity information within two hops. The TC messages of each cluster are limited within the cluster, so the routing overhead is significantly reduced. Gradually increasing the size of the drone cluster, it can be found that the routing overhead is significantly reduced. The communication overhead required for distributed formation control is only related to the local situation. As the number of drone clusters increases, the formation control overhead within the group and the formation control overhead between groups are small and stable, so its scalability is excellent. Figure 8 By comparing the routing overheads under different drone cluster sizes, we can see that this solution significantly reduces the routing overhead compared to OLSR, CLSR and other solutions. Fig. 9 It shows how the routing overhead and motion control overhead change with the size of the drone cluster.

[0153] S4, add social level field, packet movement field, cluster head sequence number field to the cluster management message, so as to realize network management and intra-packet movement control. Add packet position, packet speed, packet radius, cluster head sequence number field to the HELLO message of the second layer of the network, so as to realize the information exchange of neighbor groups within one hop.

[0154] S5, large-scale drone clusters fly in straight-line navigation scenarios and turning obstacle avoidance scenarios. The scale of the drone cluster is fixed, and the obstacle avoidance effects of drone groups with and without the starling following mechanism are compared in the turning obstacle avoidance scenario. The leader of the group with the starling following mechanism follows according to the following sequence number and the speed of the neighboring group to achieve early obstacle avoidance. When the scale of the drone cluster is fixed, when time changes, the distance between drone groups is maintained by the starling group formation control, and the density of drone groups is maintained by the pigeon-like group formation control within the group. Fig.10 , Fig.11 It shows that the cluster without starling following mode has a smaller turning radius and there is a phenomenon of hitting obstacles, while the cluster with starling following mode does not have the above problems. Fig.12 The obstacle avoidance distance with and without the starling following mechanism is shown in .

[0155] S6, a large-scale drone cluster flies in a straight-line navigation scenario and a turning obstacle avoidance scenario. The drones may switch between different groups, but the inter-group distance control of BINC will limit the distance between network clusters and reduce the occurrence of cluster switching. For different inter-group distance constraints, the size of the drone cluster is gradually increased. A smaller inter-group distance constraint will cause the occurrence of drone group switching, and a larger inter-group distance constraint can reduce the occurrence of cluster switching. Fig.13 The number of cluster switching between BINC and Boids models in different scenarios and different cluster sizes is counted in the figure. It can be seen that BINC can effectively reduce the number of cluster switching in both straight-line navigation scenarios and turning obstacle avoidance scenarios, thereby maintaining the stability of the network. Fig.14 and Fig.15 The number of cluster switching under the condition of inter-group distance constraint in the straight-line navigation scenario and the turning obstacle avoidance scenario is compared. It can be seen that the larger the inter-group distance constraint, the better the network maintenance effect.

[0156] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A biologically inspired large-scale bee swarm networking and formation control method, characterized in that: This method uses a maximum connectivity-based method to construct a hierarchical clustered network, provides a network maintenance method, and constructs routes within and between clusters; integrates the hierarchical clustered network with group formation control, and comprehensively designs the message format; constructs formation groups with network clustering, implements pigeon-like formation control within the group, uses cluster heads to update the social hierarchy of pigeons, constructs distributed interaction rules, implements starling-like formation control between groups, constructs a closed circle model, uses cluster heads to control the distance between groups, uses leaders to coordinate the speed between groups, and implements starling clustering mode, obstacle avoidance mode, and following mode at the group level.

2. A bio-inspired large-scale bee swarm networking and formation control method according to claim 1, characterized in that: The method comprises the following steps: Step 1) jointly design the message frame structure based on the information required for network architecture, routing maintenance, and formation control; Step 2) Through the interaction of neighbor messages, the connectivity information, location information and cluster information of the neighbors within the two-hop range of the drone are obtained, and the drone with the largest connectivity is elected as the cluster head. The cluster head floods the cluster management message to form a cluster. After multiple rounds of elections, the clustering of all drones is completed; Step 3) Construct a network maintenance mechanism for different network architecture failure situations, and adjust the network architecture to restore it to a stable hierarchical and clustered state; Step 4) Active routing is constructed within and between clusters to limit the routing topology control message within the cluster. The cluster head drone floods the routing topology control message in the second layer of the network and broadcasts the member information of the cluster. Step 5) Using network clustering as the basis for formation grouping, a multi-formation grouping control architecture is constructed to realize the formation control of UAVs, where network clustering and formation grouping are consistent: Step 51) constructing a pigeon-like flock formation control mechanism within the group, using cluster heads to periodically manage social levels, establishing a pigeon-like flock neighbor discovery model and a neighbor interaction model, and achieving close following flight; Step 52) Use the closed circle model to make each drone group equivalent, and define the equivalent position, equivalent radius and equivalent speed; based on the closed circle model, build a starling-like formation control mechanism between groups, obtain the position information of neighbor groups through information interaction between cluster heads, calculate and flood messages through cluster heads to control the relative position of the entire group and other groups, and the navigator adjusts its own speed by listening to the speed information of neighbor groups; map the starling's clustering mode, obstacle avoidance mode and following mode to the group formation control to achieve collision avoidance and obstacle avoidance between drone groups.

3. A bio-inspired large-scale bee swarm networking and formation control method according to claim 2, characterized in that: The neighbor message includes the neighbor's location, speed, and cluster head number information. Each drone exchanges its own location, speed, cluster head number, and connectivity information with its neighbor drones, and exchanges the location, cluster head number, and connectivity information of one-hop neighbors.

4. A bio-inspired large-scale bee swarm networking and formation control method according to claim 2, characterized in that: The step 3) is specifically as follows: designing network maintenance mechanisms for failure of member drones, failure of cluster head drones, joining of new drones, partial cluster overlap, and cluster merging, and determining the behavior of cluster heads and members under different failure scenarios, wherein: In the scenario of member drone failure, the cluster head queries the two-hop neighbor table and records the failed member drone, and the member takes no action; For the cluster head drone failure scenario, the cluster head does not take any action, and the members wait for a preset time before switching to the unclustered state; For the scenario where a new drone joins, the cluster head queries the two-hop neighbor table and records the newly joined member drone. The member replaces the cluster head ID of the drone with the ID of the new cluster head. For the scenario where some clusters overlap, the cluster head does not take any action, and the members join the new cluster with fewer hops. If there is no new cluster with fewer hops, the members remain unchanged. In the cluster merging scenario, the cluster head with smaller connectivity or the same connectivity but smaller ID gives up the cluster head status, and the members do not take any action.

5. The method of large-scale bio-inspired swarm networking and formation control according to claim 2, characterized in that: In step 51), the pigeon-like neighbor discovery model is specifically as follows: the drone is based on the exclusion radius R Rep , Alignment radius R Al And the attraction radius R Att The positions of neighbors are distinguished to determine the neighbors within a preset radius with the drone as the center. UAVs in the blind spot of the drone’s field of view are not selected as neighbors, and the neighbor discovery model has a maximum number limit.

6. A bio-inspired large-scale bee swarm networking and formation control method according to claim 2, characterized in that: In the step 51), the pigeon-like neighbor interaction model is specifically: If node i and node j are in the same cluster, and node j is in the perception set N of node i i In, N i There is a maximum limit on the number of nodes in the , and there is attraction or repulsion between the nodes, as shown below: Among them, the node is the drone, r ij is the distance between node i and node j, is the unit vector from node i to node j, R Rep is the repulsion radius, R Al is the alignment radius, R Att is the attraction radius, is the force between nodes. The superscript indicates that the force is used to control the relative position between UAVs. i is the social level of node i, A is a constant, A>0; when 0≤r ij <R Rep When R Al ≤r ij <R Att When , there is attraction between nodes; If N i Node j and node i do not belong to the same cluster. In order to avoid collision between nodes, neighbor nodes in different clusters do not attract each other but only repel each other. The force function is as follows: In order to ensure the tightness of the formation, the nearest superior pigeon generates additional following force to make up for the problem of insufficient attraction when the number of superior individuals is small. The following force is shown in the following formula: in, represents the position vector of node i, is the position vector of neighbor node j, h i is the social level of node i, node j is the set N i The node closest to node i; The drone sets its speed to the average of individuals higher up in the social hierarchy, as shown below: in, is the velocity vector of UAV i, H i is the set of neighboring drones with a higher social rank than the drone within the sensing range of drone i, |H i | is the number of drones in the set.

7. The method of large-scale bio-inspired swarm networking and formation control according to claim 2, characterized in that: In step 52), the closed circle model treats the drone group as a whole, so that the drone group is controlled as a single starling in formation, wherein the center of the closed circle model is the average position of the drones in the group, the radius of the closed circle is the distance from the drone farthest from the center to the center, and the speed of the closed circle is the speed of the leader.

8. The method of large-scale bio-inspired swarm networking and formation control according to claim 2, characterized in that: In the step 52), a starling-like neighbor discovery model is established between the groups. The center, radius, and speed information of the equivalent closed circle of the neighbor groups exchanged between cluster heads are used. The starling-like neighbor discovery model is specifically as follows: each drone group has 5 fan-shaped sensing areas, each sensing area corresponds to a sensing angle of 45°, and the remaining area is a blind area. There is only one neighbor group in each sensing area, and the radial difference between drone group i and drone group j is D ij , according to the size of the radial difference, different areas are divided. If the radial difference is less than the repulsion distance D Rep , then it is in the exclusion area. If the radial difference is less than the alignment distance D Al and greater than the repulsion distance D Rep , then it is in the alignment area. If the radial difference is less than the attraction distance D Att And greater than the alignment distance D Al , it is located in the attraction area.

9. The method of large-scale bio-inspired swarm networking and formation control according to claim 2, characterized in that: In the step 52), the action follow sequence number information is added to the routing message and the cluster management message. When each drone group informs its neighbors of its own speed, the action follow sequence number is added to identify its own changes. The action follow sequence number increases with the increase of interactive messages. The navigator saves the action follow sequence number of the received drone group and determines whether it has followed the action of the drone group based on the size of the action follow sequence number.

10. The method for bio-inspired large-scale bee swarm networking and formation control according to claim 8, characterized in that: In step 52), the starlings' clustering mode, obstacle avoidance mode and following mode are mapped to group formation control, specifically: For the cluster mode, the action mode in the repulsive area and the attractive area is as follows: The group number is i, and the neighbor group number is j. is the unit vector from group i to group j, D ij represents the radial difference between group i and group j, N i is the neighbor group set of group i, when D ij <D Rep When D Al ≤D ij <D Att When the groups have an attractive force, the radial difference will eventually converge to D Rep With D Al between; For neighbors in the alignment area, the leader adjusts its speed according to the speed of the neighbor grouping, as shown in the following formula: in, is the speed of group i, is the speed of group j, N Al Indicates the neighbors of group i in the alignment area, v0 is the cruising speed, that is, the speed of group i is the average speed of the neighbors in the alignment area; For obstacle avoidance mode, the navigator changes the flight direction on its own to avoid obstacles; In the following mode, the leader follows the neighbor with the largest speed change to avoid obstacles in advance. Specifically, first determine whether to follow the group based on the action following sequence number of the neighbor group. If following, make the following judgment: Among them, A j (n) represents the speed change index of neighbor group j. If the speed of the neighbor group is a speed that has not been imitated before, that is, the current speed of the neighbor group is not equal to the speed of group i at some time m ago The speed change of the neighbor is measured by comparing it with the speed change at the previous moment. Otherwise, it is considered that the neighbor cannot be followed and A j (n) is set to 0; In the following mode, only one neighbor group is followed by this group, so the node with the largest speed change in the neighbor group is selected. Assuming that the index of this group is I(n), its expression is as follows: Establish a speed switching threshold to evaluate the size of the packet speed change, the speed switching threshold T i It is expressed as: Among them, δ is the coefficient, N i is the set of neighbors, is the number of neighbors that can be followed in the neighbor set, γ i Reflects the change in the average speed of the neighbors that can be followed at present; Let I(n) correspond to group number k. If I(n)>T i , the leader enters the following mode, and group i imitates the speed of group k; if I(n)≤T i , the leader enters the cluster mode, group i does not imitate the speed of group k, and maintains the speed consistency between the group and the neighboring group, as shown in the following formula: Among them, v0 is the cruising speed of the drone.

Images