A Distributed Guidance Method for Multi-UAV Swarms for Cooperative Directional Observation of Moving Targets
By employing a distributed guidance method for multiple UAV swarms, the problem of maintaining observation geometry under the condition of a fixed forward-looking camera on the airframe is solved, enabling effective observation of the UAV swarm when the target is moving and communication is changing, and ensuring the stability of the observation geometry and the accuracy of the observation reference.
Patent Information
- Application Number
- CN202611554999.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-09
- Publication Date
- 2026-11-17
AI Technical Summary
When multiple UAVs conduct collaborative observations with a fixed forward-looking camera, it is difficult to simultaneously maintain effective observation distance, camera pointing, and multi-UAV observation azimuth geometry. Especially when the target moves, the communication topology changes, and the speed and turning ability of the UAVs are limited, existing methods are difficult to effectively switch information sources and transmit target status.
A distributed guidance method using multiple UAV swarms is adopted. By acquiring the navigation status and target stereo matching observation information of each UAV, direct observation nodes are determined, information sources are switched, virtual observation stations are constructed, and restricted longitudinal and normal acceleration commands are generated to ensure the effectiveness of the observation geometry.
Under the condition of a fixed forward-looking camera, a multi-UAV swarm can switch information sources according to the actual observation availability, maintain the non-contact directional observation geometry of the target, and ensure that each UAV generates an observation reference based on the target state according to the time sequence.
Smart Images

Figure CN123114010A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of UAV cooperative control and intelligent observation technology, and relates to a distributed guidance method for multi-UAV swarms for cooperative directional observation of moving targets. Background Technology
[0002] With the development of drone platforms, airborne visual sensors, and wireless communication technologies, the collaborative observation of moving targets by multiple drones can be applied to civilian scenarios such as collaborative patrols of personnel and vehicles at emergency rescue sites, dynamic observation of traffic flow, inspection of mobile equipment in ports and industrial parks, ecological monitoring, and security for large-scale events.
[0003] In moving target observation missions, UAVs typically acquire target images via onboard cameras and adjust their flight speed and heading based on the target's relative position, the position in the target image, or the target's motion state to keep the target within the camera's field of view. Some UAV target tracking solutions use rotatable cameras or gimbals to adjust camera pointing; others adjust the UAV's speed and direction based on target image information to achieve the search, tracking, or localization of moving targets.
[0004] When using a fixed forward-looking camera, the camera's optical axis is constrained by the UAV's heading. To keep the target within the camera's field of view, the UAV needs to adjust its heading; to maintain the observation distance to the target, it needs to adjust its flight speed. When the target is continuously moving or maneuvering, adjusting the heading or speed alone may prevent the target's deviation angle relative to the aircraft's optical axis, the relative distance to the target, and the UAV's speed constraints from being simultaneously satisfied. Therefore, observing moving targets under fixed forward-looking camera conditions requires simultaneous consideration of the camera's field of view, the target's relative distance, the line-of-sight angle, the UAV's speed range, and its turning capability.
[0005] For multi-UAV collaborative missions, some methods use a pre-set leader node or a fixed leader-follower communication relationship to propagate target status information. When the pre-set leader node is unable to continue providing effective observation information due to the target being out of the field of view, target tracking temporarily failing, or the communication graph switching, other UAVs find it difficult to take over the information source based on the current observation availability. If outdated target status estimates are used directly for guidance, it may result in the observation reference positions generated by each UAV being inconsistent with the target's current motion state.
[0006] Furthermore, existing distributed terminal guidance methods typically focus on the proximity and collision geometry between the UAV and the target. Under conditions of two-dimensional constant velocity motion and control using only normal acceleration, they generate cooperative guidance commands based on terminal projection, target maneuverability estimation, and neighbor information. While these methods provide a computational basis for information propagation and terminal state constraints among multiple nodes, their terminal targets are usually not observed at a positive distance, and they are not designed to address the field-of-view limitations of the fixed forward-looking camera, variable velocity conditions, and non-contact observation azimuth maintenance.
[0007] Therefore, there is an urgent need to provide a distributed guidance method for multi-UAV swarms for collaborative directional observation of moving targets. This method can switch information source nodes and propagate target state information based on the observation availability of the fixed forward-looking camera, construct a non-contact virtual observation station based on the predicted target state, and generate longitudinal and normal restricted guidance commands within the range allowed by the fixed forward-looking camera's field of view and the speed capability of the UAVs. Summary of the Invention
[0008] The purpose of this invention is to provide a distributed guidance method for multi-UAV swarms for collaborative directional observation of moving targets, in order to solve the problem that it is difficult to maintain effective observation distance, camera pointing and multi-UAV observation azimuth geometry simultaneously when the target is moving, the direct observation node is switching, the communication topology changes, and the speed and turning ability of the UAVs are limited, under the condition of fixed forward-looking camera.
[0009] To achieve the above objectives, the present invention adopts the following technical solution: A distributed guidance method for multi-UAV swarms for collaborative directional observation of moving targets, applied to a UAV swarm consisting of multiple UAVs equipped with fixed forward-looking binocular vision modules, includes the following steps: S1. In each control cycle, acquire the local navigation status of each UAV, the target stereo matching observation information output by the fixed forward-looking binocular vision module, and the timestamped messages sent by neighboring UAVs. S2. Based on the relative distance to the target, the deviation angle of the target relative to the optical axis of the aircraft, the effective tracking status of the target, and the observation quality, determine the set of direct observation nodes, and determine the information source nodes from the set of direct observation nodes; S3. Under the condition that the information source node can reach other UAVs through the directed communication graph, update the target state estimate of each UAV based on the target observation information and the timestamp message; when the information source node switches, inherit the target state estimate and its timestamp. S4. Estimate and predict the target position based on the target state, and construct a virtual observation station for each UAV based on the predicted target position, the expected observation distance, and the expected observation azimuth of each UAV. S5. Based on the predicted target speed, target relative distance, expected distance change rate, the optical axis deviation range allowed by the fixed forward-looking binocular vision module, and the speed range of the UAV, establish a joint feasibility judgment condition for the speed required to maintain the current observation geometry and the candidate line-of-sight angle correction amount. S6. Based on the predicted position of the UAV zero control system, the predicted target position, the observation projection of the virtual observation station and the UAV body heading calculation terminal, obtain the candidate line-of-sight angle correction amount; implement the optical axis deviation range limit on the candidate line-of-sight angle correction amount, and calculate the corresponding speed required to maintain the current observation geometry; when the joint feasibility judgment condition is met, generate the longitudinal acceleration command and the normal acceleration command according to the speed and the candidate line-of-sight angle correction amount respectively. S7. After applying boundary constraints to the longitudinal acceleration command and the normal acceleration command, output them to the corresponding UAV, and update the information source status, target status estimate, control command and timestamp of the current cycle to the next control cycle.
[0010] Furthermore, the target stereo matching observation information includes the timestamps of the left and right camera images, the effective markers for target stereo matching and tracking, and the target's relative three-dimensional coordinates. The fixed forward-looking binocular vision module obtains the target's relative three-dimensional coordinates through stereo matching and triangulation based on the left and right camera images, intrinsic parameters, binocular baseline, and fixed extrinsic parameters. The relative three-dimensional coordinates are then converted to the local navigation coordinate system, and the local two-dimensional position components of the ground are extracted. Combined with the local navigation status, the target position observation is obtained, and the target velocity observation is obtained based on the adjacent effective position observations and their timestamps.
[0011] Furthermore, the direct observation nodes are jointly determined by the working distance range, the field of view range, the effective target tracking status, and the observation quality with entry and exit thresholds. Determining the set of direct observation nodes includes: identifying UAVs whose relative target distance is within a preset working distance range, whose deviation angle of the target relative to the aircraft's optical axis is within the field of view range of the fixed forward-looking binocular vision module, and whose target tracking is effective, as geometrically observable nodes; calculating the observation quality based on the degree to which the relative target distance deviates from the desired observation distance, the deviation angle, and the effective target tracking status; for geometrically observable nodes not in a direct observation state, they are identified as direct observation nodes when the observation quality reaches the entry threshold; for nodes already in a direct observation state, the direct observation node status is cancelled when the observation quality falls below the exit threshold or target tracking fails, wherein the entry threshold is greater than the exit threshold.
[0012] Furthermore, determining the information source node and inheriting the target state estimate and its timestamp during information source node switching includes: if the information source node from the previous control cycle still meets the direct observation conditions, it remains unchanged; otherwise, it is taken over by a node in the current set of direct observation nodes, inheriting the target position estimate, target velocity estimate, target maneuverability estimate, timestamp, and source marker. A directed communication graph is constructed based on the current information source node, and it is determined whether the current information source node can reach other UAVs through the directed communication graph.
[0013] Furthermore, the information source node corrects the target state estimate using current observations, while non-information source nodes perform state extrapolation and propagate target maneuverability estimates based on the latest and non-outdated neighbor messages. Specifically, the information source node corrects the target position and velocity estimates based on current target observation information; non-information source nodes select timestamp-updated target state estimates from neighbor messages and perform time extrapolation based on the target state and target maneuverability estimates; for information source nodes continuously in direct observation mode, local target maneuverability estimates are obtained based on target velocity observations from adjacent control cycles; and the target maneuverability estimates of each UAV are updated based on the local target maneuverability estimates, the information source node states, and the target maneuverability estimates of neighboring nodes in the directed communication graph.
[0014] Furthermore, the terminal observation projection, combined with the predicted target position, the virtual observation station, the UAV zero-control predicted position, and the neighboring UAV predicted position, yields a candidate line-of-sight angle correction amount for correcting the observation azimuth; the candidate line-of-sight angle correction amount is used to generate restricted guidance commands only when the fixed forward-looking binocular vision module optical axis deviation constraint and the UAV speed range constraint are satisfied. The calculation of the terminal observation projection and obtaining the candidate line-of-sight angle correction includes: predicting the UAV's zero-control position based on the UAV's current speed, heading, and terminal prediction time; predicting the target position based on target position estimation, target speed estimation, and target maneuverability estimation; predicting the neighboring aircraft's position based on the neighboring aircraft's position, heading, speed, and normal acceleration commands received in the previous control cycle, and determining the azimuth interval error based on the azimuth difference between the UAV and the neighboring aircraft relative to the predicted target position; correcting the corresponding UAV's desired observation azimuth based on the azimuth interval error and constructing a virtual observation station; calculating the terminal observation projection based on the lateral deviation of the UAV's zero-control position relative to the virtual observation station and the pointing deviation of the UAV's heading relative to the reference heading, obtaining the candidate line-of-sight angle correction, and determining whether the candidate line-of-sight angle correction is jointly feasible based on the speed required to maintain the desired observation distance corresponding to the candidate line-of-sight angle correction.
[0015] Furthermore, longitudinal acceleration commands are saturated according to the upper limit of longitudinal acceleration, and normal acceleration commands are saturated according to the upper limit of normal acceleration. When the set of directly observed nodes is empty, the information source node is unreachable, the terminal prediction duration or terminal observation projection does not meet the preset conditions, or the candidate line-of-sight angle correction amount fails the joint feasibility judgment, the corresponding terminal projection control result is stopped, the corresponding control command is set to zero and written to the theoretical premise failure flag. The control state of the next control cycle is updated according to the theoretical premise failure flag, the control command saturation state, the information source state, and the message timestamp.
[0016] The present invention also provides an electronic device, comprising: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the method as described in any of the preceding methods.
[0017] A computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the method described in any of the preceding claims.
[0018] Compared with collaborative observation methods that use fixed information sources, this invention switches information sources based on actual observation availability and propagates target state information through target state inheritance, timestamps, and directed communication graphs, enabling each UAV to generate observation references based on target states with source and timing information.
[0019] This invention transforms the predicted target state into a virtual observation station that maintains a positive distance, and links the terminal observation projection, the optical axis deviation constraint of the fixed forward-looking binocular vision module, and the UAV's speed change capability; guidance commands are generated only for candidate line-of-sight angle corrections that pass joint feasibility judgment, so as to maintain the non-contact directional observation geometry of multiple UAVs on moving targets. Attached Figure Description
[0020] Figure 1 Flowchart of a distributed guidance method for multi-UAV collaborative directional observation; Figure 2 This is a geometric schematic diagram of non-contact observation under the condition of a fixed forward-looking camera; Figure 3 This is a schematic diagram of dynamic information source switching and directed target state propagation. Detailed Implementation
[0021] The following detailed description of the distributed guidance method for multi-UAV swarms for cooperative directional observation of moving targets according to the present invention is provided in conjunction with specific embodiments. These embodiments are used to explain the technical solution of the present invention and are not intended to limit the scope of protection of the present invention.
[0022] I. System Composition and Status Acquisition This implementation method uses a swarm of multiple UAVs to conduct non-contact collaborative directional observation of moving targets. The specific process is as follows: Figure 1 As shown. Each UAV includes a fixed forward-looking binocular vision module, a navigation unit, an onboard computing unit, a communication unit, and a flight control unit. The navigation unit outputs the UAV's position, heading, and speed; the communication unit sends status messages with timestamps; the onboard computing unit completes direct observation node identification, information source switching, target state estimation and prediction, virtual observation station construction, terminal observation projection, joint feasibility assessment of candidate line-of-sight angles, and guidance command generation; the flight control unit executes longitudinal acceleration commands and normal acceleration commands.
[0023] Establish a local two-dimensional rectangular coordinate system on the ground . No. A drone at all times The position, heading angle and speed are respectively , and Its control input is longitudinal acceleration. and normal acceleration The drone employs a two-dimensional variable-speed motion model:
[0024]
[0025]
[0026] in, The unit is m. The unit is rad. The unit is m / s. and The unit is m / s .
[0027] The fixed forward-looking binocular vision module synchronously outputs images from the left and right cameras, image timestamps, and valid markers for target stereo matching and tracking. For valid targets, based on the intrinsic parameters of the left and right cameras, the binocular baseline, and the fixed extrinsic parameters of the binocular vision module relative to the body coordinate system, the target's relative three-dimensional coordinates are obtained through stereo matching and triangulation. These coordinates are then transformed to the local navigation coordinate system, and the local two-dimensional position components on the ground are extracted. Combined with the local position, the target position observation is obtained. Based on the position observations of two adjacent valid targets and their image timestamps, the target velocity observation is obtained. A schematic diagram of non-contact observation geometry under fixed forward-looking camera conditions is shown below. Figure 2 As shown.
[0028] In this invention,
[0029] Indicates the angle Mapped to ;and
[0030] Indicates the variable Limited to the range Inside.
[0031] II. Direct Observation Node Identification and Information Source Switching For the The drone calculates the relative observation position, observation distance, observation line-of-sight angle, and optical axis deviation angle based on observations of the current target location, as follows:
[0032]
[0033]
[0034]
[0035] when Within the preset distance working range Not greater than the half field of view of the fixed forward-looking binocular vision module When the target stereo matching and tracking effective markers are valid, the UAV is identified as a geometrically observable node.
[0036] Calculate the overall observation quality for geometrically observable nodes:
[0037]
[0038]
[0039] in, For the desired observation distance, The distance is the scale of mass decay. The tracking quality is determined by the effectiveness of binocular matching, triangulation, and cross-frame tracking. , and These are the corresponding quality weights.
[0040] Set entry threshold and exit threshold ,and Geometrically observable nodes that were not in a directly observable state in the previous cycle are now... Nodes are added to the set of directly observed nodes at a certain time; nodes already in the directly observed state are added to the set of directly observed nodes at a certain time. Exit the set if target tracking fails.
[0041] If the current information source node still belongs to the set of directly observed nodes, it remains unchanged; otherwise, the node with the smallest node number is selected from the current set of directly observed nodes as the new information source node, such as... Figure 3 As shown. After the switch, the new information source node inherits the target position estimate, target velocity estimate, target maneuverability estimate, timestamp, and source marker from the previous cycle, and corrects the inherited state with the current target observations.
[0042] III. Target State Propagation and Prediction No. UAV maintenance target location estimation Target speed estimation And target mobility estimation Subsequent guidance calculations use the above estimates, rather than directly using the unavailable true state of the target.
[0043] The current information source node serves as the sole root node in the directed communication graph. Guidance commands are generated using the target state estimate for the current period only if the information source node can reach other UAVs via the directed communication graph.
[0044] For information source nodes that have effective target velocity observations for two consecutive control cycles, the local target maneuverability estimate is calculated according to the following formula:
[0045] Each UAV updates its own target maneuver estimate using the maneuverability estimates of its neighbors and the local maneuverability estimates of the information source nodes:
[0046] in, Indicates the first A drone towards the first A drone was used to send a message; For consistency gain, For pinning gain; when the first When the UAV is used as an information source node with continuous and effective velocity observation, ,otherwise .
[0047] The information source node uses the current target position observation and the target velocity observation as... and The correction value. Non-source nodes select the latest and non-outdated target state estimate from neighboring messages, and sort them by message age. Extrapolate:
[0048]
[0049] IV. Virtual Observation Station and Terminal Observation Projection Based on the local target state estimation, the estimated relative position, estimated range, estimated line-of-sight angle, and unit line-of-sight vector for guidance calculations are constructed as follows:
[0050]
[0051]
[0052]
[0053] No. The velocity vector of the drone is:
[0054] Terminal prediction duration is:
[0055] in, For the minimum closing velocity, and These are the lower and upper limits of the terminal prediction duration, respectively.
[0056] The predicted target location and the predicted location of the UAV zero-controller are as follows:
[0057]
[0058] Based on the neighboring aircraft's position, heading, speed, and normal acceleration commands received in the previous control cycle, the predicted position of the neighboring aircraft is obtained using a constant turning model. .make and They are respectively and Compared to If the azimuth angle is given, then the azimuth interval error and azimuth coordination amount are as follows:
[0059]
[0060] Accordingly, the following amendments are made. Determine the desired observation azimuth of the UAV and construct a virtual observation station and reference heading:
[0061]
[0062]
[0063] in, To preset the desired observation azimuth, For the desired azimuth interval, For azimuth coordination gain.
[0064] Construct the lateral unit vector, the terminal observation projection vector, the projection control matrix, and the projection control vector:
[0065]
[0066]
[0067]
[0068]
[0069] The projection control value is compressed into the viewing angle projection correction value:
[0070] in, This is the upper limit of the projection correction angle. This represents the upper limit of normal acceleration. Reference line-of-sight angle deviation. By the The preset observation azimuth and initial observation geometry of the UAV are determined. The field of view margin is set. Then, the maximum permissible line-of-sight angle deviation is:
[0071] The candidate line-of-sight angle deviation is:
[0072] V. Restricted Command Generation and Periodic Update The predicted radial velocity and expected distance change rate of the target along the current estimated line-of-sight direction are as follows:
[0073]
[0074] The velocity required to maintain the current observation geometry is:
[0075] when For finite values, satisfying ,and At this time, the candidate line-of-sight angle deviation is determined through joint feasibility assessment. The speed reference value, heading reference value, and unsaturated control command are as follows:
[0076]
[0077]
[0078]
[0079] Saturation processing of control commands:
[0080]
[0081] Each control cycle determines whether the set of directly observed nodes is empty, whether the information source nodes are reachable from other UAVs, and whether the terminal prediction duration meets the lower limit. Whether the terminal observation projection is within the preset range, and whether the candidate line-of-sight angle deviation has passed the joint feasibility judgment are all checked. If any condition is not met, the corresponding longitudinal acceleration command and normal acceleration command are set to zero, and a theoretical premise failure flag is written; otherwise, the saturated control command is output to the flight control unit. Then, the information source status, target status estimation, control command, and timestamp are updated, and the next control cycle begins.
[0082] Example In a set of three-machine coordinated orientation observation calculations, the control period Set to 0.01 s, and the calculation time domain to 30 s; the initial target position is set to... m, the target's initial velocity is set to The initial positions of the three UAVs were all located on a circle with a radius of 80 m centered on the initial position of the target. The initial observation azimuths were set to 140°, 180° and -140° respectively, and the initial velocities were all set to 8 m / s.
[0083] Half field of view of fixed forward-looking binocular vision module Set to 40°; effective observation distance range set to 15m to 150m; desired observation distance. Set to 60 m; distance-mass decay scale Set to 40 m. Distance-mass weighting. Optical axis deviation mass weight and tracking quality weights Set to 0.20, 0.20, and 0.60 respectively; observation quality enters the threshold. and exit threshold Set them to 0.75 and 0.60 respectively.
[0084] The drone's speed range is set from 3 m / s to 25 m / s, with a maximum longitudinal acceleration. Set to 3 m / s 2 Upper limit of normal acceleration Set to 10 m / s 2 Consistency gain Pinning gain Azimuth Coordination Gain Range gain and heading gain Set to 30, 3, 0.20, 0.15, and 8 respectively; upper limit of distance change rate. Set to 4 m / s.
[0085] Minimum closing speed The speed is set to 2 m / s, the terminal prediction duration range is set to 0.5 s to 8 s, and the upper limit of the projection correction angle is set to [missing information]. Set to 5°, field margin The reference line-of-sight angle deviation of the three drones is set to 5°. Set to 20°, 0° and -20° respectively.
[0086] In the information source switching scenario, when the target tracking effective marker of the original information source node fails while another direct observation node remains effective, the effective direct observation node takes over the information source and inherits the target state estimate, target maneuverability estimate, timestamp, and source marker from the previous cycle.
[0087] This embodiment illustrates the construction of the observation reference, the generation of restricted guidance commands, and the information source switching process in a scenario where three UAVs collaboratively observe a moving target. Specifically, with a desired observation distance of 60 m, a virtual observation station is constructed based on the predicted target position and the desired observation azimuth of each UAV. Corresponding guidance commands are generated through distance and heading adjustments, thereby transforming the requirement for maintaining a positive distance in non-contact directional observation into a motion control reference for each UAV.
[0088] With a fixed half-field-of-view angle of 40° and a field-of-view margin of 5° for the forward-looking binocular vision module, the maximum permissible line-of-sight angle deviation is 35°. After imposing the above-mentioned range limitation on the candidate line-of-sight angle deviations, the corresponding velocity required to maintain the current observation geometry is further calculated, and it is determined whether this velocity is a finite value and falls within the range of 3 m / s to 25 m / s. When the conditions of direct observation node, communication reachability, and terminal observation projection are all met, guidance commands are generated only for candidate results that pass the joint feasibility judgment, and are respectively set at 3 m / s2 and 10 m / s 2 The upper limit is used to saturate the longitudinal acceleration command and the normal acceleration command. The above process illustrates how the present invention incorporates camera optical axis deviation limits, UAV velocity range, and acceleration boundaries into the guidance command generation process.
[0089] In the information source switching scenario, after the original information source node exits the direct observation node set due to target tracking failure, the effective node with the smallest node number in the current set takes over the information source. The new information source node inherits the target state estimate, its timestamp, and source marker from the previous cycle and performs corrections using the current effective observations. Provided the new information source node can reach other UAVs via the directed communication graph, it continues to propagate the target state to the cluster. This process demonstrates that the present invention can adjust the information source based on actual observation availability and, through state inheritance and current observation correction, connect the target state updates before and after the switch, providing target state information with source and time basis for subsequent virtual observation station construction and collaborative guidance.
[0090] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0091] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0092] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.
[0093] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0094] The embodiments described above are merely some preferred embodiments of the present invention, and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained by equivalent substitution or equivalent transformation fall within the protection scope of the present invention.
Claims
1. A distributed guidance method for multi-UAV swarms for collaborative directional observation of moving targets, characterized in that, Includes the following steps: S1. In each control cycle, acquire the local navigation status of each UAV, the target stereo matching observation information output by the fixed forward-looking binocular vision module of the UAV, and the timestamped messages sent by neighboring UAVs. S2. Based on the relative distance to the target, the deviation angle of the target relative to the optical axis of the aircraft, the effective tracking status of the target, and the observation quality, determine the set of direct observation nodes, and determine the information source nodes from the set of direct observation nodes; S3. Under the condition that the information source node can reach other UAVs through the directed communication graph, update the target state estimate of each UAV based on the target observation information and the timestamp message; when the information source node switches, inherit the target state estimate and its timestamp. S4. Estimate and predict the target position based on the target state, and construct a virtual observation station for each UAV based on the predicted target position, the expected observation distance, and the expected observation azimuth of each UAV. S5. Based on the predicted target speed, target relative distance, expected distance change rate, the optical axis deviation range allowed by the fixed forward-looking binocular vision module, and the speed range of the UAV, establish a joint feasibility judgment condition for the speed required to maintain the current observation geometry and the candidate line-of-sight angle correction amount. S6. Based on the predicted position of the UAV zero control system, the predicted target position, the observation projection of the virtual observation station and the UAV body heading calculation terminal, obtain the candidate line-of-sight angle correction amount; implement the optical axis deviation range limit on the candidate line-of-sight angle correction amount, and calculate the corresponding speed required to maintain the current observation geometry; when the joint feasibility judgment condition is met, generate the longitudinal acceleration command and the normal acceleration command according to the speed and the candidate line-of-sight angle correction amount respectively. S7. After applying boundary constraints to the longitudinal acceleration command and the normal acceleration command, output them to the corresponding UAV, and update the information source status, target status estimate, control command and timestamp of the current cycle to the next control cycle.
2. The method according to claim 1, characterized in that, Determining the set of directly observable nodes includes: identifying UAVs whose relative target distance is within a preset working distance range, whose deviation angle of the target relative to the optical axis of the aircraft is within the field of view of the fixed forward-looking binocular vision module, and whose target tracking is effective as geometrically observable nodes; calculating the observation quality based on the degree to which the relative target distance deviates from the desired observation distance, the deviation angle, and the effective target tracking status; for geometrically observable nodes not in a directly observable state, determining them as directly observable nodes when the observation quality reaches an entry threshold; for nodes already in a directly observable state, canceling the directly observable node status when the observation quality is lower than an exit threshold or when target tracking fails, wherein the entry threshold is greater than the exit threshold.
3. The method according to claim 1, characterized in that, The target stereo matching observation information includes the timestamps of the left and right camera images, valid markers for target stereo matching and tracking, and the target's relative three-dimensional coordinates. Based on the intrinsic parameters of the fixed forward-looking binocular vision module, the binocular baseline, and the fixed extrinsic parameters relative to the body coordinate system, stereo matching and triangulation are performed on the target in the left and right camera images to obtain the target's relative three-dimensional coordinates in the camera coordinate system. The relative three-dimensional coordinates are then transformed to the local navigation coordinate system, and their position components in the local two-dimensional coordinate system on the ground are extracted. Combined with the local navigation status, the target position observation is obtained. Based on the position observations of two adjacent valid targets and their corresponding timestamps, the target velocity observation is obtained.
4. The method according to claim 1, characterized in that, The process of identifying the information source node and inheriting the target state estimate and its timestamp during information source node switching includes: if the information source node of the previous control cycle still belongs to the current direct observation node set, retaining the information source node of the previous control cycle as the current information source node; if the information source node of the previous control cycle does not belong to the current direct observation node set, selecting the node with the smallest node number from the current direct observation node set as the new information source node; inheriting the target position estimate, target velocity estimate, target maneuverability estimate, timestamp, and source marker of the previous control cycle to the new information source node; constructing a directed communication graph based on the current information source node, and determining whether the current information source node can reach other UAVs through the directed communication graph.
5. The method according to claim 1, characterized in that, Updating the target state estimates for each UAV includes: the information source node correcting the target position and velocity estimates based on the current target observation information; the non-information source node selecting timestamp-updated target state estimates from neighboring UAV messages and extrapolating the time based on the target state estimates and target maneuverability estimates; for information source nodes continuously in direct observation state, obtaining local target maneuverability estimates based on target velocity observations from adjacent control cycles; and updating the target maneuverability estimates for each UAV based on the local target maneuverability estimates, the information source node states, and the target maneuverability estimates of neighboring nodes in the directed communication graph.
6. The method according to claim 1, characterized in that, The joint feasibility determination includes: determining the expected distance change rate based on the distance deviation between the UAV and the predicted target, and imposing boundary constraints on the expected distance change rate; determining the speed required to maintain the expected observation distance based on the velocity component of the predicted target along the current line of sight, the expected distance change rate, and the candidate line of sight angle correction amount after the field of view constraint of the fixed forward-looking binocular vision module; when the speed is a finite value and is between the upper and lower limits of the UAV speed, the candidate line of sight angle correction amount is determined to be jointly feasible, otherwise it is determined to be jointly infeasible.
7. The method according to claim 1, characterized in that, The calculation of the terminal observation projection and obtaining the candidate line-of-sight angle correction includes: predicting the UAV's zero-control position based on the UAV's current speed, heading, and terminal prediction time; predicting the target position based on target position estimation, target speed estimation, and target maneuverability estimation; predicting the neighboring aircraft's position based on the neighboring aircraft's position, heading, speed, and normal acceleration commands received in the previous control cycle, and determining the azimuth interval error based on the azimuth difference between the UAV and the neighboring aircraft relative to the predicted target position; correcting the corresponding UAV's desired observation azimuth based on the azimuth interval error and constructing a virtual observation station; calculating the terminal observation projection based on the lateral deviation of the UAV's zero-control position relative to the virtual observation station and the pointing deviation of the UAV's heading relative to the reference heading, obtaining the candidate line-of-sight angle correction, and determining whether the candidate line-of-sight angle correction is jointly feasible based on the speed required to maintain the desired observation distance corresponding to the candidate line-of-sight angle correction.
8. The method according to claim 1, characterized in that, The longitudinal acceleration command is saturated according to the upper limit of the longitudinal acceleration, and the normal acceleration command is saturated according to the upper limit of the normal acceleration. When the set of direct observation nodes is empty, the information source node cannot reach other UAVs through the directed communication graph, the terminal prediction time or the terminal observation projection does not meet the preset conditions, or the candidate line-of-sight angle correction amount fails the joint feasibility judgment, the corresponding terminal projection control result is stopped, the corresponding control command is set to zero and written into the theoretical premise failure flag. The control state for the next control cycle is updated based on the theoretical premise of failure flag, control command saturation state, information source state, and message timestamp.
9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-8.
10. A computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the method of any one of claims 1-8.
