Control device design method and method for manufacturing mobile body
The control device design method estimates relative state quantities using observers and feedback controllers to maintain stable leader-follower control in multi-agent systems with restricted communication, addressing stability and tracking challenges.
Patent Information
- Application Number
- JP2023072434
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-26
- Publication Date
- 2024-11-08
AI Technical Summary
Existing leader-follower control systems in multi-agent systems face challenges in maintaining stability and tracking performance when communication between agents is restricted, particularly in environments like underwater operations where wireless communication is limited or unreliable.
A control device design method that utilizes an observer to estimate the relative state quantities of neighboring agents based on direct or indirect sensor measurements, combined with a feedback controller, to ensure stable leader-follower control even under communication limitations, using linear matrix inequalities to constrain observer and feedback gains for system stability and performance.
Enables stable and accurate leader-follower control in environments with restricted communication by estimating relative state quantities, reducing communication requirements, and ensuring system stability while improving tracking performance.
Smart Images

Figure 2024157832000001_ABST
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a method for designing a control device and a method for manufacturing a moving body. [Background technology]
[0002] Patent Document 1 describes a distributed control system in which a plurality of moving bodies (agents) recognize each other's positions and each moving body moves in an autonomous and distributed manner toward each of the high-risk points within a monitored area.
[0003] In the technology described in Patent Document 1, the consensus value to which each agent converges is time-invariant. In addition, the technology described in Patent Document 1 does not take into account the motion characteristics of the moving body. Therefore, in the technology described in Patent Document 1, in a multi-agent system in which one moving body that is a leader is followed by other moving bodies that are followers, it is difficult to control each follower to follow the leader whose state changes dynamically.
[0004] FIG. 17 is a block diagram showing a functional configuration of a conventional follower. Non-Patent Document 1 proposes a leader-follower control method for a linear multi-agent system. As shown in FIG. 17, each follower is a leader or follower (one or more agents that transmit information to a certain follower i, hereinafter also referred to as "adjacent agents") with which it can communicate. j (For example, position, velocity, attitude angle, attitude angular velocity, etc.) can be known, the state quantity x i and the state of the neighboring agent x j The control input u i is expressed by equation (2).
[0005]
number
[0006]
number
[0007] Furthermore, Non-Patent Document 1 describes a method for determining a feedback gain K so as to increase the stability of the system. [Prior art documents] [Patent documents]
[0008] [Patent Document 1] JP 2019-74918 A [Non-patent literature]
[0009] [Non-Patent Document 1] Weijun Cao, et. al., "Leader-follower consensus of linear multi-agent systems with unknown external disturbances", System & Control Letters, Vol. 82, pp. 64-70, 2015 Summary of the Invention [Problem to be solved by the invention]
[0010] For example, there are cases where communication is limited (the amount of communication is limited, and communication delays and interruptions occur), such as with UUVs operating underwater. Under conditions where communication is limited, it is necessary to take measures such as minimizing the information acquired from adjacent agents through communication and acquiring the state quantities of adjacent agents using sensors (cameras, laser scanners, etc.) without using wireless communication.
[0011] However, in the conventional technique, each follower can obtain all the states x jIn the conventional technology, it is assumed that all the state quantities x of neighboring agents can be obtained under the condition of communication restriction. j Therefore, it becomes difficult to perform leader-follower control.
[0012] The object of the present disclosure is to provide a method for designing a control device and a method for manufacturing a moving body, which can cause a follower to follow a leader in a group of moving bodies consisting of one leader moving body and one or more follower moving bodies that follow the leader, even if the follower cannot observe all of the state quantities of the other moving bodies. [Means for solving the problem]
[0013] According to one aspect of the present disclosure, a method for designing a control device is a method for designing a control device for a follower in a group of moving objects including one moving object that is a leader and one or more moving objects that are followers that follow the leader, the control device for the follower includes an observer that estimates a state quantity relative value with respect to the adjacent agent based on a relative position between the self-machine and an adjacent agent that is at least one of the leader and other followers and an observer gain, and a feedback controller that outputs a control input for the self-machine to follow the leader based on the estimated value of the state quantity relative value and a feedback gain, and the design method for the control device includes a linear matrix inequality that constrains all real parts of the poles of the observer to be away from the imaginary axis by a predetermined value on the negative side, and an H ∞ The method includes the steps of: designing an observer gain that satisfies both a linear matrix inequality that constrains the observer and the feedback controller to minimize the norm; and designing a feedback gain using the linear matrix inequality to improve the tracking control performance to the leader while ensuring the stability of the entire closed-loop system including the observer and the feedback controller designed for each of the followers.
[0014] According to one aspect of the present disclosure, a method for manufacturing a moving body is a method for manufacturing a follower moving body in a group of moving bodies consisting of one leader moving body and one or more follower moving bodies that follow the leader, and includes a step of manufacturing the control device based on a gain designed by the above-mentioned control device design method. Effect of the Invention
[0015] According to the above aspect, in a group of moving bodies consisting of one leader moving body and one or more follower moving bodies that follow the leader, the followers can be made to follow the leader even if they cannot observe all of the state quantities of the other moving bodies. [Brief description of the drawings]
[0016] [Figure 1] 1 is a diagram showing an outline of a group of moving objects according to a first embodiment. [Diagram 2] FIG. 2 is a block diagram showing a functional configuration of a follower according to the first embodiment. [Diagram 3] FIG. 1 is a first diagram showing an example of a network structure according to a first embodiment. [Figure 4] FIG. 2 is a second diagram showing an example of a network structure according to the first embodiment. [Diagram 5] 4 is a flowchart showing a flow of a method for designing a control device according to the first embodiment. [Figure 6] FIG. 4 is a diagram showing a simulation result according to the first embodiment. [Figure 7] FIG. 11 is a block diagram showing a functional configuration of a follower according to a second embodiment. [Figure 8] 10 is a flowchart showing a flow of a method for designing a control device according to a second embodiment. [Figure 9] 10 is a flowchart showing a flow of a method for designing a control device according to a third embodiment. [Figure 10] FIG. 13 is a diagram illustrating a design example of a feedback gain according to the third embodiment. [Figure 11]FIG. 13 is a first diagram showing an example of a network structure according to a fourth embodiment. [Figure 12] FIG. 13 is a second diagram showing an example of a network structure according to the fourth embodiment. [Figure 13] FIG. 13 is a diagram for explaining a simulation according to the fourth embodiment. [Figure 14] FIG. 13 is a diagram showing a simulation result according to the fourth embodiment. [Figure 15] 13 is a flowchart showing the flow of a method for designing a control device according to a modified example of the fourth embodiment. [Figure 16] 1 is a schematic block diagram illustrating a hardware configuration of a control device for a moving object according to at least one embodiment. [Figure 17] FIG. 1 is a block diagram showing a functional configuration of a conventional follower. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0017] <First embodiment> The first embodiment will be described below with reference to FIGS.
[0018] (Overall composition of the moving group) FIG. 1 is a diagram showing an outline of a group of moving objects according to the first embodiment. The moving object group 1 is made up of a plurality of moving objects that operate in cooperation with each other. The moving objects are autonomously operable drones such as UUVs (Unmanned Undersea Vehicles), UGVs (Unmanned Ground Vehicles), and UAVs (Unmanned Aerial Vehicles). As shown in FIG. 1, the moving object group 1 according to this embodiment has one leader moving object L (hereinafter also referred to as "leader L") and one or more follower moving objects F (hereinafter also referred to as "follower F") that follow the leader L while maintaining a specified formation. In the following description, the leader L and the follower F are also referred to as "agents".
[0019] Fig. 1 shows an example in which nine followers F1 to F9 follow one leader L in a predetermined formation. In the example in Fig. 1, a formation is specified in which five followers F1 to F5 and four followers F6 to F9 are arranged on concentric circles L1 and L2 centered on the leader L.
[0020] Each follower F has a network structure in which it is connected to the leader L directly or indirectly via other followers F. The network structure represents the connection relationship of information transmission between agents through wireless communication or sensor observation.
[0021] In the example of FIG. 1, the followers F1 to F5 are directly connected to the leader L. That is, the followers F1 to F5 obtain information required for leader-follower control from the leader L. The followers F6 to F9 are also connected to the other followers F1 to F4, respectively, and are indirectly connected to the leader L via these other followers F1 to F4. That is, the followers F6 to F9 obtain information required for leader-follower control from the followers F1 to F4, respectively. Therefore, the followers F6 to F9 indirectly follow the leader L via the followers F1 to F4.
[0022] In the following description, the direct connection partners of each follower F are also referred to as "adjacent agents."
[0023] The leader L dynamically changes its state (position, posture, speed, etc.) according to, for example, external operational instructions or a pre-given scenario. On the other hand, each follower F is not given operational instructions or a scenario, but automatically follows the leader L based on information obtained from adjacent agents.
[0024] (Follower function configuration) FIG. 2 is a block diagram showing a functional configuration of a follower according to the first embodiment. 2, the follower F includes a machine body 10 and a control device 20. Also, as shown in FIG. 2, the leader-follower control of the follower F is modeled by a closed loop system CL consisting of the machine body 10 and the control device 20.
[0025] The machine 10 is a control target of the control device 20, and the follower F i (i-th vehicle) includes a propulsion system (or drive system), a steering system, etc. Figure 2 shows the vehicle 10 as a linear system having the dynamics of equation (1) above.
[0026] The aircraft 10 receives the control input u i is input, and the state x i The output is the control input u i consists of multiple control inputs such as the propulsion system and steering system. For example, if the moving body (leader L and follower F) is a UUV, the control input u i is the thruster rotation speed, steering angle, etc. State quantity x i consists of multiple state quantities such as the position (three-dimensional coordinates), attitude, and velocity of the aircraft 10.
[0027] In addition, the aircraft 10 according to this embodiment has a state quantity x i Among the multiple state quantities included in i Only the above-mentioned signal is output to the control device 20.
[0028] The control device 20 includes an acquisition unit 201 , an observer 202 , and a feedback controller 203 .
[0029] The acquisition unit 201 acquires the relative position y Fi In the example of FIG. 1, when the follower F4 is a follower, the acquisition unit 201 acquires the relative position y F4 In addition, when the agent is a follower F9, the acquisition unit 201 acquires the relative position y F9 Get the.
[0030] The acquisition unit 201 receives, for example, the position y from a neighboring agent by wireless communication. j and obtain the aircraft's position y i and the relative position y Fi In addition, the acquisition unit 201 acquires the position y i Based on the position and distance of the neighboring agents measured by the sensor, the relative position y Fi may be obtained.
[0031] The observer 202 is a so-called unknown input observer (UIO). The observer 202 determines the relative position y Fi Based on the observer gains H, F, and P, the relative state values x Fi The estimated relative state value is ^x Fi ("^x" is written as x with a circumflex (^) attached.)
[0032] The feedback controller 203 estimates the relative state value ^x Fi Based on the feedback gain K, the control input u for the aircraft 10 (own aircraft) to follow the leader L is i is output to the aircraft 10. In this embodiment, the control input u i is expressed by equation (3) instead of equation (2) above.
[0033]
number
[0034] As shown in Figure 17, in conventional leader-follower control, the control input u i is the relative value of all the state quantities (x i -x j ) to enable following the leader.
[0035] In contrast, in this embodiment, the control device 20 of the follower F is operated under conditions where communication is restricted, and the position y Fi Only the relative position y Fi The other state quantity relative values other than are estimated by the observer 202. As a result, even if the control device 20 of the follower F cannot observe all state quantities of other agents, it can estimate and obtain all state quantity relative values necessary for leader-follower control.
[0036] In addition, the control device 20 of the follower F can reduce the amount of communication with the adjacent agent by limiting the information acquired from the adjacent agent to only the location information. Furthermore, since there is only one adjacent agent as a communication partner, the amount of communication and the number of communications among all agents can be reduced compared to the conventional technology in which each follower communicates with multiple nearby agents.
[0037] In addition, the control device 20 of the follower F detects the relative position y Fi In this way, even in an environment where communication delays or interruptions are likely to occur (such as underwater), the relative position y Fi By using this method, it is possible to obtain the leader L more reliably and perform resilient leader-follower control that can follow the leader L.
[0038] (Design conditions) When a general observer is applied, the observer determines the control input u i and the control input u of the neighboring agent j Relative value (input relative value) u Fi =u i -u j is required.
[0039] In contrast, in this embodiment, taking into consideration the case where communication is restricted, the information obtained from the neighboring agents is calculated based on the relative position y FiIn other words, in this embodiment, only the input relative value u Fi Therefore, in this embodiment, the input relative value is treated as an unknown input, and an observer 202 capable of estimating all state quantities while eliminating the influence of the unknown input is designed.
[0040] The estimated value of the state quantity relative value by the observer 202 is expressed by equation (4), and the true value is expressed by equation (5).
[0041]
number
[0042] Note that A, B, and C are values determined from the dynamics of the aircraft 10 shown in FIG. 2, and H, F, and P are observer gains.
[0043]
number
[0044] Observer estimation error e i is expressed by equation (6), and the dynamics of the observer estimation error is expressed by equation (7).
[0045]
number
[0046]
number
[0047] In this case, if the following equations (8)-(10) hold, then equation (11) holds.
[0048]
number
[0049]
number
[0050]
number
[0051]
number
[0052] Also, if H is stable, the estimation error e i approaches zero and equation (12) holds.
[0053]
number
[0054] However, in order for F to exist for equation (10) to hold, rank(CB) = rank(B) must hold. For example, if the number of state quantities of the moving body is less than the number of control inputs, the rank condition for equation (10) to hold cannot be satisfied. In that case, it is generally impossible to construct an unknown input observer.
[0055] For this reason, when designing the observer 202 according to this embodiment, the unknown input is the observer estimation error e i The impact on H ∞ Expressed in terms of norm, H ∞Apply RxUIO (Relaxed UIO) that minimizes the norm. RxUIO is a technology that enables unknown input observers to be designed even when the ranks of matrices do not match. In this embodiment, for example, the RxUIO technology described in "Daoliang Tan, et. al., "A Relaxed Solution to Unknown Input Observers for State and Fault Estimation", IFAC-PapersOnLine, Volume 48, Issue 21, pages 1048-1053, 2015" is used as part of the observer design.
[0056] When RxUIO is applied, equation (10) does not hold, but it is possible to design the observer gains H, F, and P so that equations (8) and (9) hold. Then, the observer estimation error e i In equation (7) showing the dynamics of the equation, two or three terms can be set to zero, and the equation can be expressed as follows:
[0057]
number
[0058] Here, let the state quantity of N followers F be x and the observer estimation error be e. Then, the state quantity x of leader L is L , input u for reader L L Using this, the closed loop system CL of the entire system (all agents) is expressed by equation (14).
[0059]
number
[0060] Here, M is the graph Laplacian with the matrix related to leader L (N+1th column, N+1th row) deleted when the N+1th agent is leader L, and is expressed by equation (15).
[0061]
number
[0062] In addition, A, B, and C in equation (14) are values determined from the dynamics of the aircraft 10 shown in FIG. 2, H, F, and P are observer gains, K is a feedback gain, and I is a unit matrix.
[0063] In equation (14), if we look at equation (16) of the system matrix A', which indicates the stability of the system, we can see that the effect of the unknown input of RxUIO cannot be completely eliminated, so the separation theorem of the observer and feedback control system does not hold. In other words, since the observer and feedback control are dependent on each other, if they are designed independently, the system will be unstable.
[0064]
number
[0065] In order to establish the separation theorem and ensure the stability of the system, the following must be satisfied in the above equation (16):
[0066]
number
[0067] The following three conditions must be met to make these sufficiently small:
[0068] (a) Reduce M, which indicates the network structure. (b) Decrease the feedback gain K. (c) H for unknown input ∞ Reduce the norm-related (FC-I)B
[0069] Regarding condition (a), in order to reduce M while satisfying the prerequisite that "each follower F has a network structure in which it is directly or indirectly connected to the leader L," it is desirable to keep the number of neighboring agents from which relative positions are obtained small.
[0070] FIG. 3 is a first diagram illustrating an example of a network structure according to the first embodiment. Figure 3 shows the network structure of the mobile group 1 shown in Figure 1. In the examples of Figures 1 and 3, each follower F has only one adjacent agent from which it obtains information so that M is minimized. In other words, followers F1, F2, F3, F4, and F5 have only leader L as an adjacent agent, and followers F6, F7, F8, and F9 have only followers F1, F2, F3, and F4 as adjacent agents, respectively.
[0071] FIG. 4 is a second diagram illustrating an example of a network structure according to the first embodiment. Fig. 4 illustrates a network structure of a moving object group 1 in which four followers F1 to F4 follow one leader L. As shown in Fig. 4, the network structure may be such that the leader L and each follower F are connected in series.
[0072] Regarding condition (b), the stability of the system can be guaranteed by reducing the feedback gain K. On the other hand, in order to improve the tracking performance for the leader L, it is necessary to make the feedback gain K sufficiently large. Therefore, if the feedback gain K is made too small in order to stabilize the system, the feedback will not be effective and the tracking performance for the leader L will deteriorate.
[0073] For condition (c), H ∞When minimizing the norm of , the real part of the pole of RxUIO approaches the imaginary axis, slowing down the convergence of the observer estimate. On the other hand, in order to improve the tracking performance to the leader L, it is necessary to move the real part of the pole of RxUIO away from the imaginary axis to the negative side in order to speed up the convergence of the observer estimate. Therefore, if (FC-I)B is made too small to stabilize the system, there is a possibility that the observer will output an erroneous estimate, and the tracking performance to the leader L will deteriorate.
[0074] In other words, the conditions (b) and (c) for ensuring the stability of the system are contradictory to the conditions for improving the tracking performance to the leader L. Therefore, in this embodiment, in order to ensure the stability of the system and improve the tracking performance to the leader L (leader-follower control performance), the conditions (b) and (c) are changed, and the observer 202 and the feedback controller 203 are designed to satisfy the following conditions (a), (b'), and (c').
[0075] (a) Reduce M, which indicates the network structure. (b') Set the feedback gain K to the maximum value within the range that guarantees stability of the entire system. (c') The real part of the eigenvalue of the gain H, which is the pole of RxUIO, is placed so that it is away from the imaginary axis by a predetermined value α or more on the negative side, and then H, which represents the influence of the unknown input, is placed. ∞ Design an RxUIO with the minimum norm
[0076] (Control device design method) FIG. 5 is a flowchart showing the flow of the method for designing the control device according to the first embodiment. A method for designing the observer 202 and the feedback controller 203 of the control device 20 will be described in detail below with reference to FIG.
[0077] First, the network structure of the mobile group 1 is defined (step S01). Specifically, to satisfy the condition (a), a network structure is defined in which all followers F are virtually connected directly or indirectly to the leader L, and the number of direct connection partners (adjacent agents) of each follower is small. In this embodiment, for example, the number of adjacent agents is set to one.
[0078] Next, the observer gains H, F, and P of the observer 202 are designed using Linear Matrix Inequality (LMI) so as to satisfy the condition (c') (step S02).
[0079] In condition (c'), α, which determines the convergence speed of the observer estimate, is a design parameter that is arbitrarily set by the designer of the control device 20. The LMI conditions are expressed by the following equations (18) and (19). Equation (18) is the unknown input H ∞ This is a constraint for minimizing the norm, and uses the RxUIO technique. In RxUIO, as mentioned above, H ∞ When minimizing the norm, the real part of the pole of RxUIO approaches the imaginary axis, which causes a problem of slowing down the convergence speed of the observer estimate. Therefore, in this embodiment, the linear matrix inequality of equation (19) is further added to design the observer gain. Equation (19) is a constraint for moving the real part of the pole of RxUIO by α to the negative side from the imaginary axis.
[0080]
number
[0081]
number
[0082] In step S02, variables P, Y, and V are designed so as to minimize ξ, which represents the unknown input, while satisfying the LMI conditions in equations (18) and (19). If equation (18) is satisfied, H ∞The norm is less than ξ, where Y=PF and V=PP1. By determining the variables P, Y, and V, the observer gains H, F, and P of the observer 202 are also determined.
[0083] Next, the feedback gain K of the feedback controller 203 is designed.
[0084] First, an upper limit (initial value) of the magnitude ||K||=γ of the feedback gain K is set (step S03). This initial value is a design parameter that is arbitrarily set by the designer of the control device 20.
[0085] Next, the feedback gain K is designed based on the LMI condition assuming that all state quantities are observable while satisfying ||K||<γ (step S04). At this time, the LMI condition is expressed by the following equations (20) and (21).
[0086]
number
[0087]
number
[0088] In step S04, variables X and Y that satisfy the LMI conditions of equations (20) and (21) are designed. Here, KX=Y, and when variables X and Y are determined, the feedback gain K is also determined (K=YX -1 ). In step S02, since RxUIO (observer 202) that minimizes the influence of unknown inputs has been designed, the estimated values of all state quantities of adjacent agents are input to the feedback controller 203 as shown in FIG. 2. Equation (20) is a constraint for guaranteeing system stability when the observer estimated values match the true values. Equation (21) is a constraint for suppressing the magnitude of the feedback gain K to γ or less.
[0089] Next, the poles of the closed loop system CL of the entire system (all followers F) are calculated (step S05) using the designed feedback gain K. The poles of the closed loop system CL are eigenvalues of the system matrix A' expressed by the above equation (16).
[0090] Next, it is determined whether the real parts of the poles are all negative (step S06).
[0091] If the real parts of the poles are all negative (step S06; YES), the system is stable. In this case, the upper limit value γ of the magnitude of the feedback gain K is increased by a predetermined amount dγ (step S07), and the process returns to step S04 where the feedback gain is designed again. This predetermined amount dγ is a design parameter that is arbitrarily set by the designer of the control device 20.
[0092] On the other hand, if at least one of the real parts of the poles is not negative (step S06; NO), the system is unstable. In this case, the value of the feedback gain K designed with the previous upper limit value γ=γ-dγ is adopted (step S08).
[0093] In this way, while the real parts of the poles are all negative, the feedback gain K is designed while gradually increasing the upper limit value γ of the magnitude of the feedback gain K (steps S04 to S07), so that the feedback gain K can be maximized within a range in which the stability of the entire system is guaranteed. In other words, the above-mentioned condition (b') can be satisfied.
[0094] In this embodiment, the example in which α, which determines the convergence speed of the observer estimate, is fixed and set by the designer in step S02 has been described, but the present invention is not limited to this. The closer the value of α is to the imaginary axis, the slower the convergence speed becomes, while the value of ξ becomes smaller, and the influence of unknown input becomes smaller. Therefore, in another embodiment, for example, step S02 may be repeated while changing the value of α within a range specified by the designer, and α may be obtained such that the value of ξ becomes equal to or smaller than a reference value preset by the designer and the convergence speed of the observer estimate becomes equal to or smaller than the speed preset by the designer. In this way, it is possible to design appropriate observer gains H, F, and P that balance both the convergence speed and the estimation accuracy of the observer 202.
[0095] In addition, in this embodiment, a network structure in which each follower F has only one adjacent agent has been described as an example, but the present invention is not limited to this. In other embodiments, a network structure in which each follower F has two or more adjacent agents may be used. For example, if the adjacent agents of follower F1 are the leader L and follower F2, the acquisition unit 201 of the control device 20 acquires the relative positions (y Fi1 =y i -y j1 ,y Fi2 =y i -y j2 ) The observer 202 also obtains the relative state values (x Fi1 ,x Fi2 In this case, in step S01 of Fig. 5, a network structure in which each follower F has a plurality of adjacent agents is defined, and the control device 20 is designed.
[0096] Based on the gain designed by the above-mentioned design method, the control device 20 and a moving body (follower F) equipped with the control device 20 are manufactured.
[0097] (Simulation results) FIG. 6 is a diagram showing a simulation result according to the first embodiment. FIG. 6 is a graph showing the results of a simulation of leader-follower control using a moving object designed (manufactured) by the design method of this embodiment. In this simulation, a moving object group 1 consisting of one leader L and nine followers F moved in the formation shown in FIG. 1. The five followers F1 to F5 were arranged at equal intervals on a concentric circle L1 with a radius of 50 m centered on the leader L, and the four followers F6 to F9 were arranged at equal intervals on a concentric circle L2 with a radius of 100 m centered on the leader L. FIG. 6 shows the time history of the deviation from the target trajectory of each follower F1 to F9 in this simulation, and the time history of the speed of the leader L. As shown in FIG. 6, although a deviation from the target trajectory occurred temporarily due to the change in the trajectory of the leader L, it was confirmed that the deviation from the target trajectory gradually converged to 0.
[0098] (Action, effect) As described above, the control device 20 of the follower F according to this embodiment i and the relative position y between adjacent agents Fi Based on the observer gains H, F, and P, the relative state value x Fi and an observer 202 that estimates the state quantity relative value ^x Fi Based on the feedback gain K, the vehicle F i The control input u for following the leader L i The control device 20 is also designed by using a linear matrix inequality that constrains all real parts of the poles of the observer 202 to be a predetermined value α away from the imaginary axis on the negative side, and a feedback controller 203 that outputs ∞ The method includes a step of designing observer gains H, F, and P that satisfy both the linear matrix inequality that constrains the observer 202 and the feedback controller 203 designed for each follower F, and a step of designing a feedback gain K using the linear matrix inequality to improve the tracking control performance for the leader L while ensuring the stability of the entire closed-loop system CL including the observer 202 and feedback controller 203 designed for each follower F.
[0099] By doing this, even if it is not possible to observe all the state quantities of neighboring agents due to communication restrictions, for example, the relative position y Fi From all the state quantity relative values x Fi It is possible to design an observer 202 capable of estimating the given values α and H ∞ By designing the observer gains H, F, and P that satisfy the constraints on the norm, even if the relative values of the control inputs to the neighboring agents are unknown, the influence of the unknown inputs to the observer 202 is sufficiently small to ensure that all state quantity relative values x Fi It is possible to estimate and suppress a decrease in the convergence speed of the estimate of the observer 202. In addition, by designing the feedback gain K using the linear matrix inequality as described above, it is possible to ensure the stability of the entire system (all agents) and improve the tracking performance of each follower F to the leader L.
[0100] In addition, the design method of the control device 20 further includes a step of designing a feedback gain K that satisfies both a linear matrix inequality that constrains the closed-loop system CL to be stable while gradually increasing an upper limit value γ of the magnitude of the feedback gain K, and a linear matrix inequality that constrains the magnitude of the feedback gain K to be equal to or less than the upper limit value γ while the real parts of the poles of the closed-loop system CL are all negative.
[0101] In this way, the feedback gain K can be increased within a range in which the stability of the entire system (all agents) is guaranteed. This makes it possible to more reliably ensure both the guarantee of the stability of the entire system and the improvement of the tracking performance of each follower F to the leader L.
[0102] Also, in the step of designing the observer gains H, F, and P, the design of the observer gains H, F, and P may be repeatedly performed while changing a predetermined value α that determines the convergence speed of the observer within a specified range, and the predetermined value α may be determined so that the convergence speed is equal to or less than a preset speed.
[0103] In this way, it is possible to design appropriate observer gains H, F, and P that balance the convergence speed and estimation accuracy of the observer 202.
[0104] In addition, the manufacturing method of the moving object is to determine the relative position y Fi Based on the observer gains H, F, and P, the relative state value x Fi and designing an observer 202 for estimating the state quantity relative value ^x Fi Based on the feedback gain K, the control input u for the robot to follow the leader L is calculated. i The step of designing the observer 202 includes a step of designing a feedback controller 203 that outputs: ∞ Design observer gains H, F, and P that satisfy both linear matrix inequalities constrained to minimize the norm.
[0105] By doing this, even if it is not possible to observe all the state quantities of neighboring agents due to communication restrictions, for example, the relative position y Fi From all the state quantity relative values x Fi It is possible to manufacture a moving body (follower F) capable of estimating the predetermined value α and H ∞ By designing the observer gains H, F, and P that satisfy the constraints on the norm, even if the relative values of the control inputs to the neighboring agents are unknown, the influence of the unknown inputs to the observer 202 is sufficiently small to ensure that all state quantity relative values x Fi It is possible to estimate the state quantity relative value with higher accuracy, and to suppress a decrease in the convergence speed of the estimated value of the observer 202. Therefore, it is possible to manufacture a moving body capable of estimating the state quantity relative value with higher accuracy.
[0106] <Second embodiment> Next, a second embodiment will be described with reference to Figures 7 and 8. Components common to the above-described embodiment will be given the same reference numerals and detailed description will be omitted.
[0107] (Follower function configuration) FIG. 7 is a block diagram showing a functional configuration of a follower according to the second embodiment. As shown in FIG. 7, the control device 20 of the follower F according to this embodiment further includes an integrator 204 and an adder 205.
[0108] The integrator 204 detects the relative position y Fi (position deviation) is accumulated, and the control input u i The correction value is output.
[0109] The adder 205 receives the control input u i1 ,u i2 At least one control input u i1 The correction value output by the integrator 204 is added to the control input u i1 ' (corrected control input) is input.
[0110] (Control device design method) FIG. 8 is a flowchart showing the flow of a method for designing a control device according to the second embodiment. Hereinafter, a method for designing the control device 20 according to this embodiment will be described with reference to Fig. 8. Note that steps S01 to S08 are the same processes as those in the first embodiment (Fig. 5), and therefore descriptions thereof will be omitted. In this embodiment, the method for designing the control device 20 further includes a step of adding an integrator 204 and an adder 205 to the control device 20 (step S10).
[0111] For simplicity, let us assume that the vehicle is a UUV and the control input u i1 ,u i2 are the thruster rotation speed command value and rudder command value, respectively.i2 The target attitude controlled by the rotation speed command value u changes stepwise, so it can be tracked without any steady-state error. i1 The target position controlled by generates a steady-state deviation because the position of the leader L, which is the tracking partner, changes from moment to moment in a ramp shape.
[0112] Therefore, in step S10, the target position (relative position y Fi ) is designed to be added. The correction value output by the integrator 204 is then converted into the rotation speed command value u i1 In this case, a design is made to add an adder 205 that adds to
[0113] As shown in the example of Fig. 6, the rudder command value u i2 In another embodiment, the correction value may not be added to all the control inputs u i1 ,u i2 , . . . a correction value may be added to the
[0114] (Action, effect) As described above, the method for designing the control device 20 according to this embodiment further includes the step of adding the integrator 204 that outputs a correction value that reduces the steady-state deviation.
[0115] In this way, it is possible to design the control device 20 that can reduce the steady-state deviation and accurately track the target position that changes from moment to moment as the reader L moves.
[0116] <Third embodiment> Next, a third embodiment will be described with reference to Figures 9 and 10. Components common to the above-described embodiment will be given the same reference numerals and detailed description will be omitted.
[0117] In this embodiment, in the design process of the observer gains H, F, P (step S02 in FIG. 5) and the design process of the feedback gain K (steps S03 to S08 in FIG. 5), the observer gains H, F, P and the feedback gain K are calculated for each speed.
[0118] (Control device design method) FIG. 9 is a flowchart showing the flow of a method for designing a control device according to the third embodiment. Step S21 in FIG. 9 is the same as step S01 (FIG. 5) in the above embodiment, and therefore a description thereof will be omitted.
[0119] The motion characteristics of the moving body (follower F) change depending on the speed. That is, the matrices A and B expressed in the linear system (FIG. 2) of the vehicle 10 change depending on the speed. For this reason, in this embodiment, a step (step S22) of designing observer gains H, F, and P for each speed and a step (step S23) of designing feedback gain K for each speed are respectively performed.
[0120] In step S22, the same process as step S02 in Fig. 5 is performed for each of a plurality of speeds specified by the designer. For example, assume that speeds of 2 knots, 2.5 knots, 3 knots, 3.5 knots, ... are specified. First, observer gains H, F, P corresponding to a speed of 2 knots that satisfy both of the above expressions (18) and (19) are designed using A and B corresponding to a speed of 2 knots. Next, observer gains H, F, P corresponding to a speed of 2.5 knots that satisfy both of the above expressions (18) and (19) are designed. Similarly, observer gains H, F, P corresponding to a speed of 3 knots, 3.5 knots, ... are designed.
[0121] In step S23, the same processes as steps S03 to S08 in Fig. 5 are performed for each of the multiple speeds specified in step S22. In the above example, first, using A and B corresponding to a speed of 2 knots, a feedback gain K corresponding to a speed of 2 knots that is maximum within a range in which the real parts of the poles of the closed loop system CL are all negative is designed. Next, using A and B corresponding to a speed of 2.5 knots, a feedback gain K corresponding to a speed of 2.5 knots that is maximum within a range in which the real parts of the poles of the closed loop system CL are all negative is designed. Similarly, feedback gains K corresponding to speeds of 3 knots, 3.5 knots, ... are designed.
[0122] FIG. 10 is a diagram illustrating a design example of a feedback gain according to the third embodiment. 10 shows an example in which the feedback gain K of each agent is designed for a speed of 2 to 6 knots (in increments of 0.5 knots) in step S23. Although not shown in the figure, the observer gains H, F, and P have also been similarly designed.
[0123] The control device 20 of each agent stores in advance a table of feedback gains K designed for each speed in a storage area. In addition, the feedback controller 203 of the control device 20 uses the linearly complemented feedback gain K shown in FIG. 10 based on the table of feedback gains K and the speed of the agent's own vehicle to calculate the control input u i This allows the control device 20 to appropriately control the aircraft 10 in accordance with the motion characteristics of the aircraft's speed, thereby further improving the ability to follow the leader L.
[0124] Similarly, the control device 20 of each agent stores in advance in a storage area a table of the observer gains H, F, P designed for each speed. Also, the observer 202 of the control device 20 calculates the state quantity relative value x using the table of the observer gains H, F, P and the observer gains H, F, P linearly complemented based on the speed of the own vehicle. FiIn this way, the control device 20 estimates the state quantity relative value x Fi The estimation accuracy can be improved.
[0125] (Action, effect) As described above, in this embodiment, in the process of designing the observer gains H, F, P, a plurality of observer gains H, F, P corresponding to a plurality of specified speeds are designed.
[0126] In this way, it is possible to design an observer 202 capable of estimating state quantity relative values with high accuracy according to the speed of the aircraft 10, whose motion characteristics change according to the speed.
[0127] In addition, in the process of designing the feedback gain K, a plurality of feedback gains K corresponding to a plurality of specified speeds are designed.
[0128] In this way, it is possible to design a control device 20 (feedback controller 203) that can appropriately control the aircraft 10 in accordance with its own speed, for an aircraft 10 whose motion characteristics change depending on its speed.
[0129] <Fourth embodiment> Next, a fourth embodiment will be described with reference to Figures 11 to 14. Components common to the above-mentioned embodiments are given the same reference numerals and detailed description will be omitted.
[0130] For example, if one follower leaves the group of mobile objects due to a malfunction, the follower that had been receiving information from the follower will reconnect to another agent (leader or follower) and will obtain information from the new agent thereafter. In other words, the network structure may change during the operation of the group of mobile objects due to the departure of a follower.
[0131] However, in conventional technology, the control system was designed on the assumption that the network structure would not change during operation of the mobile group. Therefore, in a control system designed with conventional technology, there is a possibility that leader-follower control will not be performed normally if a follower leaves the mobile group.
[0132] In this embodiment, a control device 20 is designed that enables stable leader-follower control even after a change in the network structure, taking into consideration a case where a follower leaves the mobile group 1. Note that in this embodiment, the number of adjacent agents of each follower F is limited to one, and a case where one of multiple followers F leaves is considered.
[0133] (Control device design method) In this embodiment, an example will be described in which a method of designing a feedback gain K that takes into account a change in network structure is applied to the design method of the first embodiment (FIG. 5) among the above-mentioned embodiments.
[0134] 5, a network structure is defined in which the number of adjacent agents from which each follower F obtains a relative position is limited to one. Steps S02 to S08 are the same as those in the first embodiment, except for the design process (step S04) of the feedback gain K described below.
[0135] The design process (step S04) of the feedback gain K taking into consideration cases where the network structure changes will be described.
[0136] The network structure before the network change is M α The network structure after the change (after one follower F leaves) is denoted as M β The network structure M α Let Xα and Yα be the variables obtained by solving the LMI for. In other words, the above equation (20) becomes the following equation (22).
[0137]
number
[0138] Network Structure M α The feedback gain K designed for the network structure M β If the following equation (23) holds when applied to the above equation, the system can be said to be stable even after the network structure change.
[0139]
number
[0140] As described in the above embodiment, formula (20) is a necessary condition for the system to be stable in a predefined network structure. Therefore, when formula (22) obtained by rewriting formula (20) is satisfied, formula (23) is satisfied if formula (24) below is satisfied.
[0141]
number
[0142] Therefore, for any M β For X, we solve the two LMIs in equation (22) and equation (25) below. α ,Y α The feedback gain K=Y α X α -1 Using this, we can determine whether the network structure M α ,M β Stability can be guaranteed even for
[0143]
number
[0144] In addition, if we limit it to the departure of one follower F, as described below, α -M βIt can be analytically shown that is always a semidefinite matrix. Therefore, any network structure M after one follower F leaves β For this reason, the stability of the system can be guaranteed if the following equation (26) is satisfied.
[0145]
number
[0146] Therefore, in this embodiment, in step S04 of FIG. 5, a feedback gain K that satisfies the three LMI conditions of equation (21) described in the above embodiment and equations (22) and (26) described in this embodiment is designed.
[0147] Also, (M α -M β ) is always a positive semidefinite matrix, which will be described with reference to FIGS.
[0148] (A) Case where the follower at the end of the link breaks away If a follower F at the end of the mobile group 1 leaves, there is no need to reconnect the network. Let i be the follower that left and j be the follower that was sending information to the follower that left. α -M β If only j → i is communicating (M α ) is the same as the graph Laplacian of M α -M β It is obvious that the eigenvalues of are 0 or 1 (0 or more), and M α -M β is always a positive semidefinite matrix (M α -M β ≥ 0).
[0149] FIG. 11 is a first diagram illustrating an example of a network structure according to the fourth embodiment. FIG. 11 shows an example of a mobile group 1 consisting of one leader L and four followers F1 to F4. (a) of FIG. 11 shows the network structure M before the change. α, (b) shows the network structure after the change M β , (c) M α -M β Shows.
[0150] Specifically, an example in which the end follower F4 is detached will be described with reference to FIG. 11. The network structure M after the change shown in FIG. 11(b) will be described. β In the row of followers F1 to F3, there is no change, and the row of the departing follower F4 becomes all 0. Also, as shown in FIG. 11(c), M α -M β Only the row of the defecting follower F4 (row 4) remains, and all other rows become 0. In other words, M α -M β In the case where only follower F2 communicates with follower F4 (M α ) is the graph Laplacian of M α -M β The eigenvalues of are only 0 and 1, and M α -M β It has been proven that ≥ 0.
[0151] (B) Case where a follower leaves during the connection Next, if a follower leaves during the connection, the follower who obtained information from the left follower must reconnect to another agent. Let the left follower be i, the follower who obtained information from the left follower be j, and the agent from which j newly obtains information (reconnects) be r. In this case, M α -M β The i-th row of M will remain, the i-th column of the j-th row will be -1, the r-th column will be 1 (if r is the leader L, this will not be reflected), and all other elements will be 0. Then, M α -M β It is obvious that the eigenvalues of are 0 or 1 (0 or more), and M α -M β is always a positive semidefinite matrix (M α -M β ≥ 0).
[0152] FIG. 12 is a second diagram illustrating an example of a network structure according to the fourth embodiment. FIG. 12 shows an example of a mobile group 1 consisting of one leader L and four followers F1 to F4. (a) of FIG. 12 shows the network structure M before the change. α , (b) shows the network structure after the change M β , (c) M α -M β Shows.
[0153] Specifically, an example in which follower F2 leaves in the middle of a connection will be described with reference to Fig. 12. When follower F2 leaves, follower F4 that was connected to follower F2 must reconnect to another follower F. For example, as shown in Fig. 12(b), assume that follower F4 reconnects to follower F1. The same applies when follower F4 reconnects to another agent (leader L or follower F3).
[0154] The network structure M after the change shown in FIG. 12(b) β In the example, all the rows of the departing follower F2 are set to 0. Also, the row of follower F4, which obtained information from the departing follower F2, is set to a new network structure. In other words, the column of the departing follower F2, which was the original link partner, is set to 0, and the column of follower F1, which is the new link partner, is set to -1. Also, as shown in (c) of FIG. 12, M α -M β is M α The row (2nd row) of follower F2 remains, and the network structure changes. The first column of the row (4th row) of follower F4 becomes 1, the second column becomes -1, and all other elements become 0. Therefore, even in the case where a follower leaves in the middle of the connection, M α -M β The eigenvalues of are only 0 and 1, and M α -M β It has been proven that ≥ 0.
[0155] (Simulation results) FIG. 13 is a diagram for explaining a simulation according to the fourth embodiment. Fig. 13 shows the contents of a simulation of leader-follower control by moving objects designed (manufactured) by the design method of this embodiment. In this simulation, a moving object group 1 consisting of one leader L and nine followers F moved in the formation shown in Fig. 13(a). The five followers F1 to F5 were arranged at equal intervals on a concentric circle L1 with a radius of 50 m centered on the leader L, and the four followers F6 to F9 were arranged at equal intervals on a concentric circle L2 with a radius of 100 m centered on the leader L.
[0156] In this simulation, follower F4 failed at time t1, and 5 seconds after the failure, the network structure was changed as shown in Fig. 13(b). Specifically, follower F9, which had acquired information from follower F4, was reconnected to follower F5, and thereafter acquired its relative position from follower F5.
[0157] FIG. 14 is a diagram showing a simulation result according to the fourth embodiment. FIG. 14 shows the time history of deviation from the target trajectory of each follower F1 to F9 and the time history of the speed of the leader L in the simulation of FIG. 13. As shown in FIG. 14, at time t1, follower F4 broke down and left the moving body group 1. The time history of deviation from the target trajectory of follower F4 after that is not shown. In addition, follower F9, which had acquired information from follower F4, followed the broken follower F4 for 5 seconds from time t1, so that the deviation from the target trajectory temporarily increased. However, it was confirmed that the deviation from the target trajectory of follower F9 gradually decreased and finally converged to 0 because follower F9 changed its connection partner to follower F5.
[0158] (Modification) In the fourth embodiment, the method of further adding an LMI condition when designing the feedback gain K for the control device 20 of the follower F in the first embodiment (FIG. 2) has been described, but the present invention is not limited to this. For example, in another embodiment, the method of designing the feedback gain K in the fourth embodiment may be applied to a conventional follower control device (FIG. 17).
[0159] FIG. 15 is a flowchart showing the flow of a method for designing a control device according to a modification of the fourth embodiment. When applying the design method of the feedback gain K according to the fourth embodiment to a conventional control device (FIG. 17), first, a network structure is defined in the same manner as in the fourth embodiment (step S31). That is, a network structure is defined in which the number of adjacent agents from which each follower F obtains a relative position is limited to only one.
[0160] Next, the feedback gain K is designed (step S32). In this modification, the network structure defined in step S31 is α Then, we design the feedback gain K that satisfies the two LMI conditions in equations (22) and (26) above.
[0161] The control device in the prior art is premised on being able to observe all the state quantities of the neighboring agents. Therefore, the control input u is calculated based on the relative value between the state quantity of the own agent and the state quantity of the neighboring agent, and the feedback gain K designed in step S31. i By calculating this, it is possible to realize leader-follower control that is robust to changes in the network structure caused by the departure of one follower, even for conventional followers.
[0162] (Action, effect) As described above, the design method of the control device 20 according to this embodiment includes the steps of: defining a network structure that represents the connection relationship of information transmission between the leader L and the follower F such that the follower F acquires the state quantity of an adjacent agent that is one of the leader L or the follower F; and designing a feedback gain K that is a linear matrix inequality based on the defined network structure and that satisfies both a first linear matrix inequality (equation (22)) that constrains the closed-loop system CL including the feedback controller 203 to be stable, and a second linear matrix inequality (equation (26)) based on variables obtained from the first linear matrix inequality.
[0163] In this way, it is possible to design a control device 20 that can stably continue leader-follower control even if one follower F leaves the moving object group 1 and the network structure changes. As a result, it is possible to configure a robust moving object group 1 that can continue operation even if one follower F in the moving object group 1 fails.
[0164] Furthermore, when the control device 20 is configured to include the observer 202, the step of designing the feedback gain K involves designing a feedback gain that satisfies all of the first linear matrix inequality (equation (22)), the second linear matrix inequality (equation (26)), and the third linear matrix inequality (equation (21)) that constrains the magnitude of the feedback gain K to be equal to or less than the upper limit value γ, while gradually increasing the upper limit value γ while the real parts of the poles of the closed-loop system CL are all negative.
[0165] In this way, it is possible to design a control device 20 (FIG. 2) equipped with an observer 202 that can increase the feedback gain K within a range that ensures the stability of the entire system (all agents), and that can continue to perform leader-follower control stably even if the network structure changes.
[0166] <Other embodiments> Although one embodiment has been described in detail above with reference to the drawings, the specific configuration is not limited to the above, and various design changes are possible. That is, in other embodiments, the order of the above-mentioned processes may be changed as appropriate. Also, some of the processes may be executed in parallel.
[0167] <Control device hardware configuration> FIG. 16 is a schematic block diagram showing a hardware configuration of a control device of a moving object according to at least one embodiment.
[0168] The computer 900 comprises a processor 901 , a main memory device 902 , an auxiliary memory device 903 , and an interface 904 .
[0169] The above-mentioned control device 20 is implemented in a computer 900. The operations of the above-mentioned processing units are stored in the auxiliary storage device 903 in the form of a program. The processor 901 reads the program from the auxiliary storage device 903, loads it in the main storage device 902, and executes the above-mentioned processing in accordance with the program. The processor 901 also secures storage areas in the main storage device 902 corresponding to the above-mentioned storage units in accordance with the program.
[0170] The program may be for realizing a part of the functions to be performed by the computer 900. For example, the program may be for realizing the functions by combining with other programs already stored in the auxiliary storage device 903 or by combining with other programs implemented in other devices. In another embodiment, the computer 900 may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to or instead of the above configuration. Examples of PLDs include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), and FPGA (Field Programmable Gate Array). In this case, some or all of the functions to be realized by the processor 901 may be realized by the integrated circuit.
[0171] Examples of the auxiliary storage device 903 include a hard disk drive (HDD), a solid state drive (SSD), a magnetic disk, a magneto-optical disk, a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), and a semiconductor memory. The auxiliary storage device 903 may be an internal medium directly connected to the bus of the computer 900, or an external storage device 910 connected to the computer 900 via an interface 904 or a communication line. In addition, when the program is distributed to the computer 900 via a communication line, the computer 900 that receives the program may load the program into the main storage device 902 and execute the above-mentioned process. In at least one embodiment, the auxiliary storage device 903 is a non-transitory tangible storage medium.
[0172] The program may be for realizing part of the above-mentioned functions. Furthermore, the program may be a so-called differential file (differential program) that realizes the above-mentioned functions in combination with another program already stored in the auxiliary storage device 903.
[0173] <Additional Notes> The method for designing a control device and the method for manufacturing a moving body described in the above-described embodiments can be understood, for example, as follows.
[0174] (1) According to the first aspect, the method for designing a control device is a method for designing a control device 20 for a follower F in a group of moving objects 1 including one moving object that is a leader L and one or more following moving objects F that follow the leader L. The control device 20 for the follower F controls a relative position y Fi Based on the observer gains H, F, and P, the relative state value x Fi and an observer 202 that estimates the state quantity relative value ^x Fi Based on the feedback gain K, the control input u for the robot to follow the leader L is calculated. i The control device 20 is provided with a feedback controller 203 that outputs: ∞ The method includes a step of designing observer gains H, F, and P that satisfy both the linear matrix inequality that constrains the observer 202 and the feedback controller 203 designed for each follower F, and a step of designing a feedback gain K using the linear matrix inequality to improve the tracking control performance for the leader L while ensuring the stability of the entire closed-loop system CL including the observer 202 and feedback controller 203 designed for each follower F.
[0175] By doing this, even if it is not possible to observe all the state quantities of neighboring agents due to communication restrictions, for example, the relative position y FiFrom all the state quantity relative values x Fi It is possible to design an observer 202 capable of estimating the given values α and H ∞ By designing the observer gains H, F, and P that satisfy the constraints on the norm, even if the relative values of the control inputs to the neighboring agents are unknown, the influence of the unknown inputs to the observer 202 is sufficiently small to ensure that all state quantity relative values x Fi It is possible to estimate and suppress a decrease in the convergence speed of the estimate of the observer 202. In addition, by designing the feedback gain K using the linear matrix inequality as described above, it is possible to ensure the stability of the entire system (all agents) and improve the tracking performance of each follower F to the leader L.
[0176] (2) According to the second aspect, in the design method for the control device 20 according to the first aspect, the step of designing the observer gains H, F, P involves repeatedly designing the observer gains H, F, P while varying a predetermined value α that determines the convergence speed of the observer 202 within a specified range, and determining the predetermined value α so that the convergence speed is equal to or less than a preset speed.
[0177] In this way, it is possible to design appropriate observer gains H, F, and P that balance the convergence speed and estimation accuracy of the observer 202.
[0178] (3) According to a third aspect, the design method for the control device 20 according to the first or second aspect further includes a step of designing a feedback gain K that satisfies both a linear matrix inequality that constrains the closed-loop system CL to be stable while gradually increasing an upper limit value γ of the magnitude of the feedback gain K, and a linear matrix inequality that constrains the magnitude of the feedback gain K to be equal to or less than the upper limit value γ, while the real parts of the poles of the closed-loop system CL are all negative.
[0179] In this way, the feedback gain K can be increased within a range in which the stability of the entire system (all agents) is guaranteed. This makes it possible to more reliably ensure both the guarantee of the stability of the entire system and the improvement of the tracking performance of each follower F to the leader L.
[0180] (4) According to the fourth aspect, the method for designing the control device 20 according to any one of the first to third aspects is to calculate the relative position y Fi The method further includes adding an integrator 204 that outputs a correction value for reducing a steady-state deviation of the aircraft with respect to the target position based on the above.
[0181] In this way, it is possible to design the control device 20 that can reduce the steady-state deviation and accurately track the target position that changes from moment to moment as the reader L moves.
[0182] (5) According to a fifth aspect, in a design method for a control device 20 relating to any one of the first to fourth aspects, the step of designing observer gains H, F, and P designs multiple observer gains H, F, and P corresponding to multiple specified speeds.
[0183] In this way, it is possible to design an observer 202 capable of estimating state quantity relative values with high accuracy according to the speed of the aircraft 10, whose motion characteristics change according to the speed.
[0184] (6) According to a sixth aspect, in the design method for the control device 20 according to the third aspect, the step of designing the feedback gain K designs a plurality of feedback gains K corresponding to a plurality of specified speeds.
[0185] In this way, it is possible to design a control device 20 (feedback controller 203) that can appropriately control the aircraft 10 in accordance with its own speed, for an aircraft 10 whose motion characteristics change depending on its speed.
[0186] (7) According to the seventh aspect, a method for manufacturing a moving body is a method for manufacturing a moving body that is a follower F in a moving body group 1 consisting of one moving body that is a leader L and one or more moving bodies that are followers F that follow the leader L, and includes a step of manufacturing the control device 20 based on a gain designed by a design method for the control device 20 relating to any one of the first to sixth aspects. [Explanation of symbols]
[0187] 1. Mobile Group 10 Aircraft (Controlled Object) 20 Control device 201 Acquisition Department 202 Observer 203 Feedback Controller 204 Integrator 205 Adder CL Closed loop system L Leader F Follower H,F,P observer gain K Feedback Gain
Claims
1. A method for designing a control device for a follower in a group of moving objects including one moving object that is a leader and one or more moving objects that are followers of the leader, comprising the steps of: The control device of the follower includes: an observer that estimates a relative value of a state quantity between the leader and the adjacent agent, which is at least one of the followers, based on a relative position between the leader and the adjacent agent and the agent itself and an observer gain; a feedback controller that outputs a control input for the own device to follow up the leader based on the estimated value of the state quantity relative value and a feedback gain; Equipped with The design method of the control device includes a linear matrix inequality that constrains all real parts of the poles of the observer to be a predetermined value away from the imaginary axis on the negative side, and ∞ and designing a feedback gain for improving the tracking control performance to the leader by using the linear matrix inequality while ensuring the stability of the entire closed loop system including the observer and the feedback controller designed for each of the followers. How to design a control device.
2. The step of designing the observer gain includes repeatedly designing the observer gain while changing the predetermined value that determines the convergence speed of the observer within a specified range, and determining the predetermined value so that the convergence speed is equal to or less than a preset speed. A method for designing a control device according to claim 1.
3. The method further includes a step of designing a feedback gain that satisfies both a linear matrix inequality that constrains the closed loop system to be stable while gradually increasing an upper limit value of the magnitude of the feedback gain, and a linear matrix inequality that constrains the magnitude of the feedback gain to be equal to or less than the upper limit value while the real parts of the poles of the closed loop system are all negative. A method for designing a control device according to claim 1.
4. The method further includes a step of adding an integrator that outputs a correction value for reducing a steady-state deviation of the target position of the own aircraft based on a relative position of the own aircraft with respect to the target position of the adjacent agent. The method for designing a control device according to claim 3.
5. The step of designing the observer gain includes designing a plurality of observer gains corresponding to a plurality of specified speeds. A method for designing a control device according to claim 1.
6. The step of designing the feedback gain includes designing a plurality of feedback gains corresponding to a plurality of specified speeds. The method for designing a control device according to claim 3.
7. A method for manufacturing a follower moving body in a group of moving bodies including one leader moving body and one or more follower moving bodies following the leader, the method comprising the steps of: A step of manufacturing the control device based on a gain designed by the control device design method according to any one of claims 1 to 6. A method for manufacturing a moving object.
Citation Information
Patent Citations
Control device, moving body, and distribution control program for moving body
JP2019074918A
