Loss detection and automatic replacement reconstruction method based on cooperation of multiple unmanned aerial vehicles

By improving the bird flock model and the automatic replacement and reconstruction method for redundant node identification, the problems of reconstruction delay and long travel distance in UAV swarms are solved, achieving efficient topology reconstruction and fast node replacement, which is suitable for highly mobile UAV ad hoc networks.

CN119012414APending Publication Date: 2024-11-22CHANGCHUN UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410986715.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Existing drone swarm reconstruction methods suffer from high reconstruction latency, long repair node movement distances, and the need for centralized control. Furthermore, existing algorithms are not applicable to highly mobile drone ad hoc networks.

Method used

A loss detection and automatic replacement reconstruction method based on multi-UAV collaboration is adopted. An improved bird flock model is used to construct a 1-connected network, perform local cut point detection and redundant node marking, and combine inertia, attraction, repulsion, alignment and approach-avoidance rules to optimize topology reconstruction through redundant node judgment and automatic replacement protocol.

Benefits of technology

It reduces communication overhead, improves the efficiency of rapid detection and reconstruction after drone node failure, reduces the impact of network structure, and solves the problem of high topology reconstruction complexity in highly mobile drone ad hoc networks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119012414A_ABST
    Figure CN119012414A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of unmanned aerial vehicle cluster network topology reconstruction, particularly relates to a loss detection and automatic replacement reconstruction method based on multi-unmanned aerial vehicle cooperation, introduces a redundancy judgment strategy of unmanned aerial vehicle nodes on the basis of traditional distributed reconstruction, and provides a loss detection and automatic replacement mechanism based on multi-unmanned aerial vehicle cooperation. According to the method, the difference between an unmanned aerial vehicle topology model and a traditional wireless sensor network model is comprehensively considered, and an avoidance rule and an inertia rule conforming to the flight law of the unmanned aerial vehicle are introduced; redundancy which considers the boundary effect is adopted to judge that an unmanned aerial vehicle task scene is well met, the influence of node movement on the whole network is reduced, and the overall movement distance of unmanned aerial vehicle nodes in a cluster is also reduced. According to the loss detection and automatic replacement method based on multi-unmanned aerial vehicle writing, on the basis of information exchange between unmanned aerial vehicle nodes, a competition scene caused by communication delay is considered, and the rapid detection and reconstruction efficiency after the unmanned aerial vehicle nodes fail is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle cluster network topology reconstruction, and specifically to a loss detection and automatic replacement reconstruction method based on multi-unmanned aerial vehicle collaboration. Background Art

[0002] Drone swarms are a combat system consisting of a large number of small drones, based on the combat capability of single-platform drones, supported by the collaborative interaction capabilities between drones, and centered on the emergence of group intelligence. They are built based on an open system architecture and are integrated and constructed with advantages and intelligent features such as anti-destruction, low cost, and distributed functions. Drone swarm combat will be widely used in future wars, becoming the core of future wars and having an important impact on future wars.

[0003] Topology reconstruction refers to changing the network topology by controlling the mobile nodes or links in the network, so that the network has higher performance and can provide better services to users. Topology reconstruction is an effective means to restore network connectivity. That is, when a network failure causes the network to be partitioned, connectivity is destroyed, and performance is significantly reduced, the topology reconstruction algorithm is run to restore network connectivity by adding new network nodes or redundant nodes in the mobile network.

[0004] Existing research on connectivity optimization and topology reconstruction technology is mostly concentrated in the field of wireless sensor networks, and there is little research on topology reconstruction technology specifically for drone cluster communication networks. Research on reconstruction algorithms in large-scale drone cluster topology networks still has problems such as high algorithm complexity and incompatibility with application scenarios. At the same time, in traditional ADHOC networks, the mobility of nodes is relatively weak, and it is unreasonable to directly apply these algorithms to highly mobile drone self-organizing networks. Therefore, the topology fault tolerance and reconstruction problem in drone self-organizing networks with strong mobility, higher real-time requirements, and complex working environments is an area that needs to be studied and improved urgently.

[0005] Therefore, we propose a loss detection and automatic replacement and reconstruction method based on multi-UAV collaboration to solve the above problems. Summary of the invention

[0006] 1. Technical issues to be resolved

[0007] In view of the shortcomings of the prior art, the present invention provides a loss detection and automatic replacement reconstruction method based on multi-UAV collaboration, which solves the problems of high reconstruction delay, long moving distance of repair nodes and the need for centralized control in the existing UAV cluster reconstruction method.

[0008] (II) Technical solution

[0009] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0010] A loss detection and automatic replacement reconstruction method based on multi-UAV collaboration includes the following steps:

[0011] Step 1, construct an improved bird flock model: the entire drone cluster is constructed as a 1-connected network using the improved bird flock model; the improved bird flock model includes traditional gravity rules, alignment rules, repulsion rules, and inertial rules and approach-avoidance rules that conform to the drone motion rules; the multi-drone system uses a wireless communication mode, namely the user datagram protocol;

[0012] Step 2, local cut point detection: perform k-Tarjan cut point detection on the 3-hop network of each node in the singly connected network constructed in step 1, detect the cut points in the network and mark them as key nodes when exchanging neighbor information. These nodes cannot be used as mobile nodes in the automatic replacement step;

[0013] Step 3, redundant node marking: In the drone cluster, redundancy refers to the relevance and repetitiveness of information provided by unmanned nodes. The node redundancy is judged by combining the expected redundancy and the expected coverage value considering the boundary effect, that is, preparing mobile nodes for automatic replacement in step 5;

[0014] Step 4, UAV packet loss detection: This detection protocol consists of two tasks; the purpose of task 1 is to find the suspected lost UAV node, and task 2 is to record the result of task 1; another local variable managed by both tasks is a timer to determine whether the UAV node is lost;

[0015] Step 5, automatic replacement of drones: After receiving a group of lost drones, the drones will decide whether automatic replacement is needed based on the proposed automatic replacement protocol, and propose corresponding competition schemes to determine the optimal replacement drone node.

[0016] Furthermore, the communication sensing range of a single node in the drone cluster in step 1 can be divided into a gravitational sensing area (R s ≤|V i V j |≤R a ), stable perception area (R r ≤|V i V j |≤R s ) and the repulsive force sensing area (|V k V i |≤R r ), where R a is the maximum communication distance of a single UAV node;

[0017] In addition, according to the characteristics of the movement of drone clusters, inertial rules and approach-avoidance rules are added to make it more adaptable to the movement characteristics of nodes in drone clusters;

[0018] (1) Inertia rule

[0019] In real-world scenarios, the change in the speed of individual drones is constrained by physical inertia and will not change dramatically in a short period of time. The improved flock model uses the unit direction vector of the individual speed of the drone at the current moment directly as the inertial component of the speed at the next moment, which is the following formula:

[0020]

[0021] (2) Gravitational rules

[0022] The gravitational rule simulates the cohesive characteristics of individual drone clusters; in the improved bird flock model, individual drones try to get close to the geometric center of all neighboring drones within their perception range, which is the following formula:

[0023]

[0024] Among them, V i =(x i ,y i , z i ) is the coordinate of the target node i at time n; assuming that there are N neighboring drones in the gravity sensing area of ​​drone i, and V j =(x i ,y i , z i ) represents the current coordinates of neighboring drone individual j, then the geometric center coordinates V of all neighboring nodes o =(x i ,y i , z i ) can be expressed as follows:

[0025]

[0026] (3) Repulsion rule

[0027] The repulsion rule simulates the collision avoidance behavior between individual swarm drones in real scenarios. When an individual drone senses that it is too close to its neighboring drone, that is, the distance is smaller than the repulsion radius, a repulsion effect is generated for adjustment.

[0028]

[0029] Among them, V i =(x i ,y i , z i ) is the coordinate of the target individual at time n; V k =(xk ,y k , z k ) represents the coordinates of neighboring drone k in the individual repulsion perception domain at time n;

[0030] (4) Alignment rules

[0031] The purpose of the alignment rule is to make all drones in the cluster form a unified movement trend; in the improved bird flock model, the distance between each drone and the surrounding is less than the perception radius R a The movement direction of the UAV tends to be consistent, that is, the following formula

[0032]

[0033] Where m is the number of neighboring drone nodes;

[0034] (5) Approach-avoidance rules

[0035] The approach-avoidance rule is used to simulate the behavior of individual drone swarms avoiding threat sources, as well as discovering and approaching tasks, such as reconnaissance and attacking targets;

[0036]

[0037] Among them, V r is the coordinate of the target object or threat object; when the target is a mission target, the sign after the equal sign in the above formula is positive, and when the target is a threat object, the sign is negative.

[0038] Furthermore, the redundant node marking in step 3 combines the node redundancy expectation with the expected coverage quality value considering the boundary effect to perform redundant node judgment;

[0039] The redundancy of a single node in the network is expressed as follows:

[0040]

[0041] In the formula, based on the probability perception model of a single node, SD i To mark a single node v in the drone cluster network i The perception range, Represents node v i The set of all one-hop neighbor nodes; node v i The redundancy is defined as the ratio of the overlap between the node's own communication sensing area and the neighboring node's sensing area to the node's own sensing area;

[0042] All drone nodes are deployed in area S. The probability of any node being randomly deployed in area S is (S represents the area of ​​the drone scattering area); in order to be close to the actual situation, the sensing radius of the drone node is set to obey N(r0,δ2 ) distribution; for the sensing radius, they are R i and R j Two drone nodes v i and v j If there are overlapping sensing areas in the topological structure, p ij To express the probability that two nodes have overlapping perception areas, it can be expressed as follows:

[0043]

[0044] Let E(p ij ) is p ij The mathematical expectation of E(p ij ) is calculated using the following formula:

[0045]

[0046] use To specifically represent the node v in the cluster i and v j The probability that there is no overlapping area is expressed as 1-p ij ; If node v i There are k neighbors, use p ik To express the probability that this node has an overlapping area with any neighboring node, the probability that the sensing area does not overlap with any neighboring node can be obtained. but It can be expressed as follows:

[0047]

[0048] From the above formula, we can It is expressed as the following formula:

[0049]

[0050] If the expected value of the node's redundancy is greater than EDQ, that is, node v i The expected value of the redundancy ξ(i) satisfies the following formula, which means that node v i ; The overlap ratio with its neighbor nodes is greater than the EDQ value, that is, let any node v in the network i ; The network connectivity will also reach the EDQ value when working in the minimum power state;

[0051] E(ξ(i))≥E EDQ # E in the formula EDQ The EDQ value is shown as follows:

[0052]

[0053] Combining the above formulas, if node v i , the number of neighbor nodes k satisfies the following formula, that is, node v i : Meet the basic conditions for becoming a redundant node;

[0054]

[0055] Except for the case where the node itself is used as a repair node for topology reconstruction, in the remaining cases, it is necessary to select a repair node from the redundant nodes to repair the disconnected area, that is, the topology reconstruction of the drone cluster network; after the redundant nodes in the network are selected, the repair nodes need to be repaired in the redundant nodes according to the weight ρ of the redundant nodes, which is calculated as follows:

[0056]

[0057] Where E rest is the node residual energy, d is the Euclidean distance between the redundant node and the disconnected area;

[0058] The redundant node with the largest weight is selected as the repair node to reconstruct the topology of the drone cluster network; selecting a repair node among the nodes with large redundancy values ​​means that the node has the largest overlapping area, and its movement has the least impact on the topological structure of the drone cluster network; from the above formula, it can be concluded that the greater the residual energy, the greater the probability that the redundant node with a closer distance will be selected as the repair node, which can ensure that the node can maintain a longer remaining working time after completing the topological reconstruction. The redundant node that is close to the repair area is easy to be selected as the repair node, which means selecting a node with a small moving distance, which can reduce the energy loss of the node during movement.

[0059] Furthermore, in the step 5, the automatic replacement method of the drone, after receiving a group of lost drones, the drone will decide whether automatic replacement is required, and the criteria for selecting a drone to perform the task are:

[0060] 1. The target priority it currently occupies must be lower than the lost drone, otherwise it will not be assigned to any target position

[0061] 2. It has the shortest path to the location of the lost drone

[0062] Each drone performs this process periodically, broadcasting its own cost in the node information message and obtaining the costs of other drones through the information message; under the competition scheme, each drone will decide whether to occupy the vacant target.

[0063] In the automatic replacement protocol, in order to achieve competition success, the UAV node must be the lowest priority node, and there must be no unassigned target point with a higher priority than the current target point, and it must be the nearest one among all unassigned higher priority target points.

[0064] Each UAV will compare its distance with all other UAVs with the same target point priority to ensure that it is the closest to the selected target point;

[0065] When we have many drones in the lowest priority position, it is guaranteed that at least one drone is closest to the unassigned target. When one’s own drone finds itself closest to the unassigned target position, the protocol enters the task competition scheme, which guarantees that whenever there is an unassigned target point, at least one drone is assigned to the target point even when there is a communication delay.

[0066] When UAV i updates its distance at time k+1, the distance d to the target i (k+1) has two cases;

[0067] Case 1: d i (k+1)≥cost i (k), at this time, the cost of drone i at time k+1 is as follows

[0068] cost i (k+1)=min{cost i (k), d i (k+1)}

[0069] =cost i (k)#

[0070] Case 2: d i (k+1)≤cost i (k); At this time, the cost of drone i at time k+1 is as follows

[0071] cost i (k+1)=min{cost i (k), d i (k+1)}

[0072] =d i (k+1)

[0073] ≤cost i (k)#

[0074] When there is a communication delay, if only one UAV enters the contention scheme when executing the above method, the UAV will be assigned to the vacated target according to the scheme;

[0075] When another drone enters the contention scheme, their costs at time k are cost i (k) and cost j (k), assuming cost i(k)>cost j (k), due to communication delay, min{cost i (k), cost j (k)} At time k, not every UAV is available, and both UAVs i and j use the information containing the costs of other UAVs for cost evaluation;

[0076] Consider all the possibilities for cost assessment of drones below:

[0077] 3. If drone i finds cost i (k)>cost j (kt L ), where t L Due to communication delay, UAV i retreats in this case; similarly, UAV j finds cost j (k)≤cost i (kt L ), UAV j will be assigned to the vacated target;

[0078] 4. If the drone finds cost i (k)≤cost j (kt L ) will assign UAV i to the target; similarly, UAV j finds cost j (k)≤cost i (kt L ) This is because there is no cost j (k)>cost i (kt L ); therefore, drone j will also be assigned;

[0079] It is possible that multiple drones meet the conditions of the competition scheme and fly to the unassigned target point at the same time; to prevent this from happening, we introduce another mechanism, namely the conflict prevention mechanism; the core of this mechanism is to assign a delay of n seconds from small to large according to the drone ID before the drone flies to the unassigned target.

[0080] Moreover, the proposed automatic replacement method is a dynamic process, which continuously performs automatic replacement until the replaced UAV reaches the unassigned target.

[0081] Furthermore, to prevent conflicts caused by multiple drone nodes satisfying the competing solutions at the same time, a conflict prevention mechanism is proposed. According to the conflict prevention mechanism, drones i and j have a delay of n = time_interval seconds; this means that drone i immediately goes to the unassigned target at time k, while drone j stays at its current location until time = k + time_interval; the cost of drone i at time k is reset to costi (k) = d i (k).

[0082] After one iteration of time update, the distance between drone i and the target at time k+1 is as follows:

[0083] d i (k+1)=d i (k)-v×time_interval#

[0084] Since the velocity v>0, this means that d i (k+1) <d i (k) and the following formula is obtained

[0085] cost i (k+1)=min{cost i (k), d i (k+1)}

[0086] =d i (k+1)

[0087] <d i (k+1)

[0088] =cost i (k)#

[0089] However, at time k+1, due to the conflict prevention mechanism, drone j stops at the position at time k; this means that d j (k+1)=d j (k); The cost of resetting UAV j can be obtained as follows

[0090] cost j (k+1)=min{cost j (k), d j (k+1)}

[0091] =cost j (k)#

[0092] Since at time k, cost j (k) = cost i (k), which means that at time k+1, we have cost j (k+1)>cost i (k+1)

[0093] Even if many drones have the same path to the target point, among the other drones, only the drone with the smallest ID has a cost less than the costs of the other drones; this will ensure that only one drone will automatically replace to the unassigned target location within a limited time.

[0094] (III) Beneficial effects

[0095] Compared with the prior art, the present invention provides a loss detection and automatic replacement and reconstruction method based on multi-UAV collaboration, which has the following beneficial effects:

[0096] 1. Cut point detection is performed in the local topology of the node, which avoids the large amount of inefficient information exchange in global cut point detection, reduces the communication overhead in the drone cluster network where it is difficult to obtain global information, and solves the problem of high time complexity of cut point detection in large-scale dynamic networks.

[0097] 2. A redundant node judgment strategy is introduced into the topology reconstruction algorithm, and nodes with higher redundancy are used as repair nodes in the topology reconstruction process, which reduces the impact on the overall structure of the network and reduces the number of invalid information exchanges in the original network.

[0098] 3. A loss detection and automatic replacement mechanism based on multi-UAV collaboration is proposed and theoretically proved, which improves the efficiency of rapid detection and reconstruction after UAV node failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 It is the overall flow chart of the present invention;

[0100] Figure 2 This is a schematic diagram of the two-dimensional structure of the flight node sensing area of ​​the present invention;

[0101] Figure 3 This is a schematic diagram of the consistency between the local cutting point and the global cutting point of the present invention;

[0102] Figure 4 This is a schematic diagram of the automatic replacement protocol flow of the present invention;

[0103] Figure 5 It is a schematic diagram of the competition scheme flow of the present invention. DETAILED DESCRIPTION

[0104] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0105] Example

[0106] like Figure 1-5 As shown, an embodiment of the present invention proposes a loss detection and automatic replacement and reconstruction method based on multi-UAV collaboration, including:

[0107] Step 1, build a network model: The entire drone cluster is constructed as a 1-connected network using the improved bird flock model. The model includes traditional opposition rules, cohesion rules, alignment rules, and inertia rules and avoidance rules that conform to the drone motion rules.

[0108] like Figure 2 As shown in the figure, the communication perception range of a single node in the drone cluster can be divided into the gravity perception area (R s ≤|V i V j |≤R a ), stable perception area (R r ≤|V i V j |≤R s ) and the repulsive force sensing area (|V k V i |≤R r ), where R a is the maximum communication distance of a single UAV node.

[0109] In addition, according to the characteristics of the movement of drone clusters, inertial rules and approach-avoidance rules are added to make them more adaptable to the movement characteristics of nodes in drone clusters.

[0110] (1) Inertia rule

[0111] In real-world scenarios, the change in the speed of individual drones is constrained by physical inertia and will not change dramatically in a short period of time. In the improved bird flock model, the unit direction vector of the individual speed of the drone at the current moment is directly used as the inertial component of the speed at the next moment, that is, formula (1):

[0112]

[0113] (2) Gravitational rules

[0114] The gravitational rule simulates the cohesive characteristics of individual drones in a swarm. In the improved bird flock model, individual drones try to get close to the geometric center of all neighboring drones within their sensing range, that is, formula (2):

[0115]

[0116] Among them, V i =(x i ,y i , z i ) is the coordinate of the target node i at time n. Assume that there are N neighboring drones in the gravity sensing area of ​​drone i, and V j =(x i ,y i , z i) represents the current coordinates of neighboring drone individual j, then the geometric center coordinates V of all neighboring nodes o =(x i ,y i , z i ) can be expressed as formula (3):

[0117]

[0118] (3) Repulsion rule

[0119] The repulsion rule simulates the collision avoidance behavior between individual drones in a cluster in real scenarios. When an individual drone senses that it is too close to its neighboring drone (the distance is less than the repulsion radius), a repulsion effect is generated for adjustment.

[0120]

[0121] Among them, V i =(x i ,y i ,z i ) is the coordinate of the target individual at time n. k =(x k ,y k ,z k ) represents the coordinates of the neighboring drone k in the individual repulsive sensing domain at time n.

[0122] (4) Alignment rules

[0123] The purpose of the alignment rule is to make all drones in the cluster form a unified movement trend. In the improved bird flock model, each drone and the surrounding drones l (the distance is less than the perception radius R) a )The direction of movement tends to be consistent, that is, formula (6)

[0124]

[0125] Where m is the number of neighboring drone nodes.

[0126] (5) Approach-avoidance rules

[0127] The approach and avoidance rules are used to simulate the behavior of individual drone swarms avoiding threat sources, as well as discovering and approaching mission (reconnaissance, strike) targets.

[0128]

[0129] Among them, V r is the coordinate of the target object or threat object. When the target is a task target, the sign after the equal sign in formula (7) is positive, and when the target is a threat object, the sign is negative.

[0130] Step 2: Based on the above conclusions, at time t, UAV node c can obtain its k-hop local topology G through k information exchanges. kc (t), get the adjacency table NT kc (t), and then traverse the topology structure with node c as the root node and perform the Tarjan algorithm. These marked key nodes cannot be used as mobile nodes in the automatic replacement step.

[0131] like Figure 3 The analysis shows the consistency between the global cut point and the k-jump cut point under different node densities and different k values. It can be seen that when k = 3, the node density D mean When it is greater than 4, the global cut point is consistent with the k-hop cut point.

[0132] The core of the Tarjan algorithm proposes two very important parameters (timestamps) on the traditional DFS: dfn[x] array and low[x] array. dfn[x] refers to the time of the first visit to the x node, and low[x] refers to the earliest time point that the x node can be traced back to. In Tarjan, DFS must first recursively visit the next node and then visit its own node.

[0133] Set a global variable times, which is an access time value. Each time a new node is accessed, a new time times++ is given to it. dfn is the times when it is first accessed. Then low needs to be modified and updated to determine the earliest time point that can be traced back.

[0134] There are two ways to determine whether a point is a cut point:

[0135] 1. If the node is the root node, if the node has two or more unconnected subtrees, deleting the node will get these subtrees, so the node is a cut point.

[0136] 2. If the node is not the root node, if there is no edge that can lead to the visited node during the traversal, low[next]>=dfn[now] (that is, there is no way back), then the node is a cut point.

[0137] Step 3: Redundant node marking. Some nodes in the drone cluster network are called redundant nodes. Redundant nodes can be used to reconstruct the drone cluster network topology and have little impact on the network structure.

[0138] In the research on topological reconstruction of UAV swarm networks, the boundary effect of the mission area is often ignored. On this basis, the present invention combines the node redundancy expectation with the expected coverage quality value considering the boundary effect to judge the redundant nodes. The redundancy of a single node in the network is expressed by the following formula (9):

[0139]

[0140] In the formula, based on the probability perception model of a single node, SD i To mark a single node v in the drone cluster network i The perception range, Represents node v i The set of all one-hop neighbor nodes; node v i The redundancy is defined as the ratio of the overlap between the node's own communication sensing area and the neighboring node's sensing area to the node's own sensing area.

[0141] All drone nodes are deployed in area S. The probability of any node being randomly deployed in area S is (S represents the area of ​​the drone scattering area). In order to be close to the actual situation, the sensing radius of the drone node is set to obey N(r0,δ 2 ) distribution. For the sensing radius, R i and R j Two drone nodes v i and v j If there are overlapping sensing areas in the topological structure, p ij To express the probability that two nodes have overlapping perception areas, the following formula (10) can be used to express it:

[0142]

[0143] Let E(p ij ) is p ij The mathematical expectation of E(p ij ) is calculated using the following formula (11):

[0144]

[0145] use To specifically represent the node v in the cluster i and v j The probability that there is no overlapping area is expressed as If the node v i There are k neighbors, use p ik To express the probability that this node has an overlapping area with any neighboring node, the probability that the sensing area does not overlap with any neighboring node can be obtained. but It can be expressed by the following formula (12):

[0146]

[0147] From the above formula, we can It is expressed as the following formula (13):

[0148]

[0149] If the expected value of the node's redundancy is greater than EDQ, that is, node v i The expected value of the redundancy ξ(i) satisfies the following formula (14), which means that the node v i ; The overlap ratio with its neighbor nodes is greater than the EDQ value, that is, let any node v in the network i ; When working in the minimum power state, the network connectivity will also reach the EDQ value.

[0150] E(ξ(i))≥E EDQ #(13)

[0151] Where E EDQ The EDQ value is shown in formula (15):

[0152]

[0153] Combining the above formulas, if node v i , the number of neighbor nodes k satisfies formula (16), that is, node v i : Meets the basic requirements for becoming a redundant node.

[0154]

[0155] Except for the case where the node itself is used as a repair node for topology reconstruction, in the remaining cases, a repair node needs to be selected from the redundant nodes to repair the disconnected area, that is, the topology reconstruction of the drone cluster network. After the redundant nodes in the network are selected, the repair nodes need to be repaired in the redundant nodes according to the weight ρ of the redundant nodes, which is calculated as follows (17):

[0156]

[0157] Where E rest is the node residual energy, and d is the Euclidean distance between the redundant node and the disconnected area.

[0158] The redundant node with the largest weight is selected as the repair node to reconstruct the topology of the drone cluster network. Selecting a repair node among nodes with large redundancy values ​​means that the node has the largest overlapping area, and its movement has the least impact on the topological structure of the drone cluster network; from the above formula, the greater the residual energy, the greater the probability that the redundant node with a closer distance is selected as the repair node, which can ensure that the node can maintain a longer remaining working time after completing the topological reconstruction. The redundant node close to the repair area is easy to be selected as the repair node, which means selecting a node with a small moving distance, which can reduce the energy loss of the node during movement.

[0159] Step 4, UAV packet loss detection: This protocol consists of two tasks. The purpose of Task 1 is to find the suspected lost UAV node, and Task 2 is to record the result of Task 1. Another local variable managed by both tasks is a timer to determine whether UAV q is lost.

[0160] Each drone broadcasts its node information message regularly, and other drones should receive the message. If the drone does not receive the node information message after the timeout, the suspicious set regains the ID of the node and sends it to task 2. When another layer receives the suspected node of task 1, it outputs the correct result of the detected lost drone set.

[0161] Step 5, automatic replacement of drones: After receiving a group of lost drones, the drone will decide whether automatic replacement is needed (the drone goes to the target location and occupies the target location of the lost drone).

[0162] The criteria for selecting a drone for this mission were:

[0163] 1. The target priority it currently occupies must be lower than the lost drone, otherwise it will not be assigned to any target position

[0164] 2. It should have the shortest path to the lost drone

[0165] According to this principle, the automatic replacement protocol diagram is as follows Figure 4 shown.

[0166] Each drone runs this process periodically, broadcasting its own cost (in a node information message) and obtaining the costs of other drones through information messages. Under the competition scheme, each drone will decide whether to occupy the vacated target.

[0167] In the automatic replacement protocol, in order to achieve competition success, the UAV node must be the lowest priority target point, and there must be no unassigned target point with a higher priority than it, and it must be the closest one among all unassigned higher priority target points.

[0168] Each UAV will compare its distance to all other UAVs with the same target point priority to ensure that it is closest to the selected target point.

[0169] When we have many drones at the lowest priority position, it is guaranteed that at least one drone is closest to the unassigned target. When the own drone finds itself closest to the unassigned target position, the protocol enters the task contention scheme. This scheme guarantees that at least one drone is assigned to the target point whenever there is an unassigned target point or even when there is a communication delay. Figure 5 Details of the competition proposal.

[0170] In the competition, the cost function of the i-th drone at time k is denoted as cost i (k)

[0171] Assume that drone i updates the distance to the target at time k, that is, d i (k), the cost function of UAV i is (18)

[0172] cost i (k) = min{cost i (k-1), d i (k)}#(18)

[0173] Where cost i (k-1) is the cost function of the previous time.

[0174] When drone i updates its distance at time k+1, we have the distance d to the target i (k+1). Distance d i (k+1) has two cases.

[0175] Case 1: d i (k+1)≥cost i (k), at this time, the cost of drone i at time k+1 is (19)

[0176] cost i (k+1)=min{cost i (k), d i (k+1)}

[0177] =cost i (k)#(19)

[0178] Case 2: d i (k+1)≤cost i (k). At this time, the cost of drone i at time k+1 is (20)

[0179] cost i (k+1)=min{cost i (k), d i (k+1)}

[0180] =d i (k+1)

[0181] ≤cost i (k)#(20)

[0182] In any case, the monotonicity cost i (k+1)≤cost i(k) is true, that is, the cost of each drone is monotonically decreasing. Similarly, each drone will lead to a monotonically decreasing number of drones.

[0183] Secondly, due to d i (k)≥0, which means the cost i (k) is bounded. Therefore, from the fact that a monotone decreasing function has a finite limit, it follows easily from the monotone convergence theorem that the cost of each drone in the swarm has a finite limit.

[0184] The above two conclusions, that the cost decreases monotonically and has a finite limit over time, mean that the cost function of each drone cannot grow unbounded, which also means that we can compare the costs between drones in a drone swarm under the competition scheme.

[0185] Assume that there is an unassigned target with a higher priority and one or two UAVs with the lowest priority. Under the contention scheme, by using the above cost calculation method, at least one UAV will not cancel the assignment of the target location even if there is a communication delay.

[0186] When there is communication delay, there are two situations when executing the above method:

[0187] In case 1, only one drone enters the contention scheme, so it is concluded that the drone will be assigned to the vacated target according to the scheme.

[0188] In case 2, another drone enters the contention scheme. Without loss of generality, we assume that drones i and j both enter the contention scheme, and their costs at time k are cost i (k) and cost j (k). Assume cost i (k)>cost j (k), as mentioned above, due to communication delay, min{cost i (k), cost j (k)} Not every drone is available at time k. Both drones i and j use the information containing the costs of other drones for cost evaluation.

[0189] Consider all the possibilities for cost assessment of drones below:

[0190] 5. If the drone i finds cost i (k)>cost j (kt L ), where t L Due to communication delay, UAV i retreats in this case. Similarly, UAV j finds cost j (k)≤costi (kt L ), UAV j will be assigned to the vacated target.

[0191] 6. If the drone finds cost i (k)≤cost j (kt L ) will assign UAV i to the target. Similarly, UAV j finds cost j (k)≤cost i (kt L ) This is because there is no cost j (k)>cost i (kt L ). Therefore, drone j will also be assigned.

[0192] In summary, no matter what situation the two drones are in, at least one of them will be assigned to the vacated target.

[0193] It is possible that multiple drones meet the conditions of the competition scheme and thus fly to the unassigned target point at the same time. For example, two drones are hovering at their target positions respectively and have the same path to the unassigned target at time k=0, 1, 2... In this case, in order to prevent it from happening, we introduce another mechanism, namely the conflict prevention mechanism. The core of this mechanism is to assign n seconds of delay from small to large according to the drone ID before the drone flies to the unassigned target. The detailed mechanism is:

[0194] According to the conflict prevention mechanism, drones i and j have a delay of n = time_interval seconds. This means that drone i immediately goes to the unassigned target at time k, while drone j stays at its current location until time = k + time_interval. The cost of drone i at time k is reset to cost i (k) = d i (k).

[0195] After one iteration of time update, the distance between UAV i and the target at time k+1 is:

[0196] d i (k+1)=d i (k)-v×time_interval#(21)

[0197] Since the velocity v>0, this means that d i (k+1) <d i (k) and the following formula (22) is obtained

[0198] cost i (k+1)=min{costi (k), d i (k+1)}

[0199] =d i (k+1)

[0200] <d i (k+1)

[0201] =cost i (k)#(22)

[0202] However, at time k+1, due to the conflict prevention mechanism, drone j stops at the position at time k. This means that d j (k+1)=d j (k); The cost of resetting UAV j can be obtained from equation (23):

[0203] cost j (k+1)=min{cost j (k), d j (k+1)}

[0204] =cost j (k)#(23)

[0205] Since at time k, cost j (k) = cost i (k), which means that at time k+1, we have cost j (k+1)>cost i (k+1)

[0206] Even if many drones have the same path to the target point, among the other drones, only the drone with the smallest ID has a cost less than the cost of other drones. This will ensure that only one drone will automatically replace to the unassigned target location within a limited time.

[0207] In summary, the feasibility of this scheme is proved, that is: assuming that there is an unassigned target with a higher priority and there is at least one UAV with the lowest priority. By applying the automatic replacement protocol to each UAV, it is ensured that only one UAV in the multi-UAV system occupies the vacated target.

[0208] Moreover, the proposed automatic replacement method is a dynamic process, which continuously performs automatic replacement until the replaced UAV reaches the unassigned target.

[0209] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A loss detection and automatic replacement reconstruction method based on multi-UAV collaboration, Features: The following steps are included: Step 1, construct an improved bird flock model: the entire drone cluster is constructed as a 1-connected network using the improved bird flock model; The improved flock model includes traditional gravitational rules, alignment rules, repulsion rules, as well as inertial rules and approach-avoidance rules that conform to the UAV motion rules; the multi-UAV system utilizes the wireless communication mode, namely the User Datagram Protocol; Step 2, local cut point detection: perform k-Tarjan cut point detection on the 3-hop network of each node in the singly connected network constructed in step 1, detect the cut points in the network and mark them as key nodes when exchanging neighbor information. These nodes cannot be used as mobile nodes in the automatic replacement step; Step 3, redundant node marking: In the drone cluster, redundancy refers to the relevance and repetitiveness of information provided by unmanned nodes. The node redundancy is judged by combining the expected redundancy and the expected coverage value considering the boundary effect, that is, preparing mobile nodes for automatic replacement in step 5; Step 4, UAV packet loss detection: This detection protocol consists of two tasks; the purpose of task 1 is to find the suspected lost UAV node, and task 2 is to record the result of task 1; another local variable managed by both tasks is a timer to determine whether the UAV node is lost; Step 5, automatic replacement of drones: After receiving a group of lost drones, the drones will decide whether automatic replacement is needed based on the proposed automatic replacement protocol, and propose corresponding competition schemes to determine the optimal replacement drone node.

2. According to the method of loss detection and automatic replacement and reconstruction based on multi-UAV collaboration described in claim 1, Features: The communication sensing range of a single node in the drone cluster in step 1 can be divided into a gravitational sensing area (R s ≤|V i V j |≤R a ), stable perception area (R r ≤|V i V j |≤R s ) and the repulsive force sensing area (|V k V i |≤R r ), where R a is the maximum communication distance of a single UAV node; In addition, according to the characteristics of the movement of drone clusters, inertial rules and approach-avoidance rules are added to make it more adaptable to the movement characteristics of nodes in drone clusters; (1) Inertia rule In real-world scenarios, the change in the speed of individual drones is constrained by physical inertia and will not change dramatically in a short period of time. The improved flock model uses the unit direction vector of the individual speed of the drone at the current moment directly as the inertial component of the speed at the next moment, which is the following formula: (2) Gravitational rules The gravitational rule simulates the cohesive characteristics of individual drone clusters; in the improved bird flock model, individual drones try to get close to the geometric center of all neighboring drones within their perception range, which is the following formula: Among them, V i =(x i ,y i ,z i ) is the coordinate of the target node i at time n; assuming that there are N neighboring drones in the gravity sensing area of ​​drone i, and V j =(x i ,y i ,z i ) represents the current coordinates of neighboring drone individual j, then the geometric center coordinates V of all neighboring nodes o =(x i ,y i ,z i ) can be expressed as follows: (3) Repulsion rule The repulsion rule simulates the collision avoidance behavior between individual swarm drones in real scenarios. When an individual drone senses that it is too close to its neighboring drone, that is, the distance is smaller than the repulsion radius, a repulsion effect is generated for adjustment. Among them, V i =(x i ,y i ,z i ) is the coordinate of the target individual at time n; V k =(x k ,y k ,z k ) represents the coordinates of neighboring drone k in the individual repulsion perception domain at time n; (4) Alignment rules The purpose of the alignment rule is to make all drones in the cluster form a unified movement trend; in the improved bird flock model, the distance between each drone and the surrounding is less than the perception radius R a The movement direction of the UAV tends to be consistent, that is, the following formula Where m is the number of neighboring drone nodes; (5) Approach-avoidance rules The approach-avoidance rule is used to simulate the behavior of individual drone swarms avoiding threat sources, as well as discovering and approaching tasks, such as reconnaissance and attacking targets; Among them, V r is the coordinate of the target object or threat object; when the target is a mission target, the sign after the equal sign in the above formula is positive, and when the target is a threat object, the sign is negative.

3. According to the loss detection and automatic replacement reconstruction method based on multi-UAV collaboration described in claim 1, Features: The redundant node marking in step 3 combines the node redundancy expectation with the expected coverage quality value considering the boundary effect to determine the redundant nodes; The redundancy of a single node in the network is expressed as follows: In the formula, based on the probability perception model of a single node, SD i To mark a single node v in the drone cluster network i The perception range, Represents node v i The set of all one-hop neighbor nodes; node v i The redundancy is defined as the ratio of the overlap between the node's own communication sensing area and the neighboring node's sensing area to the node's own sensing area; All drone nodes are deployed in area S. The probability of any node being randomly deployed in area S is (|S| represents the area of ​​the drone scattering area); in order to be close to the actual situation, the sensing radius of the drone node is set to obey N(r 0 ,δ 2 ) distribution; for the sensing radius, they are R i and R j Two drone nodes v i and v j If there are overlapping sensing areas in the topological structure, p ij To express the probability that two nodes have a sensed overlapping area, it can be expressed as follows: Let E(p ij ) is p ij The mathematical expectation of E(p ij ) is calculated using the following formula: use To specifically represent the node v in the cluster i and v j The probability that there is no overlapping area is expressed as If the node v i There are k neighbors, use p ik To express the probability that this node has an overlapping area with any neighboring node, the probability that the sensing area does not overlap with any neighboring node can be obtained. but It can be expressed as follows: From the above formula, we can It is expressed as the following formula: If the expected value of the node's redundancy is greater than EDQ, that is, node v i The expected value of the redundancy ξ(i) satisfies the following formula, which means that node v i ; The overlap ratio with its neighbor nodes is greater than the EDQ value, that is, let any node v in the network i ; The network connectivity will also reach the EDQ value when working in the minimum power state; E(ξ(i))≥E EDQ # E in the formula EDQ The EDQ value is shown as follows: Combining the above formulas, if node v i , the number of neighbor nodes k satisfies the following formula, that is, node v i : Meet the basic conditions for becoming a redundant node; Except for the case where the node itself is used as a repair node for topology reconstruction, in the remaining cases, it is necessary to select a repair node from the redundant nodes to repair the disconnected area, that is, the topology reconstruction of the drone cluster network; after the redundant nodes in the network are selected, the repair nodes need to be repaired in the redundant nodes according to the weight ρ of the redundant nodes, which is calculated as follows: Where E rest is the node residual energy, d is the Euclidean distance between the redundant node and the disconnected area; The redundant node with the largest weight is selected as the repair node to reconstruct the topology of the drone cluster network; selecting a repair node among the nodes with large redundancy values ​​means that the node has the largest overlapping area, and its movement has the least impact on the topological structure of the drone cluster network; from the above formula, it can be concluded that the greater the residual energy, the greater the probability that the redundant node with a closer distance will be selected as the repair node, which can ensure that the node can maintain a longer remaining working time after completing the topological reconstruction. The redundant node that is close to the repair area is easy to be selected as the repair node, which means selecting a node with a small moving distance, which can reduce the energy loss of the node during movement.

4. According to the method of loss detection and automatic replacement and reconstruction based on multi-UAV collaboration described in claim 1, Features: In the step 5, the automatic replacement method of the drone, after receiving a group of lost drones, the drone will decide whether automatic replacement is needed. The criteria for selecting a drone to perform the task are:

1. The target priority it currently occupies must be lower than the lost drone, otherwise it will not be assigned to any target position 2. It has the shortest path to the location of the lost drone Each drone performs this process periodically, broadcasting its own cost in the node information message and obtaining the costs of other drones through the information message; under the competition scheme, each drone will decide whether to occupy the vacant target. In the automatic replacement protocol, in order to achieve competition success, the UAV node must be the lowest priority node, and there must be no unassigned target point with a higher priority than the current target point, and it must be the nearest one among all unassigned higher priority target points. Each UAV will compare its distance with all other UAVs with the same target point priority to ensure that it is the closest to the selected target point; When we have many drones in the lowest priority position, it is guaranteed that at least one drone is closest to the unassigned target. When one’s own drone finds itself closest to the unassigned target position, the protocol enters the task competition scheme, which guarantees that whenever there is an unassigned target point, at least one drone is assigned to the target point even when there is a communication delay. When UAV i updates its distance at time k+1, the distance d to the target i (k+1) has two cases; Case 1: d i (k+1)≥cost i (k), at this time, the cost of drone i at time k+1 is as follows cost i (k+1)=min{cost i (k),d i (k+1)} =cost i (k)# Case 2: d i (k+1)≤cost i (k); At this time, the cost of drone i at time k+1 is as follows cost i (k+1)=min{Pcost i (k),d i (k+1)} =d i (k+1) ≤cost i (k)# When there is a communication delay, if only one UAV enters the contention scheme when executing the above method, the UAV will be assigned to the vacated target according to the scheme; When another drone enters the contention scheme, their costs at time k are cost i (k) and cost j (k), assuming cost i (k)>cost j (k), due to communication delay, min{cost i (k),cost j (k)} At time k, not every UAV is available, and both UAVs i and j use the information containing the costs of other UAVs for cost evaluation; Consider all the possibilities for cost assessment of drones below:

1. If drone i finds cost i (k)>cost j (kt L ), where t L Due to communication delay, the drone retreats in this case. ; Similarly, drone j finds cost j (k)≤cost i (kt L ), UAV j will be assigned to the vacated target; 2. If the drone finds cost i (k)≤cost j (kt L ) will assign UAV i to the target; similarly, UAV j finds cost j (k)≤cost i (kt L ) This is because there is no cost j (k)>cost i (kt L ); therefore, drone j will also be assigned; It is possible that multiple drones meet the conditions of the competition scheme and fly to the unassigned target point at the same time; to prevent this from happening, we introduce another mechanism, namely the conflict prevention mechanism; the core of this mechanism is to assign a delay of n seconds from small to large according to the drone ID before the drone flies to the unassigned target. Moreover, the proposed automatic replacement method is a dynamic process, which continuously performs automatic replacement until the replaced UAV reaches the unassigned target.

5. According to the method of loss detection and automatic replacement and reconstruction based on multi-UAV collaboration described in claim 1, Features: To prevent conflicts caused by multiple drone nodes satisfying the competing solutions at the same time, a conflict prevention mechanism is proposed. According to the conflict prevention mechanism, drones i and j have a delay of n = time_interval seconds; this means that drone i immediately goes to the unassigned target at time k, while drone j stays at its current location until time = k + time_interval; the cost of drone i at time k is reset to cost i (k) = d i (k). After one iteration of time update, the distance between drone i and the target at time k+1 is as follows: <h2 style=";text-align:left;direction:ltr">d<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> (k+1) = d<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> (k)-v×time_interval# Since velocity v>0, this means that d i (k+1) <d i (k) and the following formula is obtained cost i (k+1)=min{cost i (k),d i (k+1)} =d i (k+1) <d i (k+1) =cost i (k)# However, at time k+1, due to the conflict prevention mechanism, drone j stops at the position at time k; this means that d j (k+1)=d j (k); The cost of resetting UAV j can be obtained as follows cost j (k+1)=min{cost j (k),d j (k+1)} =cost j (k)# Since at time k, cost j (k) = cost i (k), which means that at time k+1, we have cost j (k+1)>cost i (k+1) Even if many drones have the same path to the target point, among the other drones, only the drone with the smallest ID has a cost less than the costs of the other drones; this will ensure that only one drone will automatically replace to the unassigned target location within a limited time.