Multi-agent cooperative target pointing control method based on local orientation

Abstract

The invention discloses a multi-agent cooperative target pointing control method based on a local orientation. The method comprises three modules, namely a distributed attitude estimation module, a cooperative target estimation module and a target pointing control module. Information fusion is performed through target local azimuth information, relative attitude information between neighbors and local interaction information measured by a leader agent, so that all agents can autonomously complete attitude estimation, target positioning and target pointing. According to the method, the problem of cooperative pointing control without a global coordinate system, unknown target position and arbitrary deployment in a three-dimensional space is solved. The designed distributed attitude estimator does not need global coordinate information and orthogonalization processing, the cooperative target estimator does not require all agents to be equipped with orientation sensors and is not limited by a specific deployment structure or a two-dimensional space, and a target pointing controller does not need complex geometric analysis or a switching algorithm to avoid singular points, so that the method is simple and convenient. Therefore, the method has better adaptability and expandability, and is suitable for various practical application scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of cooperative control of multi-agent systems, and particularly to a multi-agent cooperative target pointing control method based on local orientation. Background Art

[0002] Multi-agent cooperative pointing control refers to the process in which agents cooperatively adjust their own postures through local information interaction, so that the pointing directions of all agents finally converge to the same target direction. This technology has important application values in multiple fields. For example, in forest exploration and disaster rescue, a single unmanned aerial vehicle (UAV) may have difficulty obtaining target information due to path limitations or occlusion, while a group of UAVs can cooperatively point to the target from different angles to improve detection and tracking capabilities. In 3D reconstruction tasks, multiple cameras are synchronously aligned with the target, and two-dimensional images are obtained through multi-angle observations to achieve high-precision 3D reconstruction. In addition, in target tracking in crowded areas, cooperative pointing control can help multiple UAVs capture the target omnidirectionally, making it difficult for the target to escape monitoring and improving tracking efficiency and accuracy.

[0003] Currently, multi-agent cooperative pointing control methods mainly rely on deploying agents along the same baseline and designing feedback strategies using geometric relationships to achieve pointing cooperation. However, such methods usually require complex geometric analysis and are difficult to generalize in complex environments. In addition, in environments such as indoors, urban canyons, or underwater, a global coordinate system may not be available, while most existing methods assume that the global coordinate system is known, which is not always true in practice. The assumption of relying on a global coordinate system or an on-board compass is often difficult to meet in a changing environment, thus limiting the adaptability and application scope of the system.

[0004] Therefore, under the condition of no global coordinate system, the multi-agent cooperative target pointing control method based on local orientation can not only improve the autonomy and applicability of the system, but also bring improvements to the existing cooperative target pointing control technology, while enriching the theoretical system of multi-agent cooperative control, and has important research value and application prospects. Summary of the Invention

[0005] In order to solve the above-mentioned defects or problems existing in the prior art, the purpose of the present invention is to provide a multi-agent cooperative pointing control method based on local orientation, which includes three major modules: distributed attitude estimation, cooperative target estimation, and target pointing control. Through the local orientation of the target measured by some agents, the relative attitude information between neighbors, and the local interaction data information, information fusion is carried out, so that all agents can independently complete attitude estimation, target positioning, and target pointing, and has high efficiency, autonomy, and wide applicability.

[0006] The present invention is realized through the following technical solutions: A multi-agent collaborative target pointing control method based on local orientation, comprising the following steps: Step 1: The multi-agent system includes agents randomly deployed in three-dimensional space. Each agent is equipped with a local coordinate system , and cannot obtain the global coordinate system information. Some agents (called leader agents) are equipped with visual sensors to measure the local orientation relative to the target , and at least two leader agents are non-collinear with the target. The remaining agents (called follower agents) cannot measure the target local orientation; Step 2: Define the attitude of agent relative to the global coordinate system as , where is a special orthogonal group and the attitude is unknown; Define the relative attitude of agent relative to neighbor agent as , and the relative attitude can be measured by the equipped visual sensor; Define a common virtual coordinate system , and the origin and direction of the common virtual coordinate system are unknown and arbitrary to the agents; Step 3: Define the positions of agent and the target in the common virtual coordinate system as and respectively, where the superscript represents the transpose symbol; Step 4: Define the pointing of agent as a unit direction vector and the kinematic model as , where in the formula is the orthogonal projection matrix of the unit vector , is the three-dimensional identity matrix; is the control input to be designed, used to project the control signal onto the vertical plane to adjust the pointing direction; Step 5: Define the set of all agents in the multi-agent system as , the subset of leader agents as , and the subset of follower agents as , where is the number of leader agents and satisfies to ensure the uniqueness of target positioning; Step 6: Define the communication network topology of the multi-agent system as a connected undirected graph ; Step 7: The multi-agent cooperative target pointing control method based on local orientation requires the interaction information between the agent and its neighbor agents to achieve cooperative target pointing control. Using the nearest neighbor rule, define the set of neighbor agents of the agent as ; Step 8: Limited by the lack of a global coordinate system, the agent cannot directly obtain its own attitude . The agent needs to use a distributed attitude estimator to jointly determine a common virtual coordinate system and estimate its own attitude in ; ; The design expression of the distributed attitude estimator is: (1) where is the estimate of the attitude of agent in the common virtual coordinate system and ; ; is the relative attitude of agent relative to agent and ; is the proportionality coefficient; is the adjacency weight matrix of the network topology graph ; Furthermore, in step 8, there is also a step of setting the initial estimate value parameters of the distributed attitude estimator. Set , and the initial attitude of the agent needs to ensure that the rotation angle range around the rotation axis is within to ensure the convergence of the distributed attitude estimator; Step 9: Each agent in the multi-agent system cannot directly obtain the complete target position information. The multi-agent system measures the target local orientation obtained by the leader agent . The self-position information of each agent and the neighbor target position estimation information obtained through interaction with neighbor agents are used to design the following cooperative target estimator for each agent : (2) where is the target position of agent in and The estimated value; if the agent is the leader agent, then otherwise ; is the orthogonal projection matrix of the unit vector , that is ; Furthermore, the step 9 also includes the step of setting the initial target estimation value parameter of the cooperative target estimator, and the initial estimation value of the agent is not equal to the initial value estimation of the neighbor agent , that is ; Step 10: The agent obtains the target position information according to the distributed attitude estimator and the cooperative target estimator, and combines its own position information to design the following multi-agent pointing control input: (3) where the orthogonal projection matrix projects the displacement difference between the target estimated position and the agent's own position onto the plane perpendicular to the pointing direction vector , thereby driving the pointing to rotate so that the pointing directions of all agents converge to the target direction.

[0007] As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following beneficial effects: The multi-agent cooperative target pointing control method based on local orientation provided by the present invention solves the problems of no global coordinate system, arbitrary deployment of agents in three-dimensional space, and unknown target position in multi-agent cooperative target pointing control. The designed distributed attitude estimator does not require agents to obtain global coordinate system information, nor does it need to perform the Gram-Schmidt orthogonalization process and has no steady-state ambiguity, thus simplifying the calculation; at the same time, the cooperative target estimator does not require all agents to be equipped with sensors that can measure the local orientation of the target, nor does it depend on a specific agent deployment structure or be limited to two-dimensional space; in addition, the target pointing controller does not require complex geometric analysis, nor does it need to adopt a cumbersome switching algorithm to avoid singularities, thus significantly improving the adaptability and scalability of the entire system and making it more suitable for actual application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0009] Figure 1 It is a step flow chart of a multi-agent cooperative target pointing control method based on local orientation of the present invention; Figure 2 Under the action of the distributed attitude estimator (1), the attitude estimation error Convergence process; Figure 3 Under the action of the cooperative target estimator (2), the target position estimation error Convergence process; Figure 4 It is a trajectory graph of the cooperative target pointing of the multi-agent system under the simultaneous action of the cooperative target pointing control strategy, that is, formulas (1), (2) and (3). Specific implementation manner

[0010] The following further illustrates the present invention in combination with specific embodiments.

[0011] Refer to Figure 1 , the present invention provides a multi-agent cooperative target pointing control method based on local orientation. This method includes three major modules: distributed attitude estimation, cooperative target estimation, and target pointing control. Through the local orientation information of the target measured by the leader agent, the relative attitude information between neighbors, and the local interaction information, information fusion is carried out so that all agents can independently complete attitude estimation, target positioning, and target pointing. Specifically, it includes the following steps: Step 1: The multi-agent system contains Agents are arbitrarily deployed in three-dimensional space. Each agent Is equipped with a local coordinate system , and cannot obtain the global coordinate system Information. Some agents (called leader agents) are equipped with vision sensors to measure the local orientation Relative to the target, and at least two leader agents are not collinear with the target. The remaining agents (called follower agents) cannot measure the local orientation of the target; Step 2: Define the attitude of agent Relative to the global coordinate system as , where Is a special orthogonal group and the attitude is unknown; Define the relative attitude of agent Relative to the neighbor agent As , the relative pose can be measured by equipping with a vision sensor; define a common virtual coordinate system , the origin and direction of the common virtual coordinate system are unknown and arbitrary to the agent; Step 3: Define the agent and the positions of the target in the common virtual coordinate system are respectively and , where the superscript represents the transpose symbol; Step 4: Define the pointing direction of the agent as a unit direction vector and the kinematic model is , where is the orthogonal projection matrix of the unit vector , is the three-dimensional identity matrix; is the control input to be designed, used to project the control signal onto the vertical plane to adjust the pointing direction; Step 5: Define the set of all agents in the multi-agent system as , the subset of leader agents is , and the subset of follower agents is , where is the number of leader agents, and satisfies to ensure the uniqueness of target localization; Step 6: Define the communication network topology of the multi-agent system as a connected undirected graph ; Step 7: The multi-agent cooperative target pointing control method based on local orientation requires the interaction information between the agent and its neighbor agents to achieve cooperative target pointing control. Using the nearest neighbor rule, define the neighbor agent set of the agent as ; Step 8: Limited by the lack of a global coordinate system, the agent cannot directly obtain its own pose . The agent needs to use a distributed pose estimator to jointly determine a common virtual coordinate system and estimate its own pose in ; The design expression of the distributed pose estimator is: (1) where, is the pose of the agent itself in the common virtual coordinate system under estimate and ; is the agent relative to the agent relative pose and ; is the proportionality coefficient; is the network topology graph adjacency weight matrix; Continue to set the initial estimated value parameters of the distributed pose estimator. Set , and the initial pose of the agent needs to ensure that the rotation angle range around the rotation axis is within to ensure the convergence of the distributed pose estimator; Furthermore, construct a Lyapunov function to confirm the stability of the distributed pose estimator, which specifically includes: constructing a Lyapunov function to confirm that the distributed pose estimator (1) stabilizes the system, and the multi-agent reaches an agreement to determine the common virtual coordinate system , the agent the absolute error of the pose estimation satisfies , that is, it is equivalent to the absolute error value of the pose estimation of the agents in the multi-agent system all converge to the same value to complete its own pose estimation task; Step 9: Each agent in the multi-agent system cannot directly obtain complete target position information. The multi-agent system measures the target local azimuth obtained by the leader agent , and the own position information of each agent and the neighbor target position estimation information obtained through interaction with neighbor agents are used to design the following cooperative target estimator for each agent : (2) where is the estimated value of the target position by the agent at ; if the agent is the leader agent, then , otherwise ; is the orthogonal projection matrix of the unit vector , that is ; Continue to set the initial target estimated value parameters of the cooperative target estimator. The initial estimated value of the agent is the same as that of the neighbor agent The initial value estimates are not equal, i.e., ; Furthermore, construct a Lyapunov function to confirm the stability of the cooperative target estimator, which specifically includes: constructing a Lyapunov function, confirming that the cooperative target estimator (2) stabilizes the system, and confirming that the target estimate error value converges to zero, and it is concluded that under the action of this cooperative target estimator, the target estimated position can converge to the true target position , and the target position estimation task is completed; Step 10: The agent obtains the target position information according to the distributed attitude estimator and the cooperative target estimator , and combines its own position information (3) where the orthogonal projection matrix projects the displacement difference between the target estimated position and the agent's own position onto a plane perpendicular to the pointing direction vector , thereby driving the pointing to rotate so that the pointing directions of all agents converge to the target direction; In the above formulas (1), (2) and (3), the estimators include formulas (1) and (2), which are used to output the required self-attitude estimation value and the target position estimation value of the agent; the distributed attitude estimator (1) mainly solves the problem of the absence of a global coordinate system, and the multi-agent system jointly determines a common virtual coordinate system , and realizes the self-attitude estimation under the coordinate system , thereby converting the information to be fused into the coordinate system to achieve effective fusion; the cooperative target estimator (2) mainly solves the problem of unknown target position information in the multi-agent system. According to the triangulation principle, the leader agent can uniquely determine the target position based on the measured local azimuth of the target; the formula for the target pointing controller is (3), which is used to adjust the pointing direction of the agent itself until the agents point to the same target.

[0012] The above controller (3) can achieve the target pointing task, and the proof process is as follows: First, define the Lyapunov candidate function: , where is the desired unit vector pointing to the target; Taking the derivative of yields: (4) In the formula, Given any constant satisfying , since , there exists a finite time such that . Therefore, according to formula (4), we can obtain: (5) It can be seen from formula (5) that the controller (3) is stable and satisfies , indicating that the orientation of the agent finally converges to the expected value, that is, the multi-agent system realizes the cooperative orientation control of the target. Q.E.D.

[0013] In addition, this embodiment provides a simulation experiment to verify the effectiveness of the above control method.

[0014] Specifically, a multi-agent system containing 10 agents is adopted in this simulation experiment, and they are arbitrarily deployed in the three-dimensional space for simulation verification. Among them, agents numbered 1 to 3 are leader agents, and the remaining agents are follower agents. The initial positions of each agent are respectively: , , , , , , , , , , the above positions in the coordinate system are only used to initialize the control strategy and can be any different values; the target local orientations measured by the leader agents are respectively: , , ; the neighbor sets of the agents are respectively: , , , , , , , , , ; In addition, select the parameter ; the initial attitude estimate value is selected according to step 9, and the initial target estimate value is: .

[0015] Under the action of the distributed attitude estimator (1) and the cooperative target estimator (2), the convergence processes of the attitude estimation error and the target estimation error are respectively as followsFigure 2 As shown in Figure 3 . The results show that it can converge to a common value, which means that all agents reach a consensus to determine a common virtual coordinate system , and accurately estimate the attitude under this coordinate; at the same time, the position estimation error gradually converges to zero, that is, the multi-agent system can estimate the true position of the target through the cooperative target estimator (2); under the action of the entire control strategies (1), (2) and (3), the trajectory diagram of the multi-agent system cooperating to point to the target is as shown in Figure 4 . It can be observed that before 15.0 s, all agents point to the target. The results show that the multi-agent system can cooperate to achieve the pointing control of the target.

[0016] The above is a specific description of the preferred implementation of the present invention. However, the specific implementation of the present invention is only used to help understand the method and its core idea of the present invention, and cannot be used to limit the scope of the right protection. The scope of protection required by the present invention is not limited to the above specific implementation manners. Moreover, for those skilled in the art, the present invention can have various deformations and changes. Any modification, improvement and equivalent replacement made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-agent collaborative target pointing control method based on local orientation, characterized by: The following steps are involved: Step 1: The multi-agent system consists of Agents are deployed arbitrarily in three-dimensional space. All equipped with local coordinate system , and the global coordinate system cannot be obtained information, some agents (called leader agents) are equipped with visual sensors to measure their local orientation relative to the target , and at least two leading agents are not co-linear with the target, and the remaining agents (called follower agents) cannot measure the local orientation of the target; Step 2: Define the Agent The relative posture of the global coordinate system is ,in is a special orthogonal group and the pose is unknown; define the agent Relative Neighbor Agents The relative posture of , the relative posture can be measured by equipping a visual sensor; Define a common virtual coordinate system , the origin and direction of the public virtual coordinate system are unknown to the agent and are arbitrary; Step 3: Define the Agent and the target in the common virtual coordinate system The following positions are and , where the superscript represents the transpose symbol; Step 4: Define the Agent The direction is a unit direction vector And the kinematic model is , where is a unit vector The orthographic projection matrix is is the three-dimensional identity matrix; is the control input to be designed, which is used to project the control signal to vertical plane to adjust the pointing direction; Step 5: Define the set of all agents in the multi-agent system as , the leader agent subset is , followed by a subset of agents ,in is the number of leading agents, and satisfies To ensure the uniqueness of target positioning; Step 6: Define the communication network topology of the multi-agent system as a connected undirected graph ; Step 7: The multi-agent collaborative target pointing control method based on local orientation requires the interactive information between the agent and the neighboring agents to realize collaborative target pointing control. The nearest neighbor rule is used to define the agent The set of neighbor agents is ; Step 8: Due to the lack of a global coordinate system, the agent cannot directly obtain its own posture , the agents need to use distributed posture estimators to jointly determine a common virtual coordinate system And estimate itself in The posture below .

2. The method for multi-agent collaborative target pointing control based on local orientation according to claim 1, characterized in that: In step 8, the design expression of the distributed attitude estimator is: (1) in, For intelligent agents Public virtual coordinate system Lower your posture The estimate and ; For intelligent agents Relative Agent The relative posture and ; is a positive proportionality coefficient; Network topology diagram The adjacency weight matrix of .

3. The method for multi-agent collaborative target pointing control based on local orientation according to claim 2 is characterized in that: In step 8, set the initial estimate parameter of the distributed attitude estimator step setting , and the initial posture of the agent The rotation angle range around the rotation axis must be within to ensure the convergence of the distributed pose estimator.

4. The method for multi-agent collaborative target pointing control based on local orientation according to claim 1, characterized in that: Step 9: Each agent in the multi-agent system cannot directly obtain complete target location information. The multi-agent system uses the leader agent to Measure the local position of the target , each agent Your own location information and by communicating with neighboring agents Obtain neighbor target location estimation information during interaction , used for each agent Design the following collaborative target estimator: (2) in, For intelligent agents exist Next to the target position The estimated value of is the leading agent, then ,otherwise ; is a unit vector The orthogonal projection matrix is .

5. The method for multi-agent collaborative target pointing control based on local orientation according to claim 4 is characterized in that: In step 9, the collaborative target estimator initial target estimate parameter step is set, and the agent The initial estimate of the neighboring agent The initial estimates of are not equal, that is, .

6. The method for multi-agent collaborative target pointing control based on local orientation according to claim 1, characterized in that: The design of the pointing controller is as follows: Step 10: the agent obtains the target position information based on the distributed posture estimator and the collaborative target estimator. , and combined with its own location information , used to design the following multi-agent directed control input: (3) Among them, the orthogonal projection matrix The displacement difference between the estimated target position and the agent's own position Projected to the direction vector vertical plane, thus driving the pointing Rotate so that the pointing directions of all agents converge to the target direction.

Images