Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161150
K. G. Kirby
Inductive inference is the process of inferring a description of a function from a finite subset of its graph. Connectionist inductive inference typically involves gradient descent algorithms in weight space. When inferring functions of unbounded sequences such algorithms run on recurrent nets and become computationally expensive. A broader framework for inductive inference is presented, and it is shown that such problems admit a dual approach, which can be phrased in terms of the simulation-as-homomorphism perspective in systems theory. Whereas the usual approach adapts the dynamics of the net to match the dynamics of the target system, the dual approach keeps the dynamics fixed and learns a homomorphism from the net to the target. The latter technique is promising because of its efficiency and its direct applicability to learning by continuous nonconnectionist system, such as neural fields.<>
{"title":"Duality in neurocomputational inductive inference: a simulationist perspective","authors":"K. G. Kirby","doi":"10.1109/ICSYSE.1991.161150","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161150","url":null,"abstract":"Inductive inference is the process of inferring a description of a function from a finite subset of its graph. Connectionist inductive inference typically involves gradient descent algorithms in weight space. When inferring functions of unbounded sequences such algorithms run on recurrent nets and become computationally expensive. A broader framework for inductive inference is presented, and it is shown that such problems admit a dual approach, which can be phrased in terms of the simulation-as-homomorphism perspective in systems theory. Whereas the usual approach adapts the dynamics of the net to match the dynamics of the target system, the dual approach keeps the dynamics fixed and learns a homomorphism from the net to the target. The latter technique is promising because of its efficiency and its direct applicability to learning by continuous nonconnectionist system, such as neural fields.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"5 1‐2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133111063","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161107
J. D. Johnson, T. A. Grogan
The selectively attentive environmental learning system (SAELS), that is capable of formulating decision policies while operating under terminally applied, minimally descriptive, reinforcement feedback is discussed. This type of reinforcement signals only that the generated policy is correct, or incorrect, and provides no information on the closeness of the generated policy to the correct policy. SAELS uses the drive-reinforcement neuronal model that, through the predictive qualities of its learning, is capable of solving the temporal credit assignment problem that arises under these reinforcement conditions. It is shown that SAELS can generate the necessary decision policy to maneuver through a multi-intersection maze.<>
{"title":"The selectively attentive environmental learning system","authors":"J. D. Johnson, T. A. Grogan","doi":"10.1109/ICSYSE.1991.161107","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161107","url":null,"abstract":"The selectively attentive environmental learning system (SAELS), that is capable of formulating decision policies while operating under terminally applied, minimally descriptive, reinforcement feedback is discussed. This type of reinforcement signals only that the generated policy is correct, or incorrect, and provides no information on the closeness of the generated policy to the correct policy. SAELS uses the drive-reinforcement neuronal model that, through the predictive qualities of its learning, is capable of solving the temporal credit assignment problem that arises under these reinforcement conditions. It is shown that SAELS can generate the necessary decision policy to maneuver through a multi-intersection maze.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130646317","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161106
D. Hawkins, H. Adeli
A prototype knowledge-based expert system developed for design of composite floor systems in multistory buildings is discussed. The design system is in accordance with the 1986 American Institute of Steel Construction (AISC) load and resistance factor design (LRFD) specification. Design options include design for practical minimum cost or for minimum depth. The artificial intelligence (AI) technology complements the traditional numerical processing in automating the complex problem of integrated engineering design and developing intelligent software packages that can interact with the user at a high level.<>
{"title":"Knowledge-based approach for composite floor system design","authors":"D. Hawkins, H. Adeli","doi":"10.1109/ICSYSE.1991.161106","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161106","url":null,"abstract":"A prototype knowledge-based expert system developed for design of composite floor systems in multistory buildings is discussed. The design system is in accordance with the 1986 American Institute of Steel Construction (AISC) load and resistance factor design (LRFD) specification. Design options include design for practical minimum cost or for minimum depth. The artificial intelligence (AI) technology complements the traditional numerical processing in automating the complex problem of integrated engineering design and developing intelligent software packages that can interact with the user at a high level.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132611984","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161137
E. Mu, J. T. Cain
Based on strict matching conditions and a linear parameterization assumption, a certainty equivalence control law to pseudolinearize nonlinear systems in the presence of parametric uncertainty is designed. Stability of the origin and state regulation are guaranteed in the vicinity of any equilibrium point of the nonlinear system. These are the first results reported for the pseudolinearization of a class of nonlinear systems with uncertain parameter knowledge.<>
{"title":"Robust pseudolinearization of nonlinear systems","authors":"E. Mu, J. T. Cain","doi":"10.1109/ICSYSE.1991.161137","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161137","url":null,"abstract":"Based on strict matching conditions and a linear parameterization assumption, a certainty equivalence control law to pseudolinearize nonlinear systems in the presence of parametric uncertainty is designed. Stability of the origin and state regulation are guaranteed in the vicinity of any equilibrium point of the nonlinear system. These are the first results reported for the pseudolinearization of a class of nonlinear systems with uncertain parameter knowledge.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"232 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133971255","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161148
J. Wang, J. Jean
A set of neural networks is used in a two-stage character recognition system to resolve the confusion among similar characters. A snowball training algorithm is proposed to remedy the convergence problem encountered by backpropagation training. The algorithm is shown to be effective in reducing the number of hidden units and the training time. To further improve the network's generalization capability, a smoothing operation is incorporated into the snowball training. Experimental results confirm the effectiveness of the approach.<>
{"title":"Automatic rule generation for machine printed character recognition using multiple neural networks","authors":"J. Wang, J. Jean","doi":"10.1109/ICSYSE.1991.161148","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161148","url":null,"abstract":"A set of neural networks is used in a two-stage character recognition system to resolve the confusion among similar characters. A snowball training algorithm is proposed to remedy the convergence problem encountered by backpropagation training. The algorithm is shown to be effective in reducing the number of hidden units and the training time. To further improve the network's generalization capability, a smoothing operation is incorporated into the snowball training. Experimental results confirm the effectiveness of the approach.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"283 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114379914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161062
A. Sarantopoulos, T. Hartley
A modeling approach for linear isentropic flow systems based on the quasi-one-dimensional Euler equations of nonviscous, compressible flow is presented. Such systems are representative of certain high-speed propulsion systems. Accurate models useful in control system studies are developed. A supersonic inlet is considered, and the resulting set of partial differential equations with boundary conditions is solved for a linear transfer matrix using Laplace transforms.<>
{"title":"Modeling of linear isentropic flow systems","authors":"A. Sarantopoulos, T. Hartley","doi":"10.1109/ICSYSE.1991.161062","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161062","url":null,"abstract":"A modeling approach for linear isentropic flow systems based on the quasi-one-dimensional Euler equations of nonviscous, compressible flow is presented. Such systems are representative of certain high-speed propulsion systems. Accurate models useful in control system studies are developed. A supersonic inlet is considered, and the resulting set of partial differential equations with boundary conditions is solved for a linear transfer matrix using Laplace transforms.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114429481","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161083
M. Bodson, J. Chiasson
Three approaches for nonlinear control are compared; feedback linearization, generalized controller canonical forms, and input/output linearization. To establish a framework for comparison, the example of a shunt DC motor is studied. It was found that the conditions for applying the method of feedback linearization were not satisfied; the full 3D state-space system could not be linearized by state transformation. It was found that, after eliminating a state and therefore reducing the dimension from three to two, the system became feedback linearizable. In other words, the dynamics associated with the remaining two states could be linearized by change of coordinates and state feedback. A similar situation occurred with the input/output linearization, except that the dynamics that were linearized were those associated with the rotor position and velocity, while the remaining dynamics (associated with the field current) were made unobservable.<>
{"title":"Nonlinear and adaptive control of a shunt DC motor","authors":"M. Bodson, J. Chiasson","doi":"10.1109/ICSYSE.1991.161083","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161083","url":null,"abstract":"Three approaches for nonlinear control are compared; feedback linearization, generalized controller canonical forms, and input/output linearization. To establish a framework for comparison, the example of a shunt DC motor is studied. It was found that the conditions for applying the method of feedback linearization were not satisfied; the full 3D state-space system could not be linearized by state transformation. It was found that, after eliminating a state and therefore reducing the dimension from three to two, the system became feedback linearizable. In other words, the dynamics associated with the remaining two states could be linearized by change of coordinates and state feedback. A similar situation occurred with the input/output linearization, except that the dynamics that were linearized were those associated with the rotor position and velocity, while the remaining dynamics (associated with the field current) were made unobservable.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115761103","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161122
L. Fortuna, S. Graziani, S. Baglio, G. Nunnari
A parallel strategy to automatically design a fuzzy controller, i.e. to obtain the rules that form the system controller, is discussed. The strategy is proposed in accordance with the natural parallelism of the cell-to-cell mapping approach, and some improvements are suggested. To make the computation algorithm faster and allow the controller design, even if complex nonlinear systems are considered, a natural modeling approach is suggested. The procedure has been implemented on a multiprocessor architecture based on transputers. It has been applied to design several fuzzy controllers: the results using the fuzzy controller of a nonlinear system confirm the suitability of the approach in comparison with classical regulators.<>
{"title":"Improvements in fuzzy controller design","authors":"L. Fortuna, S. Graziani, S. Baglio, G. Nunnari","doi":"10.1109/ICSYSE.1991.161122","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161122","url":null,"abstract":"A parallel strategy to automatically design a fuzzy controller, i.e. to obtain the rules that form the system controller, is discussed. The strategy is proposed in accordance with the natural parallelism of the cell-to-cell mapping approach, and some improvements are suggested. To make the computation algorithm faster and allow the controller design, even if complex nonlinear systems are considered, a natural modeling approach is suggested. The procedure has been implemented on a multiprocessor architecture based on transputers. It has been applied to design several fuzzy controllers: the results using the fuzzy controller of a nonlinear system confirm the suitability of the approach in comparison with classical regulators.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"136 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116382795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161154
R. Paroda, A. Kumar
An approach to design an optimal reduced-order estimator (ROE) directly from a full-order system is proposed. The cost functional used is quadratic with positive definite weighting. The approach is based on the minimization of the associated Hamiltonian to obtain optimal ROE parameters. Examination of the Hessian matrix reveals extra conditions for ROE parameters to be optimal.<>
{"title":"Optimal reduced-order estimator design for discrete, time-varying systems","authors":"R. Paroda, A. Kumar","doi":"10.1109/ICSYSE.1991.161154","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161154","url":null,"abstract":"An approach to design an optimal reduced-order estimator (ROE) directly from a full-order system is proposed. The cost functional used is quadratic with positive definite weighting. The approach is based on the minimization of the associated Hamiltonian to obtain optimal ROE parameters. Examination of the Hessian matrix reveals extra conditions for ROE parameters to be optimal.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116454079","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.1109/ICSYSE.1991.161095
K. Tomiyama, F. Honda, H. Shinoda, J. Tanaka
An integrated testbed for the study of a smart robot control system based on a UNIX workstation is discussed. It has a VME-bus-based real-time controller which provides data acquisition capability to the system. The SCARA-type manipulator is powered by a set of pulse-controlled servo drivers. Image data are gathered and processed by the combination of image processing hardware and a CCD camera. A UNIX workstation is designated to provide intelligence to the system by understanding the environment of the robot through image data. It also provides online data analysis capability. The system was checked by an object following experiment where the robot chases a pair of alternatively lit LEDs. Data acquisition and analysis capability was built into the system. A mathematical model of the manipulator was obtained. The model was used in designing a digital compensator. The compensator was implemented in the form of a software servo and successfully suppressed oscillatory response of the manipulator.<>
{"title":"A robot control system based on a UNIX work-station","authors":"K. Tomiyama, F. Honda, H. Shinoda, J. Tanaka","doi":"10.1109/ICSYSE.1991.161095","DOIUrl":"https://doi.org/10.1109/ICSYSE.1991.161095","url":null,"abstract":"An integrated testbed for the study of a smart robot control system based on a UNIX workstation is discussed. It has a VME-bus-based real-time controller which provides data acquisition capability to the system. The SCARA-type manipulator is powered by a set of pulse-controlled servo drivers. Image data are gathered and processed by the combination of image processing hardware and a CCD camera. A UNIX workstation is designated to provide intelligence to the system by understanding the environment of the robot through image data. It also provides online data analysis capability. The system was checked by an object following experiment where the robot chases a pair of alternatively lit LEDs. Data acquisition and analysis capability was built into the system. A mathematical model of the manipulator was obtained. The model was used in designing a digital compensator. The compensator was implemented in the form of a software servo and successfully suppressed oscillatory response of the manipulator.<<ETX>>","PeriodicalId":250037,"journal":{"name":"IEEE 1991 International Conference on Systems Engineering","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128154304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：