An autonomous search method for complex scene targets based on mother-and-child drones

Abstract

The present invention relates to a complex scene target autonomous search method based on a mother-and-child drone, comprising: after the mother drone receives the search task, it determines whether it can complete the search alone, if not, it analyzes the number of child drones required for the target search and specifies the child drone separation rule; after the selected child drone receives the corresponding search task, it separates from the mother drone, and the child drone performs the target search alone or in collaboration with the mother drone; during the target search process, the child drone maintains communication with the mother drone through a formulated multi-machine mobile ad hoc network mechanism; when the child drone performs the target search alone or in collaboration with the mother drone, a Voronoi diagram is used to divide the search area into areas consistent with the number of drones, and then a particle swarm algorithm is used to update the movement speed and position of each drone. Compared with the prior art, the present invention realizes the target search of the mother-and-child drone in complex scenes, and can also be used for tasks such as multi-target tracking to improve the search efficiency.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) search, and in particular to an autonomous search method for complex scene targets based on mother-child UAVs. Background Art

[0002] The method of obtaining target scene image information through drone aerial photography has the advantages of low-altitude operation, wide coverage area, strong maneuverability, high efficiency, and no hindrance from terrain and environment. It can be widely used in post-disaster search and rescue, target search and following tasks. However, a single drone has problems such as low search efficiency and can only track a single target. In some special scenarios, the size of larger drones is limited and cannot complete specific tasks alone, which reduces the detection ability of drones. Therefore, whether in the civilian or military fields, small-scale clustering of drones is a trend. UAVs can share data and sensor status information through a certain communication mechanism and control each other's interfaces, which can greatly improve the efficiency of drone work. However, the current research on the coordination of small-scale clusters of drones is still relatively rudimentary, lacking systematic and complete application methods. Summary of the invention

[0003] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method for autonomously searching targets in complex scenes based on mother-child drones, so as to realize target search of mother-child drones in complex scenes, and can also be used for tasks such as multi-target tracking to improve search efficiency.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] A complex scene target autonomous search method based on a mother-child UAV, characterized in that the mother-child UAV includes a mother aircraft and multiple child aircrafts, and the search method includes: after the mother aircraft receives a search task, it determines whether it can complete the search alone, if not, it analyzes the number of child aircrafts required for the target search and specifies the child aircraft separation rules; after the selected child aircraft receives the corresponding search task, it separates from the mother aircraft, and the child aircraft searches for the target alone or in collaboration with the mother aircraft;

[0006] During the target search process, the slave machine maintains communication with the master machine through the established multi-machine mobile ad hoc network mechanism;

[0007] When the sub-machine searches for a target alone or in collaboration with the main machine, the Voronoi diagram is used to divide the search area into areas consistent with the number of drones, and then the particle swarm algorithm is used to update the movement speed and position of each drone.

[0008] Furthermore, the search method specifically includes the following steps:

[0009] A1. After receiving the search task, the master machine determines whether it can complete the search alone. If so, it executes step S2; if not, it executes step A3;

[0010] A2, the mother aircraft carries the sub-aircraft to perform the search mission, searching for targets until all targets are searched; if the mother aircraft fails to search for all targets within the specified time, or encounters terrain that cannot be explored, execute step A5;

[0011] A3. Analyze the number of sub-machines required for target search and specify the rules for sub-machines to disengage. After the selected sub-machine receives the search mission instruction from the mother machine, it starts to the specified speed, and the mother machine releases the sub-machine. At the same time, the mother and child machines maintain communication for data transmission.

[0012] A4, the sub-machine performs the search task and searches for the target until all the targets are searched; if the sub-machine fails to search for all the targets within the specified time, execute step A5;

[0013] A5. The main machine and the sub-machine conduct a coordinated search until all targets are searched.

[0014] Furthermore, when the sub-machine searches for a target alone or in cooperation with the main machine, it specifically includes:

[0015] S1. Before the search mission begins, the search area is divided into n sub-areas using the Voronoi diagram according to the number of drones n, and the number of particles of the particle swarm algorithm is determined. The search area is abstracted into n×m grids of equal size. The size of the grid is related to the drone perception area. The grid information corresponds to an n×m-dimensional matrix, and the initial value of the matrix is ​​set to 0.

[0016] S2, the initial point of the particle swarm is selected at multiple points with the highest probability of target occurrence, and each drone goes straight to the location where the target is most likely to appear in each sub-area at a set speed;

[0017] S3. When the drone detects a certain area, if there is no target in this area, the value of the corresponding matrix position is set to 1, indicating that this area has been detected; if a target is found in this area, the value of the corresponding matrix position is set to 2, indicating that the target is found in this area; and the search results are sent to the mother machine;

[0018] S4, calculate the individual extreme value of each drone and the global extreme value of the entire system, and update the cost function; at the same time, update the speed and position information of the drone;

[0019] S5. Repeat step S4 until the matrix values ​​corresponding to all grids are 1 or 2, or the number of values ​​2 in the matrix is ​​equal to the target number, and then stop searching.

[0020] Furthermore, in step S4, the cost function is expressed as:

[0021] Cost = α*Dis(o,sd)+β / R(o)

[0022] Among them, Dis(o, sd) represents the Euclidean distance from point o to the point sd with the highest probability of target appearance, R(o) represents the growth rate of the coverage area from the current point to point o, and α and β are the weight coefficients of distance and coverage rate.

[0023] Furthermore, in step S4, the expression for updating the speed and position information of the drone is:

[0024]

[0025] Among them, k and k+1 are the number of iterations; ω is the inertia coefficient, that is, the weight of the last speed; δ1 and δ2 are constant coefficients; γ1 and γ2 are random numbers between (0, 1); v i is the velocity of the ith particle; o i is the i-th particle; L bv is the individual extreme value of the ith particle; G bv is the global extreme value.

[0026] Furthermore, the multi-machine mobile self-organizing network mechanism includes: each sub-machine maintains a status information table of the remaining drones; in the default state, the sub-machine is directly connected to the mother machine in a single hop to form a star network; when the distance between the sub-machine and the mother machine is greater than the wireless network transmission distance threshold, the sub-machine with the same status as the drone on the side of the mother machine will be selected as the forwarding node, and the mother machine will send a forwarding communication task to the node, and maintain communication with the mother machine through multi-hop transmission.

[0027] Furthermore, the state information table is expressed as:

[0028] Addr(Nv,Wc Dis,Ori)

[0029] Among them, Nv is the drone number; Wc is the working status of the drone; Dis is the distance between the drone and the mother aircraft; Ori is the side status of the drone on the mother aircraft.

[0030] Furthermore, the sub-machine transmits information including its own flight speed, direction, remaining power and search target status to the main machine at a frequency of 1Hz.

[0031] Furthermore, when a certain sub-machine is selected as a forwarding node, the mother machine will transmit the flight speed and direction of the sub-machine that needs communication forwarding at a specific frequency. The forwarding node sub-machine will fly in the same direction while ensuring that the distance between itself and the mother machine is less than the wireless network transmission distance threshold; when the distance between the forwarding node sub-machine itself and the mother machine is greater than the wireless network transmission distance threshold, another forwarding node sub-machine needs to be selected.

[0032] Furthermore, the submachine separation rule is to follow the principle that the overall center of gravity of the submachine and the main machine is maintained near the central axis.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention effectively improves the efficiency of target search in complex scenes through the coordinated work of the mother and child UAVs. Through the configuration of the child UAV, the invention can penetrate into the places that the mother UAV cannot reach, thereby improving the site adaptability.

[0035] At the same time, the present invention utilizes the combination of the Voronoi diagram and the particle swarm algorithm to enable the UAV to reach the location where the target may appear more quickly, making the search faster and more convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of the structure of the mother-child UAV of the present invention.

[0037] Figure 2 Schematic diagram of the search method of the present invention.

[0038] Figure 3 It is a schematic diagram of the multi-machine mobile ad hoc network of the present invention.

[0039] Figure 4 Schematic diagram of the target search process of the present invention.

[0040] Figure numerals: 1-mother machine, 11-docking mechanism, 12-guiding device, 2-child machine, 21-electromagnet, 22-guiding mechanism. DETAILED DESCRIPTION

[0041] 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.

[0042] This embodiment provides a method for autonomously searching for targets in complex scenes based on a mother-and-child UAV.

[0043] like Figure 1As shown, the mother-child drone includes a mother machine 1 and multiple daughter machines 2, and the number of daughter machines 2 is at least three. In this embodiment, three daughter machines 2 are taken as an example. Three docking mechanisms 11 and guiding devices 12 are provided below the mother machine 1. The three daughter machines 2 are numbered 0, 1 and 2, respectively. No. 0 is located directly below the mother machine 1, and No. 1 and No. 2 are distributed on the left and right sides below to ensure that the overall center of gravity of the mother-child drone 1 is near the central axis of the mother machine 1. The docking mechanism 11 specifically adopts a metal plate, and the docking mechanism 11 is used to connect the power-off type electromagnet 21 on the top of each daughter machine 2. The guiding device 12 specifically includes 2 UWB positioning anchor points and 1 Apriltag label. A guiding mechanism 22 is provided on one side of the daughter machine 2, including 1 UWB module and 1 visual camera. The daughter machine 2 and the mother machine 1 use the cooperation of the guiding module and the guiding mechanism 22 to formulate a relative positioning and guiding strategy to assist the docking mechanism 11 to complete the docking of the mother-child drone 1 with each other.

[0044] The relative positioning guidance strategy is as follows: When the slave returns to the mother machine, the relative distance between the slave and the mother machine is calculated by the UWB AOA method through the UWB module on the slave side and the two UWB positioning anchor points on the mother machine, with an accuracy of decimeter level. The slave approaches the mother machine with the path planning goal of reducing the relative distance with the mother machine. When the slave approaches the mother machine, the slave uses the camera to search for the Apriltag tag on the mother machine side. After searching, it uses the Apriltag tag for visual precise positioning, completes the relative positioning guidance of the slave and mother machine, and realizes the docking of the slave and mother machine.

[0045] The search method of this embodiment is as follows: after the mother machine accepts the search task, it determines whether it can complete the search alone. If not, it analyzes the number of child machines required for the target search and specifies the child machine's disengagement rules; the selected child machine disengages from the mother machine after receiving the corresponding search task, and the child machine searches for the target alone or in collaboration with the mother machine. During the target search process, the child machine maintains communication with the mother machine through the established multi-machine mobile ad hoc network mechanism. When the child machine searches for the target alone or in collaboration with the mother machine, the Voronoi diagram is used to divide the area to be searched into areas consistent with the number of drones, and then the particle swarm algorithm is used to update the movement speed and position of each drone. Specific details are as follows: Figure 2 As shown:

[0046] Step A0: Search starts and initializes.

[0047] Step A1: After receiving the search task, the master machine determines whether it can complete the search alone. If so, execute step A2; if not, execute step A3.

[0048] Step A2: The mother machine carries the sub-machine to perform the search mission and search for the target until all the targets are searched; if the mother machine fails to search for all the targets within the specified time, or encounters terrain that cannot be explored (for example, encountering special situations such as too narrow a passage or too scattered targets), execute step A5.

[0049] Step A3: Analyze the number of sub-machines required for target search and specify the rules for sub-machine disengagement; after the selected sub-machine receives the search task instruction from the mother machine, it starts to the specified speed, and the mother machine releases the sub-machine; at the same time, the mother and child machines maintain communication for data transmission. The mother machine releases the sub-machine by specifically disconnecting the electromagnet, so that the mother-child machine docking mechanism loses its magnetic force, and the mother machine releases the sub-machine.

[0050] Step A4: The slave performs the search task and searches for the target until all the targets are searched. If the slave fails to search for all the targets within the specified time, step A5 is executed.

[0051] Step A5: The master and slave machines conduct a collaborative search until all targets are searched.

[0052] In the search method, the separation rule of the sub-machine follows the principle that the overall center of gravity of the mother and sub-machine is maintained near the central axis. If the required number of sub-machines is one, the sub-machine of the docking mechanism No. 0 will be separated from the mission; if the required number of sub-machines is two, the sub-machines of the docking mechanisms No. 1 and No. 2 will be separated from the mission; if the required number of sub-machines is three, the sub-machines of the docking mechanisms No. 0, No. 1 and No. 2 will all be separated from the mission.

[0053] In the search method, the multi-machine mobile ad hoc network mechanism includes: each sub-machine maintains the status information table of the other drones; the status information table is expressed as: Addr (Nv1, Wc1, Dis1, Ori1; Nv2, Wc2, Dis2, Ori2; Nv3, Wc3, Dis3, Ori3; Nv4, Wc4, Dis4, Ori4), where Nv is the drone number, Wc is the working status of the drone (0 means not executing the task, 1 means executing the task), Dis is the distance between the drone and the mother machine, and Ori is the side of the mother machine on which the drone is (0 means the left side, 1 means the right side). In the default state, the sub-machine and the mother machine are directly connected by a single hop to form a star network. When the distance Dis between the sub-machine and the mother machine is greater than the wireless network transmission distance threshold, the sub-machine with the same Ori will be selected as the forwarding node, and the mother machine will send the forwarding communication task to the node, and maintain communication with the mother machine through multi-hop transmission. The slave unit transmits information to the master unit at a frequency of 1Hz, including its own flight speed, direction, remaining power, search target status, etc. Figure 3 As shown, the micro sub-machine in the figure is the sub-machine.

[0054] When a sub-machine is selected as a forwarding node, the mother machine will transmit the flight speed and direction of the sub-machine that needs communication forwarding at 1HZ frequency, and the forwarding node sub-machine will fly in the same direction while ensuring that its own Dis is less than the wireless network transmission distance threshold. When the forwarding node sub-machine Dis is greater than the wireless network transmission distance threshold, if the task requires it, the same method will be used to select the second forwarding node sub-machine. And so on.

[0055] In the search method, when the sub-machine searches for a target alone or in cooperation with the main machine, it specifically includes:

[0056] Step s1: Before the search mission begins, the search area is divided into n areas using the Voronoi diagram according to the number of drones n, and the number of particles of the particle swarm algorithm is determined. The area to be searched is abstracted into n×m grids of equal size. The size of the grid is related to the drone perception area. The grid information corresponds to an n×m-dimensional matrix, and the initial values ​​of the matrix are all set to 0.

[0057] In the planar Voronoi diagram, the point set {p1, p2, p3…, p n Each point in the polygon is regarded as a seed, and expands at the same rate until they meet each other and stop expanding, thus forming a figure of the entire area. The points in the polygon expanded by any point satisfy

[0058] T i ={x:d(x,p i )<d(x,p j )|p i ,p j ∈S,p i ≠p j}

[0059] Among them, T i is the set of all points in the i-th seed polygon; x is any point in the polygon; p i is the i-th seed; p j is the jth seed; S is the space of all sets; d represents the Euclidean distance between two points, and the distance from any point in the polygon to the station in the polygon is less than the distance to the other stations. This method has a low computational complexity.

[0060] Step s2: The initial point of the particle swarm is selected from multiple points with the highest probability of target occurrence. Each drone goes to the most likely location of the target in each sub-area in a straight line, that is, the initial position and speed of the drone are initialized by making full use of known information to minimize the search range.

[0061] Step s3: When the drone detects a certain area, if there is no target in this area, the value of the corresponding matrix position is set to 1, indicating that this area has been detected; if a target is found in this area, the value of the corresponding matrix position is set to 2, indicating that the target is found in this area. The search results are sent to the mother machine.

[0062] Step s4: Calculate the individual extremum of each drone and the global extremum of the entire system. Each particle has an optimized cost function that determines its next direction of movement and speed, and finds the optimal solution through iterative updates. During the iteration process, the best position reached by a particle itself is called the individual extremum of the particle, and the best position reached by all particles is called the global extremum. Each particle will record its search history and track the best position of the target in the last iteration, and update its own speed and position according to the individual extremum and the global extremum.

[0063] Assume that there are N particles in the search space, o i represents the i-th particle, v i represents the velocity of the ith particle, L bv represents the individual extreme value of the i-th particle, G bv Represents the global extreme value, and the cost function is updated by the following formula

[0064] Cost = α*Dis(o,sd)+β / R(o)

[0065] Where Dis(o, sd) represents the Euclidean distance from point o to the point with the highest probability of target occurrence, R(o) represents the growth rate of the coverage area from the current point to point o, and α and β are the weight coefficients of distance and coverage rate.

[0066] Use the following formula to update the speed and position information of the drone

[0067]

[0068] Where k, k+1 are the number of iterations, ω is the inertia coefficient, that is, the weight of the previous velocity, δ1, δ2 are constant coefficients, and γ1, γ2 are random numbers between (0, 1).

[0069] Step s5: Repeat s4 until the matrix values ​​corresponding to all grids are 1 or 2, or the number of values ​​2 in the matrix is ​​equal to the target number, and then stop searching.

[0070] 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 complex scene target autonomous search method based on a mother-and-child drone, characterized in that: The mother-child UAV includes a mother aircraft and multiple child aircrafts. The search method includes: after the mother aircraft receives the search task, it determines whether it can complete the search alone. If not, it analyzes the number of child aircrafts required for the target search and specifies the child aircraft separation rules; after the selected child aircraft receives the corresponding search task, it separates from the mother aircraft, and the child aircraft searches for the target alone or in cooperation with the mother aircraft; During the target search process, the slave machine maintains communication with the master machine through the established multi-machine mobile ad hoc network mechanism; When the sub-machine searches for a target alone or in collaboration with the main machine, the Voronoi diagram is used to divide the search area into areas that are consistent with the number of drones, and then the particle swarm algorithm is used to update the movement speed and position of each drone; When the slave unit searches for a target alone or in coordination with the master unit, it specifically includes: S1. Before the search mission begins, the search area is divided into n sub-areas using the Voronoi diagram according to the number of drones n, and the number of particles of the particle swarm algorithm is determined. The search area is abstracted into n×m grids of equal size. The size of the grid is determined by the drone perception area. The grid information corresponds to an n×m-dimensional matrix, and the initial value of the matrix is ​​set to 0. S2, the initial point of the particle swarm is selected at multiple points with the highest probability of target occurrence, and each drone goes straight to the location where the target is most likely to appear in each sub-area at a set speed; S3. When the drone detects a certain area, if there is no target in this area, the value of the corresponding matrix position is set to 1, indicating that this area has been detected; if a target is found in this area, the value of the corresponding matrix position is set to 2, indicating that the target is found in this area; and the search results are sent to the mother machine; S4, calculate the individual extreme value of each drone and the global extreme value of the entire system, and update the cost function; at the same time, update the speed and position information of the drone; S5, repeat step S4 until the matrix values ​​corresponding to all grids are 1 or 2, or the number of values ​​2 in the matrix is ​​equal to the target number, and then stop searching; In step S4, the cost function is expressed as: Cost = α*Dis(o,sd)+β / R(o) Where Dis(o, sd) represents the Euclidean distance from point o to the point sd with the highest probability of target occurrence, R(o) represents the growth rate of the coverage area from the current point to point o, and α and β are the weight coefficients of distance and coverage rate; In step S4, the expression for updating the speed and position information of the drone is: Among them, k and k+1 are the number of iterations; ω is the inertia coefficient, that is, the weight of the last speed; δ1 and δ2 are constant coefficients; γ1 and γ2 are random numbers between (0, 1); v i is the velocity of the ith particle; o i is the i-th particle; L bv is the individual extreme value of the ith particle; G bv is the global extreme value.

2. According to claim 1, a complex scene target autonomous search method based on a mother-child UAV is characterized in that: The search method specifically includes the following steps: A1. After receiving the search task, the master machine determines whether it can complete the search alone. If so, execute step A2; if not, execute step A3; A2, the mother aircraft carries the sub-aircraft to perform the search mission, searching for targets until all targets are searched; if the mother aircraft fails to search for all targets within the specified time, or encounters terrain that cannot be explored, execute step A5; A3. Analyze the number of sub-machines required for target search and specify the rules for sub-machines to disengage. After the selected sub-machine receives the search mission instruction from the mother machine, it starts to the specified speed, and the mother machine releases the sub-machine. At the same time, the mother and child machines maintain communication for data transmission. A4, the sub-machine performs the search task and searches for the target until all the targets are searched; if the sub-machine fails to search for all the targets within the specified time, execute step A5; A5. The main machine and the sub-machine conduct a coordinated search until all targets are searched.

3. The method for autonomously searching for targets in complex scenes based on a mother-and-child UAV according to claim 1 is characterized in that: The multi-machine mobile self-organizing network mechanism includes: each sub-machine maintains the status information table of the other drones; in the default state, the sub-machine is directly connected to the mother machine with a single hop to form a star network; when the distance between the sub-machine and the mother machine is greater than the wireless network transmission distance threshold, the sub-machine with the same status as the mother machine on the side of the drone will be selected as the forwarding node, and the mother machine will send a forwarding communication task to the node to maintain communication with the mother machine through multi-hop transmission.

4. The method for autonomously searching for targets in complex scenes based on a mother-and-child UAV according to claim 3 is characterized in that: The state information table is expressed as: Addr(Nv,Wc Dis,Ori) Among them, Nv is the drone number; Wc is the working status of the drone; Dis is the distance between the drone and the mother aircraft; Ori is the side status of the drone on the mother aircraft.

5. The method for autonomously searching for targets in complex scenes based on a mother-and-child UAV according to claim 3 is characterized in that: The sub-machine transmits information to the main machine at a frequency of 1Hz, including its own flight speed, direction, remaining power and search target status.

6. The method for autonomously searching for targets in complex scenes based on a mother-and-child UAV according to claim 3 is characterized in that: When a certain sub-machine is selected as a forwarding node, the mother machine will transmit the flight speed and direction of the sub-machine that needs communication forwarding at a specific frequency. The forwarding node sub-machine will fly in the same direction while ensuring that the distance between itself and the mother machine is less than the wireless network transmission distance threshold; When the distance between the forwarding node slave and the master is greater than the wireless network transmission distance threshold, another forwarding node slave needs to be selected.

7. The method for autonomously searching for targets in complex scenes based on a mother-and-child UAV according to claim 1 is characterized in that: The rule for the submachine to detach is to follow the principle that the overall center of gravity of the submachine and the mother machine is maintained near the central axis.

Images