JPH0731522B2 - Adaptive controller - Google Patents

Adaptive controller

Info

Publication number
JPH0731522B2
JPH0731522B2 JP63302693A JP30269388A JPH0731522B2 JP H0731522 B2 JPH0731522 B2 JP H0731522B2 JP 63302693 A JP63302693 A JP 63302693A JP 30269388 A JP30269388 A JP 30269388A JP H0731522 B2 JPH0731522 B2 JP H0731522B2
Authority
JP
Japan
Prior art keywords
circuit
controlled object
output
control
object model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
JP63302693A
Other languages
Japanese (ja)
Other versions
JPH02148201A (en
Inventor
経一 遠藤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Toshiba Corp
Original Assignee
Toshiba Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Toshiba Corp filed Critical Toshiba Corp
Priority to JP63302693A priority Critical patent/JPH0731522B2/en
Publication of JPH02148201A publication Critical patent/JPH02148201A/en
Publication of JPH0731522B2 publication Critical patent/JPH0731522B2/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Landscapes

  • Feedback Control In General (AREA)

Description

【発明の詳細な説明】 [発明の目的] (産業上の利用分野) 本発明は、プラントやロボットの制御において、制御対
象の動特性が十分にわからない場合や、操作中に動特性
が変動する制御対象に対して、コントローラを自動的に
適応させる適応制御装置に関するものである。
DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial field of application) The present invention relates to the control of a plant or a robot when the dynamic characteristics of a controlled object are not sufficiently understood or the dynamic characteristics fluctuate during operation. The present invention relates to an adaptive control device that automatically adapts a controller to a controlled object.

(従来の技術) 適応制御方式には、セルフチューニングレギュレータや
モデル規範形適応制御がある。これらの方式では、制御
対象の動特性モデルが線形ダイナミカルシステムで表現
できる場合にのみ効果がある。
(Prior Art) Adaptive control methods include self-tuning regulators and model reference adaptive control. These methods are effective only when the dynamic characteristic model of the controlled object can be represented by a linear dynamical system.

(発明が解決しようとする課題) しかしながら、制御対象の動特性が非線形である場合
や、制御対象の動特性に関する十分な知識が得られない
場合には、前記適応制御方式は使うことはできず、実用
性に乏しい。
(Problems to be Solved by the Invention) However, the adaptive control method cannot be used when the dynamic characteristic of the controlled object is non-linear or when sufficient knowledge about the dynamic characteristic of the controlled object cannot be obtained. , Poor in practicality.

本発明は、以上の点を考慮してなされたもので、線形ダ
イナミカルシステムに限定されない多様な制御対象に対
して適用できる自己学習機能をもった適応制御装置の提
供を目的としている。
The present invention has been made in consideration of the above points, and an object of the present invention is to provide an adaptive control device having a self-learning function that can be applied to various controlled objects not limited to a linear dynamical system.

[発明の構成] (課題を解決するための手段) 本発明の適応制御装置は、上記した目的を達成するため
に、人間の脳を構成する神経細胞の動特性を単純化し工
学的にモデル化した神経素子を相互に結合して実現され
た神経回路網により制御回路と制御対象モデル回路を構
成する。神経回路網は、任意の連続関数を近似できると
いう性質をもつことが明らかになっている。この性質に
より上記各回路の動特性の近似可能性の範囲が増大す
る。また、神経回路網では、複数の入出力関係データか
ら、神経素子間の結合荷重を調整する事により、それら
の関数関係を得ることができる。これを神経回路網によ
る学習と呼ばれている。本発明は、神経回路網による学
習方式を適用制御に応用したものである。本発明では、
制御回路部は、制御対象を制御しながら、同時に、仮想
的に制御対象モデル回路も制御する。その結果として得
られる制御対象の出力に制御対象モデル回路の出力が一
致するように制御対象モデル回路の神経素子間の結合荷
重を調整することにより制御対象モデル回路に制御対象
を模擬する神経回路網を実現する。さらに、制御対象モ
デル回路の出力を制御目標に一致させるために制御回路
の神経素子間の結合荷重をも同時に調整する。
[Structure of the Invention] (Means for Solving the Problems) In order to achieve the above-mentioned object, the adaptive control device of the present invention simplifies the dynamic characteristics of nerve cells constituting the human brain and models them in an engineering manner. A control circuit and a control target model circuit are configured by a neural network realized by mutually connecting the neural elements. Neural networks have been shown to have the property of approximating arbitrary continuous functions. This property increases the range of possibilities of approximating the dynamic characteristics of the above circuits. Further, in the neural network, it is possible to obtain the functional relationships from a plurality of input / output relationship data by adjusting the coupling load between the neural elements. This is called learning by a neural network. The present invention applies a learning method based on a neural network to applied control. In the present invention,
While controlling the control target, the control circuit unit also virtually controls the control target model circuit. A neural network that simulates the controlled object in the controlled object model circuit by adjusting the coupling load between the neural elements of the controlled object model circuit so that the output of the controlled object model circuit matches the output of the resulting controlled object To realize. Further, in order to match the output of the controlled object model circuit with the control target, the coupling load between the neural elements of the control circuit is also adjusted at the same time.

(作 用) 本発明の適応制御装置により、より多くの制御対象に対
して、その動特性が十分にわからなくても学習により動
特性の知識を自動的に獲得し、同時に制御目標を満足す
るコントローラも自動的に実現することができる。
(Operation) With the adaptive control device of the present invention, knowledge of dynamic characteristics is automatically acquired by learning for more controlled objects even if the dynamic characteristics are not sufficiently understood, and at the same time, the control target is satisfied. The controller can also be realized automatically.

(実施例) 以下、本発明の実施例を図面に基づいて詳細に説明す
る。
(Example) Hereinafter, the Example of this invention is described in detail based on drawing.

第1図は本発明の一実施例に係る適応制御装置の構成を
示すブロック図である。
FIG. 1 is a block diagram showing the configuration of an adaptive control device according to an embodiment of the present invention.

1は制御対象モデル回路を示し、2は制御回路を示す。
3は制御対象モデル回路の神経素子間の結合を調整する
ための制御対象モデル部荷重調整回路を示し、4は制御
回路の神経素子間の結合を調整するための制御部荷重調
整回路を示す。5は制御目標信号発生回路を示す。6は
プラントやロボットなどの制御対象を示す。
Reference numeral 1 represents a controlled object model circuit, and 2 represents a control circuit.
Reference numeral 3 represents a controlled object model load adjusting circuit for adjusting the coupling between the neural elements of the controlled object model circuit, and 4 represents a controller load adjusting circuit for adjusting the coupling between the neural elements of the control circuit. Reference numeral 5 indicates a control target signal generation circuit. Reference numeral 6 indicates a control target such as a plant or a robot.

制御回路2は、制御対象6から制御量信号Yおよび制御
目標信号発生回路5から目標値信号Rを入力して操作量
信号Uを出力する。Uは、制御対象モデル回路1および
制御対象6に同時に入力される。その結果、制御対象の
応答信号である制御量信号Yが新たに発生する。一方、
制御対象モデル回路1はUを入力して、Yの近似信号Y
を出力する。制御対象モデル部荷重調整回路3は、Y
とYを入力してYとYとのずれからYをYに近ず
けるために、制御対象モデル回路の荷重調整量dWを計
算して制御対象モデル回路1に出力する。さらに、制御
部荷重調整回路4は、目標値信号RとYを入力して、
とRのずれからYをRに近ずけるために、制御回
路2の荷重調整量dWを計算して制御回路2に出力す
る。
The control circuit 2 inputs the controlled variable signal Y from the controlled object 6 and the target value signal R from the controlled target signal generation circuit 5, and outputs the manipulated variable signal U. U is simultaneously input to the controlled object model circuit 1 and the controlled object 6. As a result, a controlled variable signal Y, which is a response signal of the controlled object, is newly generated. on the other hand,
The controlled object model circuit 1 inputs U and outputs an approximate signal Y of Y.
Output M. The controlled model load adjustment circuit 3 is Y
In order to bring Y M closer to Y from the deviation between Y M and Y by inputting Y and Y M , the load adjustment amount dW M of the controlled object model circuit is calculated and output to the controlled object model circuit 1. Further, the control unit load adjusting circuit 4 inputs the target value signals R and Y M ,
From the deviation of Y M and R to Y M for Keru not a close to R, and outputs to the control circuit 2 calculates a load adjustment amount dW C of the control circuit 2.

制御対象モデル回路1と制御回路2は、例えば第2図に
示す神経回路網により構成される。神経回路網を構成す
る神経素子の動特性として、種々の動特性式が考えられ
る。時刻tでの神経素子iの出力をXi(t)は、 あるいは、第2図(b)に示す層状神経回路網では、 などで与えられる。ここで、Xjは、神経素子iの入力部
(シナプス)に結合している神経素子jの出力を示す。
Wijは、神経素気iの入力部(シナプス)に結合してい
る神経素子jとの結合荷重を示す。θiは、神経素子i
のしきい値を示す。
The controlled object model circuit 1 and the control circuit 2 are configured by, for example, the neural network shown in FIG. Various dynamic characteristic equations can be considered as the dynamic characteristics of the neural elements that make up the neural network. Xi (t) is the output of the neural element i at time t, Alternatively, in the layered neural network shown in FIG. 2 (b), Given in. Here, Xj represents the output of the neural element j coupled to the input part (synapse) of the neural element i.
Wij represents the coupling load with the neural element j coupled to the input part (synapse) of the nerve element i. θi is the neural element i
Indicates the threshold of.

(2)式では、Wijは神経素子iと、それの属する層の
直前の層を構成する神経素子jとの間の結合荷重を、Wi
lは神経素子iと、それの属する層の神経素子lとの間
の結合荷重をそれぞれ示している。fは、アクティベー
ト関数と呼ばれ、例えば次式に示すようなシグモイド関
数が使われる。
In the equation (2), Wij is the connection load between the neural element i and the neural element j forming the layer immediately before the layer to which Wij
l represents the coupling load between the neural element i and the neural element l of the layer to which it belongs. f is called an activate function, and for example, a sigmoid function as shown in the following equation is used.

f(x)=1/(1+exp(−a・x)) (3) 次に、このような神経素子から構成された神経回路網で
実現された制御対象モデル回路1と制御回路2の学習方
法、すなわち、制御対象モデル部荷重調整回路3および
制御部荷重調整回路4における荷重調整方法について説
明する。
f (x) = 1 / (1 + exp (−a · x)) (3) Next, a learning method for the controlled object model circuit 1 and the control circuit 2 realized by the neural network composed of such neural elements. That is, the load adjusting method in the controlled object model part load adjusting circuit 3 and the control part load adjusting circuit 4 will be described.

制御対象がm入力n出力のシステムとするとき、時刻t
での前記各信号U、Y、Y、Rを次のようなベクトル
で表すことにする。
When the controlled object is a system with m inputs and n outputs, time t
The respective signals U, Y, Y M , and R in the above equation are represented by the following vectors.

操作量信号U U(t) =(u1(t),u2(t),…,ui(t)…,Um(t)) 制御量信号Y U(t) =(y1(t),Y2(t),…,yi(t),Yn(t)) 制御対象モデル回路からの出力Y(t) =(y (t),y (t),…,Yi(t),…,Y
n(t)) 制御目標信号R R(t) =(r1(t),r2(t),…,ri(t),…,rn(t))
(Tは転置を表す) このとき、制御対象モデル回路1の動特性Fおよび制
御回路2の動特性Fを次式で表す。
Manipulated variable signal U U (t) = (u 1 (t), u 2 (t), ..., ui (t) ..., Um (t)) T controlled variable signal Y U (t) = (y 1 (t ), Y 2 (t), ..., yi (t), Yn (t)) T Output from the controlled model circuit Y M Y m (t) = (y M 1 (t), y M 2 (t) , ..., Y M i (t), ..., Y
M n (t)) T control target signal R R (t) = (r 1 (t), r 2 (t), ..., ri (t), ..., rn (t))
T (T represents transposition) At this time, the dynamic characteristic F M of the controlled object model circuit 1 and the dynamic characteristic F C of the control circuit 2 are expressed by the following equations.

U(t)=Fc(R(t),Y(t),Xc(t−1),Wc
(t)) (4) Y(t)=F(U(t)、X(t−1)、W
(t)) (5) ここで、Xcは制御回路を構成する神経素子の出力ベクト
ルを、Wcは制御回路を構成する神経素子間の結合荷重マ
トリクス(しきい値も含む)を示す。同様に、Xは制
御対象モデル回路を構成する神経素子の出力ベクトル
を、Wは制御対象モデル回路を構成する神経素子間の
結合荷重マトリクス(しきい値も含む)を示す。制御対
象モデル回路の出力Yを、制御対象の出力Yに近ずけ
るために次のような評価関数Eを設定する。
U (t) = Fc (R (t), Y (t), Xc (t-1), Wc
(T)) (4) Y M (t) = F M (U (t), X M (t−1), W
M (t)) (5) Here, Xc represents the output vector of the neural element forming the control circuit, and Wc represents the coupling weight matrix (including the threshold value) between the neural elements forming the control circuit. Similarly, X M indicates the output vector of the neural element forming the controlled object model circuit, and W M indicates the coupling weight matrix (including the threshold value) between the neural elements forming the controlled object model circuit. In order to bring the output Y M of the controlled object model circuit closer to the output Y of the controlled object, the following evaluation function E M is set.

=(Y−Y)・(Y−Y)/2 (6) 同様に、制御対象モデル回路の出力Yを、制御目標R
に近づけるために次のような評価関数Ecを設定する。
E M = (Y M −Y) T · (Y M −Y) / 2 (6) Similarly, the output Y M of the controlled object model circuit is set to the control target R
The following evaluation function Ec is set in order to approach.

Ec=(Y−R)・(Y−R)/2 (7) 荷重修正量マトリクスdW、dWcは、例えば神経回路網
の学習方法の1つであるバックプロパゲーション法を使
うと次式で表すことができる。
Ec = (Y M -R) T · (Y M -R) / 2 (7) load correction amount matrix dW M, DWC, for example when using a back propagation method, which is one of the learning process of the neural network It can be expressed by the following equation.

dW(t+1)=−αc(t)Ec/Wc/Wc(t) +βc(t)・dWc(t−1) (8) dW(t+1)=−α(t)E/W/W(t) +β(t)・dW(t−1) (9) α、β、αc、βcは、修正の程度を決めるパラメ
ータである。(8)および(9)式から結合荷重の修正
を次式で行なう。
dW C (t + 1) = − αc (t) Ec / Wc / Wc (t) + βc (t) · dWc (t−1) (8) dW M (t + 1) = − α M (t) E M / W M / W M (t) + β M (t) · dW M (t-1) (9) α M, β M, αc, βc is a parameter that determines the degree of correction. The coupling load is corrected from the equations (8) and (9) using the following equation.

(t+1)=Wc(t)+dW(t+1) (10) W(t+1)=W(t)+dW(t+1) (11) 次に処理手順について説明する。第3図は、本発明の処
理手順のフローチャートである。毎サンプル時刻ごとに
この処理が実行される。サンプル時刻tで、ステップSP
1において処理を開始した後、ステップSP2において制御
目標値信号Rと制御信号Yを読み取り、記憶する。ステ
ップSP3において記憶されたRとYから(4)式を計算
して操作量信号Uを求め、記憶する。ステップSP4で
は、操作量信号Uから(5)式を計算して制御対象モデ
ル回路の出力信号Yを求め、記憶する。ステップSP6
では、記憶されていた制御目標値信号Rと制御対象モデ
ル回路の出力信号Yから(8)式により制御回路の荷
重修正量dWを計算し、制御回路に記憶されている結合
荷重を(10)式により修正する。ステップSP7で制御量
信号Yを読み取り、記憶する。ステップSP8では、記憶
されていた制御量信号Yと制御対象モデル回路の出力信
号Yから(9)式により制御対象モデル回路の荷重修
正量dWを計算し、制御対象モデル回路に記憶されてい
る結合荷重を(11)式により修正する。ステップSP9で
は、サンプル時刻tの処理を終了して次のサンプル時刻
t+1まで処理プロセスを待機状態にする。
W c (t + 1) = Wc (t) + dW c (t + 1) (10) W M (t + 1) = W M (t) + dW M (t + 1) (11) Next, the processing procedure will be described. FIG. 3 is a flowchart of the processing procedure of the present invention. This process is executed at every sample time. At sample time t, step SP
After starting the process in 1, the control target value signal R and the control signal Y are read and stored in step SP2. Equation (4) is calculated from R and Y stored in step SP3 to obtain and store the manipulated variable signal U. In step SP4, the equation (5) is calculated from the manipulated variable signal U to obtain and store the output signal Y M of the controlled object model circuit. Step SP6
Then, the load correction amount dW c of the control circuit is calculated from the stored control target value signal R and the output signal Y M of the controlled object model circuit by the equation (8), and the coupling load stored in the control circuit is calculated as Correct according to equation 10). At step SP7, the control amount signal Y is read and stored. In step SP8, the load correction amount dW M of the controlled object model circuit is calculated from the stored controlled variable signal Y and the output signal Y M of the controlled object model circuit by the equation (9), and is stored in the controlled object model circuit. Correct the combined load by using equation (11). At step SP9, the processing at the sample time t is completed, and the processing process is put in a standby state until the next sample time t + 1.

[発明の効果] 上述したように本発明によれば、より多くの制御対象に
対して、その動特性が十分にわからなくても学習により
動特性の知識を自動的に獲得し、同時に制御目標を満足
する適応制御装置を実現することができる。
[Effects of the Invention] As described above, according to the present invention, knowledge of dynamic characteristics can be automatically acquired by learning for a larger number of control objects even if the dynamic characteristics are not sufficiently understood, and at the same time, control targets can be obtained. It is possible to realize an adaptive control device that satisfies

【図面の簡単な説明】[Brief description of drawings]

第1図は本発明による適応制御方式を適用した制御シス
テムの全体構成を示すブロック図、第2図は制御対象モ
デル回路1と制御回路2を実現する神経回路網の構成
図、第3図は適応制御の処理手順を示すフローチャート
である。 1……制御対象モデル回路、2……制御回路、3……制
御対象モデル部荷重調整回路、4……制御部荷重調整回
路、5……制御目標信号発生回路、6……制御対象。
FIG. 1 is a block diagram showing the overall configuration of a control system to which the adaptive control method according to the present invention is applied, FIG. 2 is a configuration diagram of a neural network for realizing a controlled object model circuit 1 and a control circuit 2, and FIG. It is a flow chart which shows the processing procedure of adaptive control. 1 ... Control object model circuit, 2 ... Control circuit, 3 ... Control object model part load adjusting circuit, 4 ... Control part load adjusting circuit, 5 ... Control target signal generating circuit, 6 ... Control object.

Claims (1)

【特許請求の範囲】[Claims] 【請求項1】制御対象と、 この制御対象を第1の神経回路網によりモデル化してな
る制御対象モデル回路と、 前記制御対象の制御目標値を出力する制御目標信号発生
回路と、 第2の神経回路網からなり、前記制御目標信号発生回路
からの出力と前記制御対象モデル回路からの出力とに基
づいて、前記制御対象及び前記制御対象モデル回路を同
時に制御する制御回路と、 前記制御対象の出力と前記制御対象モデル回路の出力と
が一致するように、前記制御対象の出力と前記制御対象
モデル回路の出力とに基づいて、前記制御対象モデル回
路における第1の神経回路網の結合荷重を調整する制御
対象モデル回路部調整回路と、 前記制御対象に対する制御目標値と前記制御対象モデル
回路の出力とが一致するように、前記制御目標信号発生
回路の出力と前記制御対象モデル回路との出力とに基づ
いて、前記制御回路における第2の神経回路網の結合荷
重を調整する制御回路部荷重調整回路と を具備することを特徴とする適応制御装置。
1. A controlled object, a controlled object model circuit obtained by modeling the controlled object by a first neural network, a control target signal generating circuit for outputting a control target value of the controlled object, and a second Consisting of a neural network, based on the output from the control target signal generation circuit and the output from the controlled object model circuit, a control circuit for simultaneously controlling the controlled object and the controlled object model circuit, Based on the output of the controlled object and the output of the controlled object model circuit, the coupling load of the first neural network in the controlled object model circuit is set so that the output and the output of the controlled object model circuit match. The control target model circuit adjusting circuit to be adjusted, and the control target signal generation circuit so that the control target value for the control target and the output of the control target model circuit match. Based on the output and the output of said controlled object model circuit, the adaptive control apparatus characterized by comprising a control circuit unit load adjusting circuit for adjusting the connection weights of the second neural network in the control circuit.
JP63302693A 1988-11-30 1988-11-30 Adaptive controller Expired - Fee Related JPH0731522B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP63302693A JPH0731522B2 (en) 1988-11-30 1988-11-30 Adaptive controller

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP63302693A JPH0731522B2 (en) 1988-11-30 1988-11-30 Adaptive controller

Publications (2)

Publication Number Publication Date
JPH02148201A JPH02148201A (en) 1990-06-07
JPH0731522B2 true JPH0731522B2 (en) 1995-04-10

Family

ID=17912057

Family Applications (1)

Application Number Title Priority Date Filing Date
JP63302693A Expired - Fee Related JPH0731522B2 (en) 1988-11-30 1988-11-30 Adaptive controller

Country Status (1)

Country Link
JP (1) JPH0731522B2 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104216291A (en) * 2014-09-04 2014-12-17 广州文冲船厂有限责任公司 Kinetic energy and momentum control system and control method thereof of working ship

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3221497B2 (en) * 1991-06-20 2001-10-22 三菱電機株式会社 Control method and power system voltage-reactive power control device using the control method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4368510A (en) * 1980-10-20 1983-01-11 Leeds & Northrup Company Automatic identification system for self tuning process controller

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
計測と制御,Vol.27No.11(S63.11.10)PP.61−69

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104216291A (en) * 2014-09-04 2014-12-17 广州文冲船厂有限责任公司 Kinetic energy and momentum control system and control method thereof of working ship

Also Published As

Publication number Publication date
JPH02148201A (en) 1990-06-07

Similar Documents

Publication Publication Date Title
Limanond et al. Neural network-based control design: An LMI approach
US5471381A (en) Intelligent servomechanism controller
US7647284B2 (en) Fixed-weight recurrent neural network controller with fixed long-term and adaptive short-term memory
Widrow et al. Adaptive inverse control based on linear and nonlinear adaptive filtering
JPH05127706A (en) Neural network type simulator
US6768927B2 (en) Control system
JPH0731522B2 (en) Adaptive controller
CN111624882A (en) Zero and differential game processing method for supply chain system based on reverse-thrust design method
JP2851901B2 (en) Learning control device
Lin et al. Neurofuzzy-model-following control of MIMO nonlinear systems
JPH0635510A (en) Model norm adaptive controller using neural network
JP2862308B2 (en) Controller adjustment method and adjustment system
JP2980421B2 (en) Controller using neural network model
Antony et al. Real-time nonlinear optimal control using neural networks
JPH11506553A (en) Dynamic process modeling equipment
JPH03182902A (en) Waste time compensation controller
Koivisto Minimum prediction error neural controller
JPH05333902A (en) Method and device for control by grading learning system
JPH04302302A (en) Two degree of free controller
Wang et al. Intelligent controller using neural network
JP3039154B2 (en) How to adjust the controller
JP3288066B2 (en) Learning processing method and system for network configuration data processing device
Warwich The control of dynamical systems by neural networks
Cajueiro et al. A chemical reactor benchmark for parallel adaptive control using feedforward neural networks
Sakai et al. Statistical learning method of discontinuous functions using simultaneous recurrent networks

Legal Events

Date Code Title Description
LAPS Cancellation because of no payment of annual fees