Dubravko Majetic - Academia.edu (original) (raw)

Papers by Dubravko Majetic

Control of autonomous robot motion in radial mass density field is presented. In that sense the r... more Control of autonomous robot motion in radial mass density field is presented. In that sense the robot motion is described as the function of the radial mass density parameters. The radial mass density field is between the maximal radial mass density and the minimal radial mass density. Between these two limited values one can use n points (n = 1, 2, . . . nmax) and calculate the related radial mass density for each point. The radial mass density is maximal at the minimal gravitational radius and minimal at the maximal gravitational radius. This conclusion is valid for Planck scale, but also for the scales that are less or higher of that one. Using the ratio of the Planck mass and Planck radius it is generated energy conservation constant with value κ = 0.99993392118. Further, in this theory it is possible to connect Planck’s and gravitational parameters as functions of the maximal (or minimal) radial mass density. In that sense the autonomous robot motion in radial mass density field is important for the control of the robot motion at micro and nano scales.

Automatika, 2002

Robot motion planning in multidimensional space is very time-consuming and requires a big model; ... more Robot motion planning in multidimensional space is very time-consuming and requires a big model; therefore, it is not very suitable for a real-time purpose. Limited space method (LSM) used here works with 3D real physical space (two translations and one rotation) and finds out a logical path (in the sense of the human solution faced with the same problem), rather than the optimal path (in the mathematical sense). Its main advantages are small model and short solution time. Although LSM is not as universal as the C-space, it has good potentiality for engineering applications.

The paper presents a student research work on interactive graphical interface for simulation and ... more The paper presents a student research work on interactive graphical interface for simulation and control of an experimental water level control setup. The setup has been designed for educational purposes of control system demonstration and implementation. A brief description of the experimental setup is given with the emphasis on two operating modes, simulation and control. Discrete-time PID control algorithm was integrated in water pump control loop. For both operating modes, the computer interactive graphical interface was built up. Both, mechanical and electrical parts of the setup have been realized in the Laboratory for automation and robotics. The communication with PC was established through serial port using MatLab software.

An attempt has been made to establish a nonlinear dynamic discrete-time neuron model. This dynami... more An attempt has been made to establish a nonlinear dynamic discrete-time neuron model. This dynamic discrete-time neuron disposes of local memory, in that it has dynamic states. Based on such neuron, a Dynamic Multi Layer Perceptron Neural Network is proposed for the identification of a dynamic discrete-time nonlinear system. To accelerate the convergence of proposed extended dynamic error back propagation learning algorithm, the momentum method is applied. The learning results are presented in terms that are insensitive to the learning data range and allow easy comparison with other learning algorithms, independent of machine architecture or simulator implementation.

Strojarstvo, 1991

Rad prikazuje dobivanje matematickog modela dinamike cirkulacijskog kruga u generatoru pare, line... more Rad prikazuje dobivanje matematickog modela dinamike cirkulacijskog kruga u generatoru pare, linearizaciju tog modela i sintezu modernih koncepata upravljanja tim visevelicinskim sistemom. Ukazuje se na ogranicenja i mogucnosti linearne teorije. Pokazuje se da je, koristeci rezultate dobivene simulacijom linearnog modela, moguce procijeniti znacenje mjerenja pojedinih fizickih velicina i dati odgovore na pitanja integriteta sistema upravljanja. Rezultati simulacije upucuju i na nužnost uvažavanja stvarnih nelinearnih karakteristika izvrsnih organa i na utjecaj tih uređaja na dinamiku cijelog upravljackog sistema. Pristup, metode i rezultati dobiveni u ovom radu mogu korisno poslužiti i u simulacijama posve razlicitih sistema u robotici, procesnoj tehnici, biotehnologiji ili pri upravljanju letjelicama.

Forecasting performances of feed-forward and recurrent neural networks (NN) trained with differen... more Forecasting performances of feed-forward and recurrent neural networks (NN) trained with different learning algorithms are analyzed and compared using the Mackey-Glass nonlinear chaotic time series. This system is a known benchmark test whose elements are hard to predict. Multilayer Perceptron NN was chosen as a feed-forward neural network because it is still the most commonly used network in financial forecasting models. It is compared with the modified version of the so-called Dynamic Multi-layer Perceptron NN characterized with a dynamic neuron model, i.e., Auto Regressive Moving Average filter built into the hidden layer neurons. Thus, every hidden layer neuron has the ability to process previous values of its own activity together with new input signals. The obtained results indicate satisfactory forecasting characteristics of both networks. However, recurrent NN was more accurate in practically all tests using less number of hidden layer neurons than the feed-forward NN. This study once again confirmed a great effectiveness and potential of dynamic neural networks in modeling and predicting highly nonlinear processes. Their application in the design of financial forecasting models is therefore most recommended.

IFAC Proceedings Volumes, Sep 1, 1997

A new nonconventional analytic fuzzy logic robot control synthesis is proposed. For this purpose ... more A new nonconventional analytic fuzzy logic robot control synthesis is proposed. For this purpose the following objectives are preferred and reached: (i) a new interpretation of the grade of membership or fuzziness of fuzzy control systems , (ii) a determination of a new analytic activation function, instead of using of min-max operators, (iii) a definition of a new analytic function that determines the positions of centres of output fuzzy sets, instead of definition of rule base, (iv) an introduction of an analytic defuzzification formula and Cv) an analytic fuzzy logic control synthesis of robot of RRTR-structure, using proposed analytic approach.

... Pregled bibliografske jedinice broj: 5792. Zbornik radova. Autori: Majetić, Dubravko; Novakov... more ... Pregled bibliografske jedinice broj: 5792. Zbornik radova. Autori: Majetić, Dubravko; Novaković, Branko; Oluić, Čedomir. Naslov: An approach to combination of fuzzy logic control and neural networks. ... IRB. Pripremili: Ivo Batistić i Jadranka Stojanovski. Dizajn: Studio8. ...

In this paper the nonlinear dynamic discrete-time neuron model, the so-called Dynamic Elementary ... more In this paper the nonlinear dynamic discrete-time neuron model, the so-called Dynamic Elementary Processor (DEP) is proposed. This dynamic neuron disposes of local memory, in that it has dynamic states. To accelerate the convergence of proposed extended dynamic error-back propagation learning algorithm, the adaptive neuron activation is applied. Instead of most popular unipolar and bipolar Sigmoidal neuron activation functions, the Gauss activation function with adaptive parameters is proposed. Based on the DEP neuron with adaptive activation function in hidden layer, and without Bias neuron for hidden layer, a Dynamic Multi Layer Neural Network is proposed and used for the identification of discrete-time nonlinear dynamic system.

... Bibliographic record number: 194396. Journal. Authors: Novaković, Branko; Majetić, Dubravko; ... more ... Bibliographic record number: 194396. Journal. Authors: Novaković, Branko; Majetić, Dubravko; Josip, Kasać; Danko, Brezak. Title: An analitic fuzzy logic robot control synthesis using a new profile of fuzzy sets. ... IRB. Made by: Ivo Batistić and Jadranka Stojanovski. Design: Studio8 ...

In this paper a modification of the nonlinear dynamic discrete-time neuron model, the so-called D... more In this paper a modification of the nonlinear dynamic discrete-time neuron model, the so-called Dynamic Elementary Processor (DEP), is proposed. DEP disposes of local memory, in that it has dynamic states. Instead of the most popular unipolar and bipolar Sigmoidal neuron activation functions, the Gauss activation function with adaptive parameters is applied. Based on the DEP neurons in hidden layer a modified dynamic neural network (MDNN) without any Bias neurons is proposed. For such neural network, the Error Back-Propagation and RPROP learning strategies are compared in solving of two benchmarks, the Glass-Mackey time series prediction and XOR classification problem.

Strojarstvo, 2005

The paper presents a real-time domain implementation of a robot fuzzy logic controller (FLC) with... more The paper presents a real-time domain implementation of a robot fuzzy logic controller (FLC) with a neural network. An adaptive and analytic FLC system synthesis of RRTR robot control, without any rule base, is implemented by Radial Basis Function (RBF) neural network. The self-organizing topology of RBF neural network with new one step learning procedure is established. The proposed neural network learning algorithm shows the great potential in solution of various regression tasks.

This paper presents contribution to the combined learning method for Radial Basis Function (RBF) ... more This paper presents contribution to the combined learning method for Radial Basis Function (RBF) neural network in the sense of introducing another way for the (RBF) network structure determination. Instead of using only K-means clustering method for disposing the network structure, the modified method based on neurone centre membership parameter beta_Ais proposed. In many cases this method leads toward better solutions with a special emphasis on approximation problems.

... Bibliographic record number: 179108. Journal. Authors: Novaković, Branko; Majetić, Dubravko; ... more ... Bibliographic record number: 179108. Journal. Authors: Novaković, Branko; Majetić, Dubravko; Kasać, Josip; Brezak, Danko. ... Copyright © 1997-2011. IRB. Made by: Ivo Batistić and Jadranka Stojanovski. Design: Studio8. Software: postgresql.

This work deals with the problem of potential field based mobile robot motion planning in unorgan... more This work deals with the problem of potential field based mobile robot motion planning in unorganised environment. The new approach, using a combination of negative gradient and vortex field based on Gauss potential functions is proposed. Radial Basis Function Neural Network (RBF Neural Network) learns the dependence between Gauss function parameters and velocity of mobile robot (or relative velocity between robot and obstacle in dynamical environment) ensuring passage between two closely spaced obstacles and smooth path condition for different mobile robot initial conditions. This approach overcomes some standard problems in classical potential field methods like local minima avoidance, problems of no passage between closely spaced obstacles, strong variation of the repulsive force near the minimum distance, and avoidance of moving obstacles. The method is illustrated on the example of mobile robot navigation between several closely spaced obstacles and example of avoidance of moving obstacle.

Kako neprestano svjedocimo rastu trenda automatizacije sustava koje svakodnevno koristimo, metode... more Kako neprestano svjedocimo rastu trenda automatizacije sustava koje svakodnevno koristimo, metode iz podrucja teorije automatskog upravljanja sve vise predstavljaju nezaobilazni dio inženjerske prakse, pa tako i inženjerske edukacije. U tom je svjetlu nastala i ova zbirka zadataka. Ona predstavlja presjek najznacajnijih metoda analize i sinteze sustava automatskog upravljanja ciji su smisao i primjena detaljno objasnjeni i ilustrirani nizom izloženih primjera. Metode su podijeljene u nekoliko cjelina koje pokrivaju podrucja analize i sinteze visevarijabilnih (multivarijabilnih) i jednovarijabilnih linearnih vremenski kontinuiranih i diskretnih sustava u prostoru stanja, te nelinearnih sustava. U želji da se olaksa shvacanje i jednostavnije sagleda potencijal primjene razmatranih metoda teorije automatskog upravljanja, uz uobicajene forme zadataka, izložen je i niz primjera konkretnih tehnickih sustava. Uz gotovo sve zadatke osim rjesenja priložen je i postupak rjesavanja. Pojedine su cjeline dodatno popracene kratkim uvodnim razmatranjima, a rjesenja citatelju približena dodatnim grafickim ilustracijama.

S obzirom na kontinuirani trend posvemasnje automatizacije sustava koji nas okružuju, metode iz p... more S obzirom na kontinuirani trend posvemasnje automatizacije sustava koji nas okružuju, metode iz podrucja teorije automatskog upravljanja predstavljaju nezaobilazni dio inženjerske edukacije, a njihov razvoj predstavlja istraživacki izazov koji vremenom dobiva sve vise na znacaju. U tom je smislu nastala i ova zbirka zadataka kao presjek najznacajnijih metoda analize i sinteze sustava automatskog upravljanja ciji su smisao i primjena detaljno objasnjeni u nizu izloženih primjera. Metode prikazane u nekoliko cjelina pokrivaju podrucja analize i sinteze jednovarijabilnih linearnih kontinuiranih i diskretnih sustava. Uz standardne opce forme zadataka, izložen je i niz primjera konkretnih tehnickih sustava, u nastojanju da se olaksa shvacanje i jednostavnije sagleda potencijal primjene razmatranih metoda iz teorije automatskog upravljanja. Uz gotovo sve zadatke, osim rjesenja, priložen je i postupak rjesavanja, a pojedine su cjeline jos i dodatno popracene kratkim uvodnim razmatranjima. Jednako tako brojni su zadaci popraceni dodatnim grafickim ilustracijama dobivenih rezultata.

Control of autonomous robot motion in radial mass density field is presented. In that sense the r... more Control of autonomous robot motion in radial mass density field is presented. In that sense the robot motion is described as the function of the radial mass density parameters. The radial mass density field is between the maximal radial mass density and the minimal radial mass density. Between these two limited values one can use n points (n = 1, 2, . . . nmax) and calculate the related radial mass density for each point. The radial mass density is maximal at the minimal gravitational radius and minimal at the maximal gravitational radius. This conclusion is valid for Planck scale, but also for the scales that are less or higher of that one. Using the ratio of the Planck mass and Planck radius it is generated energy conservation constant with value κ = 0.99993392118. Further, in this theory it is possible to connect Planck’s and gravitational parameters as functions of the maximal (or minimal) radial mass density. In that sense the autonomous robot motion in radial mass density field is important for the control of the robot motion at micro and nano scales.

Automatika, 2002

Robot motion planning in multidimensional space is very time-consuming and requires a big model; ... more Robot motion planning in multidimensional space is very time-consuming and requires a big model; therefore, it is not very suitable for a real-time purpose. Limited space method (LSM) used here works with 3D real physical space (two translations and one rotation) and finds out a logical path (in the sense of the human solution faced with the same problem), rather than the optimal path (in the mathematical sense). Its main advantages are small model and short solution time. Although LSM is not as universal as the C-space, it has good potentiality for engineering applications.

The paper presents a student research work on interactive graphical interface for simulation and ... more The paper presents a student research work on interactive graphical interface for simulation and control of an experimental water level control setup. The setup has been designed for educational purposes of control system demonstration and implementation. A brief description of the experimental setup is given with the emphasis on two operating modes, simulation and control. Discrete-time PID control algorithm was integrated in water pump control loop. For both operating modes, the computer interactive graphical interface was built up. Both, mechanical and electrical parts of the setup have been realized in the Laboratory for automation and robotics. The communication with PC was established through serial port using MatLab software.

An attempt has been made to establish a nonlinear dynamic discrete-time neuron model. This dynami... more An attempt has been made to establish a nonlinear dynamic discrete-time neuron model. This dynamic discrete-time neuron disposes of local memory, in that it has dynamic states. Based on such neuron, a Dynamic Multi Layer Perceptron Neural Network is proposed for the identification of a dynamic discrete-time nonlinear system. To accelerate the convergence of proposed extended dynamic error back propagation learning algorithm, the momentum method is applied. The learning results are presented in terms that are insensitive to the learning data range and allow easy comparison with other learning algorithms, independent of machine architecture or simulator implementation.

Strojarstvo, 1991

Rad prikazuje dobivanje matematickog modela dinamike cirkulacijskog kruga u generatoru pare, line... more Rad prikazuje dobivanje matematickog modela dinamike cirkulacijskog kruga u generatoru pare, linearizaciju tog modela i sintezu modernih koncepata upravljanja tim visevelicinskim sistemom. Ukazuje se na ogranicenja i mogucnosti linearne teorije. Pokazuje se da je, koristeci rezultate dobivene simulacijom linearnog modela, moguce procijeniti znacenje mjerenja pojedinih fizickih velicina i dati odgovore na pitanja integriteta sistema upravljanja. Rezultati simulacije upucuju i na nužnost uvažavanja stvarnih nelinearnih karakteristika izvrsnih organa i na utjecaj tih uređaja na dinamiku cijelog upravljackog sistema. Pristup, metode i rezultati dobiveni u ovom radu mogu korisno poslužiti i u simulacijama posve razlicitih sistema u robotici, procesnoj tehnici, biotehnologiji ili pri upravljanju letjelicama.

Forecasting performances of feed-forward and recurrent neural networks (NN) trained with differen... more Forecasting performances of feed-forward and recurrent neural networks (NN) trained with different learning algorithms are analyzed and compared using the Mackey-Glass nonlinear chaotic time series. This system is a known benchmark test whose elements are hard to predict. Multilayer Perceptron NN was chosen as a feed-forward neural network because it is still the most commonly used network in financial forecasting models. It is compared with the modified version of the so-called Dynamic Multi-layer Perceptron NN characterized with a dynamic neuron model, i.e., Auto Regressive Moving Average filter built into the hidden layer neurons. Thus, every hidden layer neuron has the ability to process previous values of its own activity together with new input signals. The obtained results indicate satisfactory forecasting characteristics of both networks. However, recurrent NN was more accurate in practically all tests using less number of hidden layer neurons than the feed-forward NN. This study once again confirmed a great effectiveness and potential of dynamic neural networks in modeling and predicting highly nonlinear processes. Their application in the design of financial forecasting models is therefore most recommended.

IFAC Proceedings Volumes, Sep 1, 1997

A new nonconventional analytic fuzzy logic robot control synthesis is proposed. For this purpose ... more A new nonconventional analytic fuzzy logic robot control synthesis is proposed. For this purpose the following objectives are preferred and reached: (i) a new interpretation of the grade of membership or fuzziness of fuzzy control systems , (ii) a determination of a new analytic activation function, instead of using of min-max operators, (iii) a definition of a new analytic function that determines the positions of centres of output fuzzy sets, instead of definition of rule base, (iv) an introduction of an analytic defuzzification formula and Cv) an analytic fuzzy logic control synthesis of robot of RRTR-structure, using proposed analytic approach.

... Pregled bibliografske jedinice broj: 5792. Zbornik radova. Autori: Majetić, Dubravko; Novakov... more ... Pregled bibliografske jedinice broj: 5792. Zbornik radova. Autori: Majetić, Dubravko; Novaković, Branko; Oluić, Čedomir. Naslov: An approach to combination of fuzzy logic control and neural networks. ... IRB. Pripremili: Ivo Batistić i Jadranka Stojanovski. Dizajn: Studio8. ...

In this paper the nonlinear dynamic discrete-time neuron model, the so-called Dynamic Elementary ... more In this paper the nonlinear dynamic discrete-time neuron model, the so-called Dynamic Elementary Processor (DEP) is proposed. This dynamic neuron disposes of local memory, in that it has dynamic states. To accelerate the convergence of proposed extended dynamic error-back propagation learning algorithm, the adaptive neuron activation is applied. Instead of most popular unipolar and bipolar Sigmoidal neuron activation functions, the Gauss activation function with adaptive parameters is proposed. Based on the DEP neuron with adaptive activation function in hidden layer, and without Bias neuron for hidden layer, a Dynamic Multi Layer Neural Network is proposed and used for the identification of discrete-time nonlinear dynamic system.

... Bibliographic record number: 194396. Journal. Authors: Novaković, Branko; Majetić, Dubravko; ... more ... Bibliographic record number: 194396. Journal. Authors: Novaković, Branko; Majetić, Dubravko; Josip, Kasać; Danko, Brezak. Title: An analitic fuzzy logic robot control synthesis using a new profile of fuzzy sets. ... IRB. Made by: Ivo Batistić and Jadranka Stojanovski. Design: Studio8 ...

In this paper a modification of the nonlinear dynamic discrete-time neuron model, the so-called D... more In this paper a modification of the nonlinear dynamic discrete-time neuron model, the so-called Dynamic Elementary Processor (DEP), is proposed. DEP disposes of local memory, in that it has dynamic states. Instead of the most popular unipolar and bipolar Sigmoidal neuron activation functions, the Gauss activation function with adaptive parameters is applied. Based on the DEP neurons in hidden layer a modified dynamic neural network (MDNN) without any Bias neurons is proposed. For such neural network, the Error Back-Propagation and RPROP learning strategies are compared in solving of two benchmarks, the Glass-Mackey time series prediction and XOR classification problem.

Strojarstvo, 2005

The paper presents a real-time domain implementation of a robot fuzzy logic controller (FLC) with... more The paper presents a real-time domain implementation of a robot fuzzy logic controller (FLC) with a neural network. An adaptive and analytic FLC system synthesis of RRTR robot control, without any rule base, is implemented by Radial Basis Function (RBF) neural network. The self-organizing topology of RBF neural network with new one step learning procedure is established. The proposed neural network learning algorithm shows the great potential in solution of various regression tasks.

This paper presents contribution to the combined learning method for Radial Basis Function (RBF) ... more This paper presents contribution to the combined learning method for Radial Basis Function (RBF) neural network in the sense of introducing another way for the (RBF) network structure determination. Instead of using only K-means clustering method for disposing the network structure, the modified method based on neurone centre membership parameter beta_Ais proposed. In many cases this method leads toward better solutions with a special emphasis on approximation problems.

... Bibliographic record number: 179108. Journal. Authors: Novaković, Branko; Majetić, Dubravko; ... more ... Bibliographic record number: 179108. Journal. Authors: Novaković, Branko; Majetić, Dubravko; Kasać, Josip; Brezak, Danko. ... Copyright © 1997-2011. IRB. Made by: Ivo Batistić and Jadranka Stojanovski. Design: Studio8. Software: postgresql.

This work deals with the problem of potential field based mobile robot motion planning in unorgan... more This work deals with the problem of potential field based mobile robot motion planning in unorganised environment. The new approach, using a combination of negative gradient and vortex field based on Gauss potential functions is proposed. Radial Basis Function Neural Network (RBF Neural Network) learns the dependence between Gauss function parameters and velocity of mobile robot (or relative velocity between robot and obstacle in dynamical environment) ensuring passage between two closely spaced obstacles and smooth path condition for different mobile robot initial conditions. This approach overcomes some standard problems in classical potential field methods like local minima avoidance, problems of no passage between closely spaced obstacles, strong variation of the repulsive force near the minimum distance, and avoidance of moving obstacles. The method is illustrated on the example of mobile robot navigation between several closely spaced obstacles and example of avoidance of moving obstacle.

Kako neprestano svjedocimo rastu trenda automatizacije sustava koje svakodnevno koristimo, metode... more Kako neprestano svjedocimo rastu trenda automatizacije sustava koje svakodnevno koristimo, metode iz podrucja teorije automatskog upravljanja sve vise predstavljaju nezaobilazni dio inženjerske prakse, pa tako i inženjerske edukacije. U tom je svjetlu nastala i ova zbirka zadataka. Ona predstavlja presjek najznacajnijih metoda analize i sinteze sustava automatskog upravljanja ciji su smisao i primjena detaljno objasnjeni i ilustrirani nizom izloženih primjera. Metode su podijeljene u nekoliko cjelina koje pokrivaju podrucja analize i sinteze visevarijabilnih (multivarijabilnih) i jednovarijabilnih linearnih vremenski kontinuiranih i diskretnih sustava u prostoru stanja, te nelinearnih sustava. U želji da se olaksa shvacanje i jednostavnije sagleda potencijal primjene razmatranih metoda teorije automatskog upravljanja, uz uobicajene forme zadataka, izložen je i niz primjera konkretnih tehnickih sustava. Uz gotovo sve zadatke osim rjesenja priložen je i postupak rjesavanja. Pojedine su cjeline dodatno popracene kratkim uvodnim razmatranjima, a rjesenja citatelju približena dodatnim grafickim ilustracijama.

S obzirom na kontinuirani trend posvemasnje automatizacije sustava koji nas okružuju, metode iz p... more S obzirom na kontinuirani trend posvemasnje automatizacije sustava koji nas okružuju, metode iz podrucja teorije automatskog upravljanja predstavljaju nezaobilazni dio inženjerske edukacije, a njihov razvoj predstavlja istraživacki izazov koji vremenom dobiva sve vise na znacaju. U tom je smislu nastala i ova zbirka zadataka kao presjek najznacajnijih metoda analize i sinteze sustava automatskog upravljanja ciji su smisao i primjena detaljno objasnjeni u nizu izloženih primjera. Metode prikazane u nekoliko cjelina pokrivaju podrucja analize i sinteze jednovarijabilnih linearnih kontinuiranih i diskretnih sustava. Uz standardne opce forme zadataka, izložen je i niz primjera konkretnih tehnickih sustava, u nastojanju da se olaksa shvacanje i jednostavnije sagleda potencijal primjene razmatranih metoda iz teorije automatskog upravljanja. Uz gotovo sve zadatke, osim rjesenja, priložen je i postupak rjesavanja, a pojedine su cjeline jos i dodatno popracene kratkim uvodnim razmatranjima. Jednako tako brojni su zadaci popraceni dodatnim grafickim ilustracijama dobivenih rezultata.