Adaptive trajectory prediction compound control method and device for visual servo system

Abstract

The invention discloses a self-adaptive trajectory prediction compound control method and device for a visual servo system, and belongs to the field of target tracking control of a photoelectric tracking system. Comprising the following steps: constructing a transfer function of a photoelectric tracking system based on a kinetic equation of a reflector and a kinetic equation of a voice coil motor in combination with a motion relation of the fast reflector in a deflection process, and obtaining a mathematical model of the photoelectric tracking system through the transfer function; identifying system input delay, designing an optimal position controller based on a photoelectric tracking system mathematical model, and constructing a single feedback controller; designing an adaptive trajectory prediction feedforward controller based on the established single feedback controller and the photoelectric tracking system mathematical model; and constructing adaptive trajectory prediction compound control based on a single feedback controller and an adaptive trajectory prediction feedforward controller. According to the invention, composite control is constructed based on position feedback and feedforward, and the tracking performance of the photoelectric tracking system is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of adaptive trajectory prediction compound control method and device for visual servo system, belong to photoelectric tracking system control technical field. BACKGROUND

[0002] System tracking accuracy is an important indicator to measure photoelectric tracking performance, which is directly reflected in the size of line of sight (LOS) error. However, because the photoelectric tracking system based on vision needs strong exposure time to obtain high-quality images by CCD, the system will contain non-negligible delay, which limits the control bandwidth. Since the delay will limit the gain of the controller, it is difficult to obtain sufficient pointing accuracy only by relying on basic feedback control. Therefore, a control strategy specially designed to overcome the impact of delay must be used to overcome the diffraction limit of the optical system as much as possible.

[0003] In order to balance the sharp contradiction between the gain and phase angle margin of the controller with time delay of the system, researchers first consider improving the design of the feedback controller. By adding an integral link, the system is converted into a double-type conditionally stable system using the PID-I method. A dynamic high-type control method of approximate adaptive controller is proposed, which uses different types of controllers according to the size of the error. In order to balance the relationship between dynamic performance and stability margin more finely, fractional order control and cascade compensation control are introduced in the visual tracking system. However, although the above methods can improve the accuracy of the system, they essentially sacrifice part of the stability margin of the system to exchange for accuracy.

[0004] In the technical solution to improve the pointing accuracy of the system, in addition to optimizing the traditional feedback controller, introducing a feedforward control branch is recognized as a key breakthrough. However, there are two core difficulties in practical application: on the one hand, the image sensor can only provide real-time line of sight deviation data based on feedback, while the feedforward control needs to obtain the motion trajectory of the target, which is often difficult to measure directly; on the other hand, due to the lag effect of target trajectory generation, combined with the high uncertainty of the motion pattern of the target (such as maneuvering turning, variable speed motion, etc.), it is difficult to build an accurate trajectory prediction model and to effectively integrate the prediction results into the feedforward compensation link. SUMMARY

[0005] The present application aims to overcome the deficiencies in the prior art and provide an adaptive trajectory prediction compound control method and device for visual servo system. The method is a composite structure of feedback and feedforward, which can improve the tracking performance of the system, and an interactive multiple model Kalman predictor is added to adapt to the change of target state, reasonably estimate the moving target and be used for feedforward. To achieve the above purpose, the present application provides the following technical solutions:

[0006] An adaptive trajectory prediction compound control method for a visual servo system, the method steps are as follows:

[0007] Step (1): based on the dynamic equation of the mirror and the dynamic equation of the voice coil motor, and combined with the motion relationship of the fast mirror in the deflection process, the transfer function of the photoelectric tracking system is constructed, and the mathematical model of the photoelectric tracking system is obtained through the transfer function;

[0008] Step (2): identify the system input delay , design an optimal position controller based on the mathematical model of the photoelectric tracking system, and construct a single feedback controller;

[0009] Step (3): design an adaptive trajectory prediction feedforward controller based on the established single feedback controller and the mathematical model of the photoelectric tracking system;

[0010] Step (4): adaptive trajectory prediction compound control based on the single feedback controller and the adaptive trajectory prediction feedforward controller.

[0011] An adaptive trajectory prediction compound control device for a visual servo system, comprising:

[0012] Mathematical model construction unit: based on the dynamic equation of the mirror and the dynamic equation of the voice coil motor, and combined with the motion relationship of the fast mirror in the deflection process, the transfer function of the photoelectric tracking system is constructed, and the mathematical model of the photoelectric tracking system is obtained through the transfer function;

[0013] Single feedback controller construction unit: identify the system input delay , design an optimal position controller based on the mathematical model of the photoelectric tracking system, and construct a single feedback controller;

[0014] Feedforward controller construction unit: design an adaptive trajectory prediction feedforward controller based on the established single feedback controller and the mathematical model of the photoelectric tracking system;

[0015] Compound control unit: adaptive trajectory prediction compound control based on the single feedback controller and the adaptive trajectory prediction feedforward controller.

[0016] An electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0017] A computer-readable storage medium having executable instructions stored thereon, the instructions being executed by a processor to cause the processor to implement the method.

[0018] Compared with the prior art, the present application has the following beneficial effects:

[0019] (1)The present application uses an interactive multi-model Kalman predictor to fuse different motion models, so as to achieve adaptive prediction and delay compensation of the current position, and to realize optimization of control performance in cooperation with feedforward control.

[0020] (2)The present application has clear ideas, simple steps, and is easy to implement in engineering. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings, which form a part of the specification, illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0022] The present application can be more clearly understood and appreciated from the following detailed description, taken in conjunction with the following drawings, in which:

[0023] Figure 1 is a flow chart of a photoelectric tracking control method of the present application;

[0024] Figure 2 is a control block diagram of adaptive trajectory prediction compound control of the present application;

[0025] Figure 3 is a comparison chart of error suppression capabilities of different control methods in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the objectives, technical solutions and advantages of the present application clearer and more apparent, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other. In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions.

[0027] Figure 1 is a flow chart of an adaptive trajectory prediction compound control method for a visual servo system in embodiment 1 of the present application. This flow chart only shows the logical order of the method described in this embodiment, and the steps shown or described can be completed in an order different from that shown in other possible embodiments of the present application as long as they do not conflict with each other. Figure 1

[0028] Figure 2 is a control block diagram of adaptive model assisted compound feedforward control of the present application. It includes a single position feedback module and an adaptive compensation module. Among them represents the transfer function mathematical model of the controlled object, C is a position controller, Q is an interactive multi-model Kalman predictor, F is a feedforward controller, is a system delay, ​for the added artificial delay, is an equivalent model of the controlled object G, , r is the synthesized target trajectory with delay.

[0029] Figure 1 The flowchart shown, comprising the following steps:

[0030] Step (1): Based on the dynamic equation of the fast mirror and the dynamic equation of the voice coil motor, and combined with the motion relationship of the fast mirror in the deflection process, the transfer function of the photoelectric tracking system is constructed, and the mathematical model of the photoelectric tracking system is obtained through the transfer function;

[0031] Step (2): The sampling rate of the CCD sensor is low, and the image processing unit needs a long time to extract information, resulting in a large delay in the extracted target miss distance. The system input delay is obtained by frequency response identification , based on the mathematical model of the photoelectric tracking system, an optimal position controller is designed, and a single position feedback controller is constructed;

[0032] Step (3): Based on the established single position feedback controller and the mathematical model of the photoelectric tracking system, an adaptive trajectory prediction feedforward controller is designed;

[0033] Step (4): Based on the single position feedback controller and the adaptive trajectory prediction feedforward controller, an adaptive trajectory prediction composite control is carried out.

[0034] Further, the fast mirror is a key component of the photoelectric tracking system, which is used to accurately and quickly adjust the direction of the reflected light beam. The dynamic equation of the fast mirror can be expressed as:

[0035] ;

[0036] wherein, is the torque applied in the X-axis direction, is the moment of inertia of the mirror and the flexible support rotating part, is the deflection angle of the mirror, is the differential of the deflection angle, i.e. the deflection speed, is the differential of the deflection speed, i.e. the deflection acceleration, is the X-axis torsional stiffness of the flexible support, is the equivalent damping coefficient of the flexible support and the voice coil motor, is the moving mass of the voice coil motor, is the distance from the force action point of the voice coil motor to the rotating shaft of the mirror.

[0037] Since the fast mirror usually uses a voice coil motor as a driver, the voice coil motor is further modeled to obtain the dynamic equation of the voice coil motor, which can be expressed as:

[0038] ;

[0039] wherein, is the axial displacement of the voice coil motor, is the velocity when it is displaced, is the acceleration when it is displaced, is the working voltage of the voice coil motor, is the working current of the voice coil motor, is is the derivative, is the inductance of the motor coil, is the resistance of the voice coil motor, is the motor torque coefficient, is the motor back electromotive force coefficient, is other equivalent damping except the motor back electromotive force, is the elastic force suffered by the voice coil motor, is the elastic force coefficient. And the displacement of the deflection angle is generally small, which can be approximated as , and the motion relationship of the fast steering mirror in the deflection process is:

[0040] ;

[0041] The transfer function of the photoelectric tracking system is obtained based on the dynamic equation of the fast steering mirror and the dynamic equation of the voice coil motor, and combined with the motion relationship of the fast steering mirror in the deflection process:

[0042] ;

[0043] wherein, represents the complex frequency variable in the Laplace transform domain, is the transfer function of the photoelectric tracking system, is the position deflection of the photoelectric tracking system, is the given input voltage of the photoelectric tracking system.

[0044] The mechanical part of the photoelectric tracking system composed of a typical oscillation link and a typical first-order inertia link is represented as follows:

[0045] ;

[0046] wherein, represents the mathematical model of the photoelectric tracking system, represents the natural oscillation frequency, represents the damping coefficient, represents the Laplace operator, represents the inertia link coefficient;

[0047] In combination step, further, the single feedback control based on the mathematical model of the photoelectric tracking system comprises a transfer function mathematical model of the photoelectric tracking system and system delay , a zero-pole cancellation method is used to design the position controller. Engineering requirements are that the system has a magnitude margin greater than 6db and a phase angle margin greater than , so we can get:

[0048] ;

[0049] Through calculation, the maximum value of the position controller gain is , so the position controller is designed as:

[0050] ;

[0051] The fitting model of the actual object G is represented. An integral link negative feedback is added to the position controller to form a position closed loop; the error transfer function of the formed position closed loop is:

[0052] ;

[0053] In combination step, further, the adaptive trajectory prediction feedforward control is based on the estimation of the position controlled object and is designed from the trajectory synthesized from the designed position controller, a target miss and a model output, wherein an interactive multi-model Kalman predictor for trajectory prediction and a feedforward controller for improving the photoelectric tracking precision are included; the error transfer function of the adaptive trajectory prediction compound control structure is:

[0054] ;

[0055] Comparing the error transfer function and the disturbance transfer function of the single feedback control and the present example, it can be seen that the error suppression ability of the system is enhanced, and the control precision is improved. If F=1 and Q= , it means that the predictor can completely compensate for the delay, and the system precision will be greatly improved.

[0056] The feedforward controller is designed as a first-order low-pass link . The interactive multi-model Kalman predictor describes different target states by incorporating different models, that is, by incorporating different state equations; the prediction results of multiple models are obtained by parallel computing the Kalman predictions of multiple state models; the model selection is completed by updating the probabilities of each model through the likelihood value; wherein the steps include:

[0057] 1) input interaction:

[0058] The transition probability of the model to j is The initial state transition matrix selected by prior knowledge and experience is:

[0059] ;

[0060] is the probability of model j at time k-1, is the estimate of the target position at this time by the predictor, represent the corresponding state covariance matrix, the interaction input of the r Kalman predictors at time k:

[0061] ;

[0062] 2) Calculate the current time state model Kalman prediction based on the mixed state estimate:

[0063] State prediction:

[0064] ;

[0065] is the state space equation of the target;

[0066] Prediction error covariance:

[0067] ;

[0068] is the process noise of the system;

[0069] Residual:

[0070] ;

[0071] , are the measurement matrix and observation matrix of the predictor, respectively;

[0072] Residual covariance:

[0073] ;

[0074] is the measurement noise of the system;

[0075] Kalman gain:

[0076] ;

[0077] State update:

[0078] ;

[0079] Prediction error covariance update:

[0080] ;

[0081] The extrapolation step is:

[0082] ;

[0083] 3) Calculate the model likelihood value:

[0084] The likelihood value is calculated by the residual and residual covariance to determine the goodness of the model:

[0085] ;

[0086] 4) Model probability update:

[0087] ;

[0088] ;

[0089] 5) Fusion state estimation output:

[0090] ;

[0091] 6) Fusion extrapolation output:

[0092] ;

[0093] Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

[0094] Embodiment 2:

[0095] This embodiment takes a certain photoelectric tracking system as an example to verify the adaptive trajectory prediction compound control method for visual servo system provided in embodiment 1.

[0096] 1) Obtain the estimation of the controlled object model by frequency domain response fitting, because the fitting accuracy is high, so in the design process, the estimation of the controlled object model obtained by fitting can be regarded as the real object. The sampling rate of the CCD image sensor is 50Hz, and the time delay obtained by frequency domain fitting is 0.02s. The system speed transfer function measured by the frequency response instrument is as follows:

[0097] ;

[0098] 2) The position controller designed by the transfer function of the controlled object is:

[0099] ;

[0100] ;

[0101] 3) The feedforward controller is designed as a first-order low-pass filter:

[0102] ;

[0103] 4) Design a uniform acceleration model and a current statistical model, and use a dual-model interactive extrapolation Kalman predictor to handle the target's non-maneuvering and maneuvering processes, respectively. The state equations are as follows:

[0104] ;

[0105] Let be the target's maneuver frequency, be the derivative of the motion time constant, and T be the sampling period.

[0106] Under the same experimental conditions, the error suppression capabilities of visual servoing systems with single feedback and those with adaptive trajectory prediction composite control were compared, such as... Figure 3 As shown, this plot is the logarithm of the ratio of error to input amplitude under different input frequencies. It can be seen that the adaptive trajectory prediction composite control visual servo tracking system designed using this method is stable and maintains good tracking accuracy under target motion at any frequency. Compared with a simple position closed-loop system, its error suppression capability is greatly improved. Although the error suppression capability at mid-frequency is somewhat weakened, this is acceptable because the motion information of the tracked target is usually concentrated at low frequencies, making low-frequency error suppression capability relatively more important.

Claims

1. An adaptive trajectory prediction composite control method for a visual servoing system, characterized in that, The method includes the following steps: Step (1): Based on the dynamic equations of the reflector and the voice coil motor, and combined with the motion relationship of the fast reflector during the deflection process, construct the transfer function of the photoelectric tracking system, and obtain the mathematical model of the photoelectric tracking system through the transfer function; Step (2): Identify system input delay The optimal position controller is designed based on the mathematical model of the photoelectric tracking system, and a single feedback controller is constructed. Step (3): Design an adaptive trajectory prediction feedforward controller based on the established single feedback controller and the mathematical model of the photoelectric tracking system; Step (4): Perform adaptive trajectory prediction composite control based on a single feedback controller and an adaptive trajectory prediction feedforward controller.

2. The adaptive trajectory prediction composite control method for a visual servoing system according to claim 1, characterized in that, The transfer function of the photoelectric tracking system is constructed, and the mathematical model of the photoelectric tracking system is obtained through the transfer function. The specific process is as follows: The photoelectric tracking system includes a reflector, and the dynamic equation of the reflector is expressed as: ； in, The torque applied in the X-axis direction. It is the moment of inertia of the rotating parts of the reflector and the flexible support. The angle of deflection of the mirror. It is the differential of the deflection angle, i.e., the deflection velocity. It is the differential of the deflection velocity, i.e., the deflection acceleration. To provide flexible support for the torsional stiffness of the X-axis. This represents the equivalent damping coefficient of the flexible support and the voice coil motor. For the mass of the voice coil motor mover. This is the distance from the point of application of the voice coil motor force to the axis of rotation of the reflecting mirror; The reflector uses a voice coil motor as its driver. Further modeling of the voice coil motor yields its dynamic equations, which are expressed as follows: ； in, It is the axial displacement of the voice coil motor. It is the velocity during its displacement. It is the acceleration during its displacement. This is the operating voltage of the voice coil motor. This is the operating current of the voice coil motor. yes The derivative, It is the inductance of the motor coil. It is the resistance of the voice coil motor. This is the motor torque coefficient. This is the back electromotive force coefficient of the motor. Other equivalent damping besides the back electromotive force of the motor, The elastic force acting on the voice coil motor. It is the elastic coefficient, which is the deflection angle of the displacement when the mirror is working. It is believed to be The motion relationship of the mirror during the deflection process: ； Based on the dynamic equations of the reflector and the voice coil motor, and combined with the motion relationship of the reflector during the deflection process, the transfer function of the photoelectric tracking system is obtained as follows: ； in, Represents the complex frequency variable in the Laplace transform domain. It is the transfer function of the photoelectric tracking system. It is the position deflection of the photoelectric tracking system. It is the given input voltage of the photoelectric tracking system; The mathematical model of the photoelectric tracking system, consisting of an oscillating element and a first-order inertial element, is expressed as follows: ； in, Mathematical model representing photoelectric tracking system Indicates the natural oscillation frequency. Indicates the damping coefficient. Represents the Laplace operator. This represents the coefficient of inertial elements.

3. The adaptive trajectory prediction composite control method for a visual servoing system according to claim 1, characterized in that, A single feedback controller is established based on the mathematical model of the photoelectric tracking system, including the calculation of the transfer function of the photoelectric tracking system and the system delay. The designed position controller system has an amplitude margin greater than 6dB and greater than The phase margin is thus obtained as follows: ； The maximum value of the position controller gain is calculated. Therefore, the position controller was designed as follows: ； Let G be the fitted model of the actual object G; add an integral negative feedback loop to the position controller to form a position closed loop, and the position closed loop error transfer function of the single feedback controller is: 。 4. The adaptive trajectory prediction composite control method for a visual servoing system according to claim 1, characterized in that, The adaptive trajectory prediction feedforward controller includes an interactive multi-model Kalman predictor for trajectory prediction and a feedforward controller for improving photoelectric tracking accuracy; the error transfer function of the adaptive trajectory prediction feedforward controller is: ； Where Q is the interactive multi-model Kalman predictor, F is the feedforward controller, and C is the position controller designed earlier. For system latency, Added artificial delay, .

5. The adaptive trajectory prediction composite control method for a visual servoing system according to claim 4, characterized in that, Design the feedforward controller as a first-order low-pass circuit .

6. The adaptive trajectory prediction composite control method for a visual servoing system according to claim 4, characterized in that, The interactive multi-model Kalman predictor describes different target states using different target state equations. The interactive multi-model Kalman predictor obtains the prediction results of multiple state models by computing the Kalman predictions of multiple state models in parallel. Then, it completes the adaptive model selection by updating the probability of each state model through the likelihood value.

7. The adaptive trajectory prediction composite control method for a visual servoing system according to claim 6, characterized in that, The working steps of the interactive multi-model Kalman predictor include: 1) Input interaction: State Model The transition probability to j is The initial state transition matrix selected based on prior knowledge and experience is: ； Let be the probability of state model j at time k-1. It is the predictor's estimate of the target's position at that moment. Representing the corresponding state covariance matrix, the interactive inputs of the r Kalman predictors at time k: ； 2) Calculate the Kalman prediction of each state model at the current time based on mixed state estimation: State prediction: ; in, The state-space equation for the objective; Prediction error covariance: ; in, This refers to the process noise of the system. Residual: ; in, , These are the measurement matrix and observation matrix of the predictor, respectively; Residual covariance: ; in, The measurement noise of the system; Kalman gain: ; Status Update: ; Prediction error covariance update: ; The extrapolation step is: ; 3) Calculate the model likelihood value; determine the goodness of the model by obtaining the likelihood value through the residuals and residual covariance. ； 4) Model probability update: ； ； 5) Fusion state estimation output: ； 6) Fusion extrapolation output: 。 8. An adaptive trajectory prediction composite control device for a visual servoing system, characterized in that, include: Mathematical model construction unit: Based on the dynamic equations of the reflector and the voice coil motor, and combined with the motion relationship of the fast reflector during the deflection process, the transfer function of the photoelectric tracking system is constructed, and the mathematical model of the photoelectric tracking system is obtained through the transfer function; Single feedback controller building block: Identifying system input delay The optimal position controller is designed based on the mathematical model of the photoelectric tracking system, and a single feedback controller is constructed. Feedforward controller building block: Based on the established single feedback controller and the mathematical model of the photoelectric tracking system, an adaptive trajectory prediction feedforward controller is designed; Composite control unit: Adaptive trajectory prediction composite control based on a single feedback controller and an adaptive trajectory prediction feedforward controller.

9. An electronic device, characterized in that, include: One or more processors; A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, cause the processor to perform the method described in any one of claims 1 to 7.

Images