TWI313536B

TWI313536B - Control method for on-line tuning vector of parameters of controller based on simultaneous perturbation stochastic approximation (spsa)

Info

Publication number: TWI313536B
Application number: TW095117079A
Authority: TW
Inventors: Ying-Yi Hong; Wen-Ming Luo
Original assignee: Univ Chung Yuan Christian
Priority date: 2006-05-15
Filing date: 2006-05-15
Publication date: 2009-08-11
Also published as: TW200743296A

Description

1313536 _吻調修正替換頁 98-5-20 九、發明說明：【發明所屬之技術領域】本發明是有關於一種控制一受控裝置的方法，且特別是有關於一種基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法。【先前技術】1313536 _ kiss adjustment correction page 98-5-20 IX. Description of the invention: [Technical field of the invention] The present invention relates to a method for controlling a controlled device, and more particularly to a method based on simultaneous perturbation random approximation The control method of the controller parameter vector line adjustment. [Prior Art]

比例積分微分控制器（pr〇p〇rti〇nal_Integral_Derivative Controller，以下簡稱；pID)，是廣泛應用於工業界的一個基礎的控制益’功用主要是根據偏差值依比例輸出修正。p 指的是一項常用的線性控制策略以調節初始響應時間；工表示積分(Integral)控制，其功用主要是根據偏差值之積分輪出修正以消除穩態偏差；D表示微分(Derivative)控制，其功用主要在調節阻尼與衝量(oversh〇〇t)。爲使控制效能達到目標，在製程變化等諸多情形下即必須調整piD三個參數’此動作即爲控制器參數向量調節。以下為比例積分微分器的轉移函數：The proportional integral derivative controller (pr〇p〇rti〇nal_Integral_Derivative Controller, hereinafter referred to as pID) is a basic control benefit widely used in the industry. The function is mainly based on the deviation value proportional output correction. p refers to a commonly used linear control strategy to adjust the initial response time; the worker represents integral (Integral) control, its function is mainly based on the integral of the deviation value to correct the correction to eliminate the steady-state deviation; D represents the derivative control Its function is mainly to adjust the damping and impulse (oversh〇〇t). In order to achieve the control performance, the three parameters of piD must be adjusted in many cases such as process changes. This action is the controller parameter vector adjustment. The following is the transfer function of the proportional integral differentiator:

ι-ζ~ 其中參數，Ki為積分她，Kd為微分失數。目前工業界的馬達控制架構1都是使用piD的方法 =控制受控裝置，例如馬達，尤其是㈣較精麵祠服馬達或步進馬_錄(Positioning)，所謂蚊蚊三位置運動到另-點靜止位置之動作。其基本運動 (^物體纏麟定位置，朗在所定精度_止遷動。 1313536 )’Κ % 98-5-20 (2) 物體啟動到某速度’直到在所定精度内等速移動的運動。 (3) 物體密切追隨空間、時間變化之目標的運動。而定位精度的優劣。可由準確性（Accuracy )與精峰性（precisi〇n) 這兩項性能規格看出。上述的PID控制器必須要根據馬達的不同，選定上述的比例參數Kp、積分參數K i以及微分參數K d以使得馬達控制精準。在實際的產品上，會根據不同種類的馬達給定鬱不同的上述比例參數Kp、積分參數&以及微分參數Kd。然而，即使是相同種類的馬達，亦有可能在生產的過程中產生誤差或應用的場合不同，即使根據馬達的種類給定上述三個參數’亦無法達到每一個產品皆能準確的受到控制。【發明内容】 • 本發明的目的就是在提供一種基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，用以改善受控裝置的動態響應’實現線上調整控制器參數向量，提高整 • 體控制系統的強健性，進一步達到準確地控制不同的受控裝置。 fI月提出種基於同時擾動隨機近似法之控制器參數向里線上調整的控制方法，此方法至少包括下列步驟：，據又控置的種類，提供控制器參數向量Θ ;定義第— 衰咸值A帛―衰減值c以及隨機擾動向量△ ; 隨機取擾動向里△的符號；b.作0+°χΔ運算，得到正擾動才工制.„减向里0 ;e.以正擾動控制器參數向量控制 1313536 98-5-20 馬達，得到第一位置誤差值γ+ β 擾動控制器參數向量Μ△運算，知到負得到梯度估測值g(0)，其中梯度估· _=ι^χΔ，其中㈣實數；“運控制馬達，得到第二位置，^值負，制器參數向量Θ _ --、… 夏祆至值γ ;f·作梯度估測運算，算如下： g. 作θ—_運算，將此運算結果取代控制 h. 重複上述步驟a〜匕；器參數向量β ; 而此控制器的轉移函數為：确〜占略外Κρ代表—比例參數，Ι-ζ~ where the parameter, Ki is the integral of her, Kd is the differential number. At present, the motor control architecture 1 of the industry is a method using piD = control of controlled devices, such as motors, especially (four) fine-faced squat motor or stepping horse positioning, the so-called mosquito three-position movement to another - The action of the point of rest position. The basic motion (^ object entangles the position, sings at the specified precision _ stop moving. 1313536 ) 'Κ % 98-5-20 (2) The object starts to a certain speed ' until the movement moves at a constant speed within the specified accuracy. (3) The movement of objects closely following the target of space and time. And the accuracy of positioning accuracy. It can be seen from the two performance specifications of Accuracy and Precisi〇. The above PID controller must select the above-mentioned proportional parameter Kp, integral parameter K i and differential parameter K d according to the motor to make the motor control accurate. In the actual product, different proportional parameters Kp, integral parameters & and differential parameters Kd are given according to different types of motors. However, even the same type of motor may have errors or applications in the production process, and even if the above three parameters are given according to the type of the motor, it is impossible to accurately control each product. SUMMARY OF THE INVENTION The object of the present invention is to provide a control method for adjusting a controller parameter vector line based on a simultaneous perturbation stochastic approximation method, which is used to improve the dynamic response of a controlled device. • The robustness of the body control system further enables accurate control of different controlled devices. In fI month, a control method based on the simultaneous disturbance random approximation controller parameter inward adjustment is proposed. The method includes at least the following steps: According to the type of control, the controller parameter vector is provided; the first-damper value is defined. A帛—Attenuation value c and random perturbation vector △; Randomly take the sign of △ in the direction of motion; b. Perform 0+°χΔ operation to get the positive disturbance to work. „Retraction 0; e. Positive disturbance controller The parameter vector controls 1313536 98-5-20 motor, and obtains the first position error value γ+ β to perturb the controller parameter vector Μ△ operation, and knows that the negative gradient estimation value g(0) is obtained, where the gradient estimation _=ι^ χΔ, where (4) the real number; “Operation control motor, get the second position, ^ value is negative, the controller parameter vector Θ _ --, ... Xia 祆 to the value γ ; f · for the gradient estimation operation, as follows: g. θ - _ operation, this operation result replaces the control h. Repeat the above steps a ~ 匕; the device parameter vector β; and the transfer function of this controller is: YES ~ account for Κ ρ represents - proportional parameter,

Ki代表積分參數，Kd代表士 1 ±1 ±1 微分參數；隨機擾動向量△表示為：，且隨機選取隨機擾動向量 Q的付唬包括隨機選取隨機擾動向量△中的元素之正負符號。、、依照本，的較佳實施例所述之基於同時擾似法之控制斋參數向量線上調整的控制方法，、思為近參數向量為： '’上述控制器 —K: Θ: 似法=====時擾動隨機近上述步, h 1313536 州月、，正替約丨 98-5-20 -------jKi represents the integral parameter, Kd represents the ±1 ±1 ±1 differential parameter; the random perturbation vector △ is expressed as: , and the random perturbation vector Q is randomly selected to include the positive and negative signs of the elements in the random perturbation vector Δ. According to the preferred embodiment of the present invention, the method for controlling the adjustment of the parameter vector line based on the simultaneous disturbance method is as follows: the parameter vector is: ''the above controller-K: Θ: like= ==== When the disturbance is random near the above steps, h 1313536 state month, positive replacement about 丨98-5-20 -------j

之前更包括下列步驟：提供旗標值k;根據旗標值k，將第一衰減值A作第一衰減運算後，取代第一衰減值；根據旗標值k’將第二衰減值c作第二衰減運算後，取代第二衰減值；將旗標值k累加預設值作為旗標值k。在更進—步的貝化例中’第一衰減運算為：Α=α1/(&2+1ζ：)Λη，其中al、 a2、η為實數，且ai為0.775、^為〇.602。在另一實施例中，第一衰減運算為：al/(a2+k+l)、，其中al、a2、η為實數’且al為0.775、ng 0.602。在又—實施例中，第二衷減運算為：A=al/(k)、，其中al、n為實數，且al為〇 775、 η 為 0.602 〇依照本發明的較佳實施例所述之基於同時擾似法之控制器參數向量線上調整的控制方法，上述 = 減運算為：⑽释，其中cl、r為實數，且為^ 二為謹。在另一實施例中，第二The method further includes the following steps: providing a flag value k; according to the flag value k, the first attenuation value A is used as the first attenuation operation, replacing the first attenuation value; and the second attenuation value c is determined according to the flag value k' After the second attenuation operation, the second attenuation value is replaced; the flag value k is added to the preset value as the flag value k. In the further step-by-step example, the first attenuation operation is: Α = α1/(& 2+1ζ:) Λη, where al, a2, η are real numbers, and ai is 0.775, ^ is 〇.602 . In another embodiment, the first decay operation is: al / (a2 + k + l), where al, a2, η are real ' and al is 0.775, ng 0.602. In still another embodiment, the second subtraction operation is: A = al / (k), where al, n are real numbers, and a is 〇 775, η is 0.602 〇 in accordance with a preferred embodiment of the present invention. Based on the control method of the vector parameter adjustment of the controller parameter based on the simultaneous disturbance method, the above = subtraction operation is: (10), where cl and r are real numbers, and ^2 is a singularity. In another embodiment, the second

C=cl/(k+1卜其中心為實數，切為本發明因採用類似擾動隨機近似的清 to)直錢變控制ϋ中預設的控 ^^，除了可以提高控制系統的強健性，改怂二，’因此應外’進-步可相準確地控制不同的受控裝置。 w為讓本:ί:,上Ϊ和其他目的、特徵;優點能更明顯明如下。 °所關式，作詳細說【實施方式】由於上述以馬達作為受控裝置的習知技術中，即使同 7 I ；1313536 98-5-20 種類的馬達(受控裝置），不同的單體亦有可能會有相對的誤差’若給定相同的參數，將會造成馬達定位控制的誤差，因此本發明提& -餘㈣喊動隨機近似法之控制突參數向量線上調整的控制方法，用以改善受控裳置的動:變 srrt上調整控制器參數向量，提高控制系統二健性’進一步達到準確的控制不同的受控裝置。以下.將舉 -種同樣以馬達作為受控裝置的實施例，並明本發明。圚不况似法明實施例之一種基於同時擾動隨機近似法之控—參數向量線上調㈣ 1 會示為根據圖1實施例的方法所實施的硬碑:二;二2 達碟音圈馬達驅動器26。料匕24數位類比轉換器25、馬須說明的是，上述圖2實施例雖為舉例，示本發明可㈣實施之特^^圈^達作而本發明所提到的硬碟音圈馬達控制，僅是1中1=。.因此，非用來限制本發明所應用的領域。 —種貫施例，請參考圖1及圖2，首弈，拼械Λ 供控制器參數向以（步㈣㈣如置的種類，指為音圈馬達21，便預先提供與音圈馬；2=::馬達參數向量义给控制器24，在此 f 21相闕的控制器 __作為例子’娜麵C=cl/(k+1, whose center is a real number, which is a clear control of the invention using a similar perturbation random approximation). In addition to improving the robustness of the control system, In the second place, the 'and therefore' external steps can accurately control different controlled devices. w is for this: ί:, captain and other purposes, features; advantages can be more obvious as follows. °Closed, detailed description [Embodiment] Due to the above-mentioned conventional technology using a motor as a controlled device, even if it is the same type of motor (controlled device) as the 7 I; 1313536 98-5-20, different monomers There may also be a relative error 'if the same parameters are given, it will cause the error of the motor positioning control. Therefore, the present invention proposes a control method for adjusting the vector of the control parameters on the vector of the random approximation method. To improve the movement of controlled skirts: change the controller parameter vector on srrt and improve the control system's two robustness' to further achieve accurate control of different controlled devices. Hereinafter, an embodiment in which a motor is also used as a controlled device will be described, and the present invention will be clarified.圚圚似似法法法法法法法法法法法法法法法法参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数参数Driver 26. The 匕24 digital analog converter 25, the horse must be described, the above embodiment of Fig. 2 is an example, and the hard disk voice coil motor mentioned in the present invention can be implemented by the invention. Control, only 1 in 1 =. Therefore, it is not intended to limit the field in which the invention is applied. - For a variety of examples, please refer to Figure 1 and Figure 2, the first game, the machine tool for the controller parameters to (step (4) (four) such as the type, referred to as the voice coil motor 21, it is provided in advance with the voice coil horse; 2 =:: Motor parameter vector is given to controller 24, where f 21 is opposite to controller __ as an example 'Nano

98-5-2 P ’積分參數，％制器參數向量/ h分參數Kd。在此以向量的形式表98-5-2 P ’ integral parameter, % controller parameter vector / h sub-parameter Kd. Here in the form of a vector

K P Θ：步驟下俊C以及隨機標值k、第—衰減值A、第二衰減第〜衰減值A定^里△(步驟_。而在本實施例中，預先给定之實激二al/(a2+k+l)An ’其中ai、a2、n皆為〇·6〇2。同樣的，第較佳的值分別是al為Un為 r亦為預先給定之=減值C定義為姆价，其中cl、 0·101。另外，士2數，其較佳的值分別是cl為〇.卜n為 ±1· ±1 ±1 了說明本翻Ή控彻參數向量㈣上述开Μ表示，為只施例，隨機擾動向量Α表示為： Δ：例來Ϊ機選!1隨機擾動向量△的符號(步驟1G3)，以此實施量中"、]就是隨機選取如上述隨機擾動向量△所表示的向摄、一素之正負付號。接下來，作θ + CxA運算，得到ff ί 参數向量0+(步驟104)，也就是說以第二衰減制器灸、'擾ί值，以隨機擾動向量△作為擾動因子，對控 =麥數向量作正擾動。以正擾動控制器參數向量Θ +控二*三得到第—位置誤差值γ+(步驟105)，以圖2的實施馬控制态24以正擾動控制器參數向量0+控制音圈 1，並且透過回授電路23得到第一位置誤差值γ+。 1313536 1 98：5：20 ^ 一 ·~ 1%) 二量△作為擾動因子，對控制器參數擾動控制器參數向量吻馬達，得到第二位置，值^步驟浙），以圖2的實施例來說以、控制器參數向量θ-控制音圈馬達路23得到第二位置誤差值γ-。透過回授電測運=估測運算’得到梯度估測值_，其中梯度估，其中m為實數(步驟ι〇8)。下一步驟，作Θ-運算，將此運瞀姓向量_。9)。接下來，將旗標:;累:控:= H後取倾標值k1IQ)。雜錢上述=乍值^ 由上述圖1與圖2的實施例可以看出，是將比例參數Kp ’齡參數Ki，齡*數=主要參數向量f，並且對初始的控制器參數向量“擾;= 作梯度運异，以獲取新的控制器參數向量0，經 / 梯度運算後便可以得到更加趨近㈣器參數= 值，使得㈣馬達的定位可以更加的準確。絲知卜广者應當知道’ _參數Κρ，積分參數&，微衫數術為控制器參數向量Θ僅僅是針對PID控制器所提出^ 實施例，當控制器是比例積分⑼控制器時， 1313536 向量0可表示 X K,· θ 實施方式即變成比例參數Kp，積分參數Ki作為控制器參數向量0，另外，隨機擾動向量△便會依照控制器參數向量Θ的維度表示如下： Γ+Γ 土1 ΔKP Θ: Step C and the random value k, the first attenuation value A, the second attenuation value - the attenuation value A ^ 里 Δ (step _. In the present embodiment, the predetermined real excitation two a / (a2+k+l)An 'where ai, a2, n are all 〇·6〇2. Similarly, the preferred values are al for Un, r is also given in advance = depreciation C is defined as Price, where cl, 0·101. In addition, the number of ±2, the preferred value is cl is 〇.b n is ±1·±1 ±1 The description of the translation control parameter vector (4) For the only example, the random perturbation vector Α is expressed as: Δ: For example, the symbol of the random perturbation vector △ (step 1G3), in which the quantity ",] is randomly selected as the above random perturbation vector △ indicates the positive and negative sign of the camera, and then the θ + CxA operation, the ff ί parameter vector 0 + (step 104), that is, the second attenuation device moxibustion, 'disturbing value' Using the random perturbation vector △ as the disturbance factor, the control = mai number vector is positively perturbed. The positive disturbance controller parameter vector Θ + control two * three to obtain the first position error value γ + (step 105) The first position error value γ+ is obtained by the positive control device parameter vector 0+ with the horse control state 24 of Fig. 2, and the first position error value γ+ is obtained by the feedback circuit 23. 1313536 1 98:5:20 ^一·~ 1 %) Two quantities △ as the disturbance factor, the controller parameter perturbs the controller parameter vector kiss motor to obtain the second position, the value ^ step Zhejiang), in the embodiment of Fig. 2, the controller parameter vector θ-control tone The circle motor path 23 obtains a second position error value γ-. The gradient estimate _ is obtained by the feedback test = estimation operation, where the gradient is estimated, where m is a real number (step ι 8). In the next step, for the Θ-operation, this is the last name vector_. 9). Next, the flag:; tired: control: = H after the threshold value k1IQ). Miscellaneous money above = 乍 value ^ As can be seen from the above embodiment of Fig. 1 and Fig. 2, the proportional parameter Kp 'age parameter Ki, age * number = main parameter vector f, and the initial controller parameter vector "disturbed" ;= Gradient migration to obtain a new controller parameter vector 0, after / gradient operation can get closer to the (four) device parameter = value, so that (4) motor positioning can be more accurate. Knowing that ' _ parameter Κ ρ, integral parameter & micro-shirt number is the controller parameter vector Θ is only proposed for the PID controller ^ embodiment, when the controller is a proportional integral (9) controller, 1313536 vector 0 can represent XK The θ implementation becomes the proportional parameter Kp, and the integral parameter Ki is used as the controller parameter vector 0. In addition, the random disturbance vector △ is expressed as follows according to the dimension of the controller parameter vector :: Γ+Γ土1 Δ

X X 因此本發明應當不限於此實施例。上述的實施例中，第一衰減值A、第二衰減值c並非一定要如上述實施例中使用上述的特定值，上述實施例僅敘述了一個較佳的數值，然熟知此技術者應當知道，上述實施例中的第一衰減值A、第二衰減值c只要是隨著上述實施例所述迴圈:錄而齡，也就是隨著迴賊衰減就可以達到收斂的㈣，也就是最終會❹近理想的控制器參數向量Θ。例如第一衰減值A可以al/(k)An實施， 2 al、η為實數。例如第二衰減值c可以师)々實施， ^ ^為實數。然而即使第―、第二衰減值設為常數林發明，僅僅是迴圈次數必須要增加才可得到車父仏=繼參數㈣Θ，因此本發明並糾此為限。㈣^ ^上述的(步螺1G5)以正擾動控制器參數向量Θ + 控制馬達21，得刭坌—从班如、，第位置块差值γ+以及（步驟107)以 , 里Θ控制馬達，得到第二位置誤差值 Y 3以重禝發生，也轷3 就疋5兄，上述的步驟105可以利用正攮動控制态參數向+ 0 一再的控制馬達例如200次，並將 1313536 Οζ 'TP 98-5-20 回授的結果累計後產生第一位置誤差值γ+。同樣的道理，上述的步驟107可以利用負擾動控制器參數向量再的控制馬達例如200次，並將回授的結果累計後產生第一位置誤差值Υ+。因此本發明不限於此。X X Therefore the invention should not be limited to this embodiment. In the above embodiment, the first attenuation value A and the second attenuation value c are not necessarily used as the specific values described above in the above embodiment, and the above embodiment only describes a preferred value, but those skilled in the art should know The first attenuation value A and the second attenuation value c in the above embodiment may be converged as long as they are recorded according to the above embodiment, that is, as the thief decays (four), that is, finally Will be close to the ideal controller parameter vector Θ. For example, the first attenuation value A can be implemented by al/(k)An, and 2 al and η are real numbers. For example, the second attenuation value c can be implemented, ^ ^ is a real number. However, even if the first and second attenuation values are set to constant, the invention is only necessary to increase the number of loops to obtain the car father = the following parameter (four), so the present invention is limited to this. (4) ^ ^ The above (step screw 1G5) controls the motor 21 with the positive disturbance controller parameter vector Θ +, from the class, from the class position, the position block difference γ+ and (step 107), to control the motor The second position error value Y 3 is obtained to occur repeatedly, and 轷3 is 疋5 brothers. The above step 105 can use the positive sway control state parameter to control the motor to + 0 again, for example, 200 times, and 1313536 Οζ ' The result of the feedback of TP 98-5-20 is accumulated to generate the first position error value γ+. By the same token, the above step 107 can use the negative disturbance controller parameter vector to control the motor, for example, 200 times, and accumulate the results of the feedback to generate the first position error value Υ+. Therefore, the invention is not limited thereto.

圖3〜圖6繪示為依照本發明實施例圖1及圖2的俨制架構所模擬出的結果。請先參考圖3，其中圖3表示二階輸出響應，301表示實際值，302表示理想值。接;來^ 考圖4,其中圖4表示251^弦波輸出響應，4〇ι表示二^ 值’402表示理想值。接下來參考圖5，其中圖$ 、一不只示方波輪出響應，501表示實際值，5〇2表示理相7、，考圖6 ’其中圖6表示2服三角波輸出響應實際值，002表示理想值。叫i表不3 to 6 illustrate the results simulated by the throttle architecture of Figs. 1 and 2 in accordance with an embodiment of the present invention. Please refer to Figure 3 first, where Figure 3 shows the second-order output response, 301 for the actual value, and 302 for the ideal value. Figure 4, wherein Figure 4 shows the 251 chord wave output response, and 4 〇 ι means the value of '402' represents the ideal value. Referring next to FIG. 5, wherein the graph $, one not only shows the square wave rotation response, 501 represents the actual value, 5 〇 2 represents the physics phase 7, and the test figure 6 'where the figure 6 represents the actual value of the 2 service triangle wave output response, 002 Indicates the ideal value. Call i

圖7〜圖9繪示為依照本發明實施例圖1 制架構所f施出⑽量結果。請先參考圖7，^ 2的控不弦波輸出響應’ 701表示類比數位轉換器2H 號，702表示數位類比轉換器25的輸出信號圖8,其中圖8表示方波輸出響應，剛表示:來多考器23的輸人信號，表示數位類比轉換器μ ^職 ::接下來參考圖9’其中圖9表示三角波輪’出：表不類比數位轉換器23的輪入信號，9〇2 ，應，901 換器25的輸出信號。〜、數位類比轉綜上所述，本發_制__隨機近似的㈣在、'表上直接改變控彻中預設的控參數向量，因:除 12 1313536 )ί 1— 9K-5-20 了可以提高控制系統的強健性，改善受控裝置的動態響應外’進一步可達到準確的控制不同的受控裝置，並且此方法易於在軟體或硬體上實施。雖然本發明已以較佳實施例揭露如上，然其並非用以限^本發明’任㈣f此㈣者，在賴離本發明之精神 =範圍内，當可作些許之更動與潤飾，因此本發明之保護範圍當視後附之申請專利範圍所界定者為準。FIG. 7 to FIG. 9 are diagrams showing the results of the (10) quantity applied to the architecture of FIG. 1 according to an embodiment of the invention. Referring first to FIG. 7, the control sinusoidal output response ' 701 represents the analog-to-digital converter 2H number, 702 represents the output signal of the digital analog converter 25, and FIG. 8 shows the square wave output response, which just indicates: The input signal of the multi-tester 23, representing the digital analog converter μ:: Next, referring to FIG. 9', wherein FIG. 9 shows the triangular wave wheel' out: the rounding signal of the analog-to-digital converter 23, 9〇2 , should, 901 converter 25 output signal. ~, digital analogy on the above, the _ system __ random approximation (four) in, 'the table directly changes the control parameter vector preset in the control, because: except 12 1313536) ί 1-9K-5- 20 can improve the robustness of the control system and improve the dynamic response of the controlled device. Further, it is possible to accurately control different controlled devices, and the method is easy to implement on software or hardware. Although the present invention has been disclosed in the above preferred embodiments, it is not intended to limit the invention to any of the four inventions, and it is possible to make some modifications and refinements within the scope of the spirit of the present invention. The scope of the invention is defined by the scope of the appended claims.

【圖式簡單說明】’ 似^ ^ I本發明實施例之—縣關時擾動隨機： =磁向量線上調整的控制方法流程圖。馬達:體==:方法所實施的硬碟音· 制架構所模擬出依'本發明實施例圖1及圖2你[Simple Description of the Drawings] ‘I ^ ^ I In the embodiment of the present invention - the county off-time disturbance random: = flow chart of the control method of the magnetic vector line adjustment. Motor: Body ==: The hard disk sound system implemented by the method is simulated according to the embodiment of the present invention. Figure 1 and Figure 2

圖7〜圖9繪示為依照本發明制架構所實施出的測量結果。也1 Θ θ 2的: 【主要元件符號說明】 20 21 23 24 25 26 101 110 .本發明貫施例的步驟硬碟碟片音圈馬達類比數位轉換器控制器數位類比轉換器馬達驅動器 13 1313536 许年4月2a日修正買 98-5-20 301 :步階輸出響應實際值 302 :步階輸出響應理想值 401 : 25Hz弦波輸出響應實際值 402 : 25Hz弦波輸出響應理想值 501 : 25Hz方波輸出響應實際值 502 : 25Hz方波輸出響應理想值 601 : 25Hz三角波輸出響應實際值Figures 7 through 9 illustrate measurement results performed in accordance with the architecture of the present invention. Also 1 Θ θ 2: [Main component symbol description] 20 21 23 24 25 26 101 110. The steps of the present invention are hard disk disc voice coil motor analog digital converter controller digital analog converter motor driver 13 1313536 Amendment April 2a, modified buy 98-5-20 301: step output response actual value 302: step output response ideal value 401: 25Hz sine wave output response actual value 402: 25Hz sine wave output response ideal value 501 : 25Hz Square wave output response actual value 502 : 25Hz square wave output response ideal value 601 : 25Hz triangle wave output response actual value

602 : 25Hz三角波輸出響應理想值 701 :弦波輸出響應中類比數位轉換器23的輸入信號 702 :弦波輸出響應中數位類比轉換器25的輸出信號 801 :方波輸出響應中類比數位轉換器23的輸入信號 802 :方波輸出響應中數位類比轉換器25的輸出信號 901 :三角波輸出響應中類比數位轉換器23的輸入信號 902 :三角波輸出響應中數位類比轉換器25的輸出信602: 25Hz triangular wave output response ideal value 701: input signal 702 of analog-to-digital converter 23 in sine wave output response: output signal 801 of digital analog converter 25 in sine wave output response: analog wave digital converter 23 in square wave output response Input signal 802: output signal 901 of the digital analog converter 25 in the square wave output response: input signal 902 of the analog to digital converter 23 in the triangular wave output response: output signal of the digital analog converter 25 in the triangular wave output response

1414

Claims

1313536 ! …一——98-5-20'"" 十、申請專利範圍： 1. 一種基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，此方法包括：根據一受控裝置的種類，提供一控制器參數向量0 ; 定義一第一衰減值A、一第二衰減值C以及一隨機擾動向量△; a. 隨機選取該隨機擾動向量△的符號； b. 作θ+cx△運算，得到正擾動控制器參數向量<9+; c. 以該正擾動控制器參數向量6>+控制該受控裝置，得到一第一位置誤差值Y+ ; d. 作Θ-Cx△運算，得到負擾動控制器參數向量; e. 以該負擾動控制器參數向量Θ —控制該受控裝置，得到一第二位置誤差值Y-; f. 作一梯度估測運算，得到一梯度估測值g(e)，其中該梯度估測運算如下：，其中m為實數； mC g. 作θ-Αβθ)運算，將此運算結果取代控制器參數向量 0 ;以及 h. 重複上述步驟a〜！1 ; 其中該控制器的轉移函數為：，Kp 代表一比例參數，& 丄一Z 代表一積分參數，Kd代表一微分參數；該隨機擾動向量A 表示為： 15 I3l3536 98-5-20 ，且隨機遽取該隨機擾動向量△沾> L±iJ 巧付號包括隨機選取該隨機擾動向量△中的元素之正負符號。 2.如申請專利範圍第丨項所述之基於同時擾動似法之控制器參數向量線上調整的控制方法，发近器參數向量為： 4該控制 K. 鲁 ±1. ±1 ±1 3. -如申請專利範圍第1項所述之基於同時擡似法之控制器參數向量線上調整的控制方法，发思機近器的轉移函數為： /、該控制 1-Z'1 其中Kp代表一比例參數，Ki代表一積分參數。 4·如申請專利範圍第3項所述之基於同時以法之控制器參數向量線上調整的控制方法，复思近器參數向量為： ^中該控制 θ = [\]。 _κ, 5·如申清專利範圍第3項所述之基於同似法之控㈣參數向量線上調整的控制方法，=隨機近擾動向量△表示為 /、中5亥隨機，且隨機選取該隨機擾動向量厶。括Ik機選取該隨機擾動向量△中的元素之正負符說付就包 6·如申請專利範圍第i項所述之基於同時擾動。隨機近 16 1313536 祖; .............98-5-20 '……’ 似法之控制器參數向量線上調整的控制方法，其中步驟h 之前更包括：提供一旗標值k ; 根據該旗標值k，將該第一衰減值A作一第一衰減運算後，取代該第一衰減值；根據該旗標值k，將該第二衰減值C作一第二衰減運算後，取代該第二衰減值；以及 > 將該旗標值k累加一預設值作為該旗標值k。 7. 如申請專利範圍第6項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中該第一衰減運算為： A=al/(a2+k)An，其中 al、a2、η 為實數。 8. 如申請專利範圍第6項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中該第一衰減運算為： . al/(a2+k+l)An，其中 al、a2、η 為實數。 9. 如申請專利範圍第6項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中該第一衰減運算為： A=al/(k)An，其中al、η為實婁欠。 10. 如申請專利範圍第6項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中該第二衰減運算為： 17 1313536 声κ月 98-5-20 >ϋ日修替捷頁 C=cl/(k)Ar，其中 cl、r 為實數。、如”專圍第6項所述之基於同時近似法之控難參數向量線上調整的蝴方法，=、二衰減運算為：荜弟 C—cl/(k+l)Ar，其中 cl、r 為實數。該梯 12·如申請專利範圍第丨項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中度估測運算中的實數m為2。 13. 如申請專利範圍第丨項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中該受控裝置為一馬達。 14. 如申請專利範圍弟13項所述之基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法，其中該馬達為音圈馬達。 18 1313536 1ΐ·ψ^ΐ：} 98-5-20 五、中文發明摘要： —種基於同時擾動隨機近似法之控制器參數向量線上調整的控制方法。此方法主要是利用同時擾動隨機近似 (SPSA)演算法’藉由改變控制器的參數以控制受控裝置。在適應控制的架構之下，利用SPSA來作為線上調^控制器參數向量的法則，自動搜尋受控裝置的最佳控制器^數向量。六、英文發明摘要：籲七、指定代表圖： (一) 本案指定代表圖為：圖（1)。 (二) 本代表圖之元件符號簡單說明： 101〜110 :本發明實施例的步驟八、本案若有化學式時，請揭示最能顯示發明特徵的化學式：無 A control method for on-line tuning parameters of the controller based on simultaneous perturbation stochastic approximation (SPSA) is disclosed. The method mainly uses simultaneous perturbation stochastic approximation (SPSA) algorithm, by tuning the parameters of the PID controller, to control a controlled apparatus. In an adaptive controlling structure, using SPSA algorithm to be an approach for on-line adjusting parameters of a controller so that could reach automatically the optimum parameters of the controller.1313536 ! ...一——98-5-20'"" X. Patent application scope: 1. A control method for vector parameter adjustment of controller parameters based on simultaneous perturbation stochastic approximation method, the method includes: a type of device, providing a controller parameter vector 0; defining a first attenuation value A, a second attenuation value C, and a random perturbation vector Δ; a. randomly selecting the sign of the random perturbation vector Δ; b. The cx △ operation obtains the positive disturbance controller parameter vector <9+; c. controls the controlled device with the positive disturbance controller parameter vector 6>+ to obtain a first position error value Y+; d. Θ-Cx △ operation, obtaining a negative disturbance controller parameter vector; e. controlling the controlled device with the negative disturbance controller parameter vector 得到 to obtain a second position error value Y-; f. performing a gradient estimation operation to obtain a The gradient estimate g(e), wherein the gradient estimation operation is as follows: , where m is a real number; mC g. is a θ-Αβθ) operation, replacing the controller parameter vector 0 with the result of the operation; and h. repeating the above steps a~! 1; wherein the transfer function of the controller is: , Kp represents a proportional parameter, & 丄Z represents an integral parameter, and Kd represents a differential parameter; the random perturbation vector A is expressed as: 15 I3l3536 98-5-20, And randomly acquiring the random perturbation vector Δ &> L±iJ 巧付号 includes randomly selecting the positive and negative signs of the elements in the random perturbation vector Δ. 2. For the control method of the vector parameter adjustment of the controller parameter based on the simultaneous perturbation like method according to the scope of the patent application, the vector parameter of the approximator is: 4 The control K. Lu ±1. ±1 ±1 3. - The control function of the controller parameter vector vector adjustment based on the simultaneous lift method is as follows: /, the control 1-Z'1 where Kp represents a Proportional parameter, Ki represents an integral parameter. 4. If the control method based on the simultaneous adjustment of the controller parameter vector line described in the third paragraph of the patent application scope, the reconsideration parameter vector is: ^ The control θ = [\]. _κ, 5· For example, according to the third paragraph of the patent scope of the patent, the control method based on the same method (4) parameter vector line adjustment, = random near disturbance vector △ is expressed as /, medium 5 hai random, and randomly selected the random Disturbed vector 厶. The Ik machine selects the sign of the element in the random perturbation vector Δ and says that it is based on the simultaneous perturbation as described in item i of the patent application. Randomly close to 16 1313536 ancestors; .............98-5-20 '...' The control method of the controller parameter vector line adjustment of the like method, wherein before step h, it further includes: a flag value k; according to the flag value k, after the first attenuation value A is subjected to a first attenuation operation, replacing the first attenuation value; and according to the flag value k, the second attenuation value C is made into a After the second attenuation operation, the second attenuation value is replaced; and > the flag value k is accumulated by a preset value as the flag value k. 7. The control method for adjusting a vector parameter of a controller parameter based on a simultaneous perturbation stochastic approximation according to claim 6, wherein the first attenuation operation is: A=al/(a2+k)An, where al , a2, η are real numbers. 8. The control method for adjusting the vector parameter of a controller parameter based on the simultaneous perturbation stochastic approximation method as claimed in claim 6, wherein the first attenuation operation is: . al/(a2+k+l)An, wherein Al, a2, and η are real numbers. 9. The control method for adjusting a vector parameter of a controller parameter based on a simultaneous perturbation stochastic approximation method as claimed in claim 6, wherein the first attenuation operation is: A=al/(k)An, where al, η To be owed. 10. The control method for adjusting the vector parameter of the controller parameter based on the simultaneous perturbation stochastic approximation method according to claim 6 of the patent application scope, wherein the second attenuation operation is: 17 1313536 sound κ月 98-5-20 > The day is replaced by the C=cl/(k)Ar, where cl and r are real numbers. For example, the butterfly method based on the simultaneous approximation method for adjusting the difficult parameter vector line according to the sixth item, =, and the second attenuation operation is: C-cl/(k+l)Ar, where cl, r The real number. The ladder 12 is a control method for adjusting the vector parameter of the controller parameter based on the simultaneous perturbation stochastic approximation as described in the scope of the patent application, wherein the real number m in the degree estimation operation is 2. The control method for adjusting the vector parameter of the controller parameter based on the simultaneous perturbation stochastic approximation method described in the third paragraph of the patent scope, wherein the controlled device is a motor. 14. The random disturbance based on simultaneous disturbance according to claim 13 The control method of the vector parameter adjustment of the controller parameter of the approximation method, wherein the motor is a voice coil motor. 18 1313536 1ΐ·ψ^ΐ:} 98-5-20 V. Chinese invention summary: a kind of simultaneous approximation based on random approximation The control method for adjusting the controller parameter vector line. This method mainly uses the simultaneous disturbance random approximation (SPSA) algorithm to control the controlled device by changing the parameters of the controller. Under the architecture, SPSA is used as the rule of the parameter vector of the online controller, and the optimal controller number vector of the controlled device is automatically searched. VI. English invention summary: VII, designated representative map: (1) The representative figure is: Figure (1). (b) The symbol of the representative figure is briefly described: 101~110: Step 8 of the embodiment of the present invention. If there is a chemical formula in this case, please disclose the chemical formula that best shows the characteristics of the invention: none A control method for on-line tuning parameters of the controller based on simultaneous perturbation stochastic approximation (SPSA) is disclosed. The method mainly uses simultaneous perturbation stochastic approximation (SPSA) algorithm, by tuning the parameters of the PID controller, to control a controlled In an adaptive controlling structure, using SPSA algorithm to be an approach for on-line adjusting parameters of a controller so that could reach automatically the optimum parameters of the controller.