http://sedici.unlp.edu.ar:80/handle/10915/183318 2026-04-13T23:02:33Z 2026-04-13T23:02:33Z Imada, Akira http://sedici.unlp.edu.ar:80/handle/10915/185600 2025-10-07T20:10:10Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) We simulate an associative memory model using spiking neurons instead of McCulloch-Pitts neurons. To store a set of patterns, we employ a Hebbian-like learning algorithm. The learning behavior, however, is somewhat of a different one of the traditional Hopfield model. To study the difference we explore the fitness landscape defined on synaptic weights space when they evolve searching for the optimal learning.

2003-01-01T00:00:00Z We simulate an associative memory model using spiking neurons instead of McCulloch-Pitts neurons. To store a set of patterns, we employ a Hebbian-like learning algorithm. The learning behavior, however, is somewhat of a different one of the traditional Hopfield model. To study the difference we explore the fitness landscape defined on synaptic weights space when they evolve searching for the optimal learning. Huapaya, Constanza R. Arona, Graciela M. Lizarralde, Francisco A. http://sedici.unlp.edu.ar:80/handle/10915/185221 2025-09-29T20:09:47Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) Our long term aim is to reach an easily available, flexible and complete instructor model of an Intelligent Tutoring System. In order to get our objective, we propose a new instructor model where intentions of actual instructors are shaped into adaptive tutoring strategies. The key idea involves an instructor acting as she if was a knowledge engineer, proposing goals and softgoals, and choosing resources and tasks in order to reach them through an ITS. We present a preliminary conceptual approach to an explicit instructor model. When an author/instructor delivers to system her personal experience, ITS captures these individual features. Each instance of the model would be an instructor with a clear profile acting according to specific behavior. So, a student would choose a personal instructor that matches better her learning style.

2003-01-01T00:00:00Z Our long term aim is to reach an easily available, flexible and complete instructor model of an Intelligent Tutoring System. In order to get our objective, we propose a new instructor model where intentions of actual instructors are shaped into adaptive tutoring strategies. The key idea involves an instructor acting as she if was a knowledge engineer, proposing goals and softgoals, and choosing resources and tasks in order to reach them through an ITS. We present a preliminary conceptual approach to an explicit instructor model. When an author/instructor delivers to system her personal experience, ITS captures these individual features. Each instance of the model would be an instructor with a clear profile acting according to specific behavior. So, a student would choose a personal instructor that matches better her learning style. Schiaffino, Silvia Amandi, Analía http://sedici.unlp.edu.ar:80/handle/10915/185220 2025-09-29T20:09:48Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) Interface agents are computer programs that learn users’ preferences to provide them personalized assistance with their computer-based tasks. In order to personalize the interaction with users, interface agents must learn how to best interact with each user and how to provide them assistance of the right sort at the right time. Particularly, an interface agent has to discover when the user needs a suggestion to solve a problem, when he requires only a warning about it, when he wants the agent to execute an action and when he wants the agent to do just nothing. In this work we propose a learning algorithm, named WATSON, to tackle this problem. The WATSON algorithm enables an interface agent to adapt its behavior and its interaction with a user to the user’s assistance requirements. Our algorithm uses association rules to discover associations among problem situations and a user’s assistance requirements in a given application domain.

2003-01-01T00:00:00Z Interface agents are computer programs that learn users’ preferences to provide them personalized assistance with their computer-based tasks. In order to personalize the interaction with users, interface agents must learn how to best interact with each user and how to provide them assistance of the right sort at the right time. Particularly, an interface agent has to discover when the user needs a suggestion to solve a problem, when he requires only a warning about it, when he wants the agent to execute an action and when he wants the agent to do just nothing. In this work we propose a learning algorithm, named WATSON, to tackle this problem. The WATSON algorithm enables an interface agent to adapt its behavior and its interaction with a user to the user’s assistance requirements. Our algorithm uses association rules to discover associations among problem situations and a user’s assistance requirements in a given application domain. Araya, Juan Francisco Cipriano, Aldo http://sedici.unlp.edu.ar:80/handle/10915/185218 2025-09-29T20:09:48Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) In this paper, a parameter estimation problem for a Takagi-Sugeno (TS) fuzzy dynamical system is formulated under the assumption that the premises in the membership functions are known. A linear expression in consequent parameters is obtained under this assumption. If the system is timevarying (TV), the parameters can be determined by recursive estimation techniques. As an alternate approach, the use of multi-resolution wavelets is proposed. Furthermore, a parameter estimation toolbox for fuzzy dynamical models is developed which is then applied to a simple example and to the Mackey-Glass chaotic time series.

2003-01-01T00:00:00Z In this paper, a parameter estimation problem for a Takagi-Sugeno (TS) fuzzy dynamical system is formulated under the assumption that the premises in the membership functions are known. A linear expression in consequent parameters is obtained under this assumption. If the system is timevarying (TV), the parameters can be determined by recursive estimation techniques. As an alternate approach, the use of multi-resolution wavelets is proposed. Furthermore, a parameter estimation toolbox for fuzzy dynamical models is developed which is then applied to a simple example and to the Mackey-Glass chaotic time series. Flores, Cecília Gluz, João Seixas, Louise Vicari, Rosa Coelho, Helder http://sedici.unlp.edu.ar:80/handle/10915/185217 2025-09-29T20:09:48Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) AMPLIA is an Intelligent Learning Multi-Agent Environment. It is designed to support training of diagnostic reasoning and modeling of domains with complex and uncertain knowledge. AMPLIA focuses on the medical area, where learner’s modeling tasks will consist of creating a Bayesian network for a problem the system will present. A pedagogic negotiation process (managed by an intelligent Mediator Agent) will treat the differences of topology and probability distribution between the model the learner built and the one built-in in the system. That negotiation process occurs between the agents that represent the expert knowledge domain and the agent that represents the learner knowledge. The possibility of using Bayesian networks to create knowledge representation allows the learner to visualize his/her ideas organization, create and test hypothesis.

2003-01-01T00:00:00Z AMPLIA is an Intelligent Learning Multi-Agent Environment. It is designed to support training of diagnostic reasoning and modeling of domains with complex and uncertain knowledge. AMPLIA focuses on the medical area, where learner’s modeling tasks will consist of creating a Bayesian network for a problem the system will present. A pedagogic negotiation process (managed by an intelligent Mediator Agent) will treat the differences of topology and probability distribution between the model the learner built and the one built-in in the system. That negotiation process occurs between the agents that represent the expert knowledge domain and the agent that represents the learner knowledge. The possibility of using Bayesian networks to create knowledge representation allows the learner to visualize his/her ideas organization, create and test hypothesis. Cofman, Fernando Sáez, Doris Gómez, Juan C. http://sedici.unlp.edu.ar:80/handle/10915/185216 2025-09-29T20:09:49Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) This paper presents the results of the implementation of Fuzzy PID control strategies on a complex, critically-stable system, namely the PID-Pong machine. Two different fuzzy algorithms are implemented in real-time and compared to a linear PID controller, used as the evaluation basis. Conclusions are established based on regulation, implementation and tuning indicators

2003-01-01T00:00:00Z This paper presents the results of the implementation of Fuzzy PID control strategies on a complex, critically-stable system, namely the PID-Pong machine. Two different fuzzy algorithms are implemented in real-time and compared to a linear PID controller, used as the evaluation basis. Conclusions are established based on regulation, implementation and tuning indicators Destéfanis, Eduardo A. Canali, Luis R. Steiner, Guillermo M. http://sedici.unlp.edu.ar:80/handle/10915/185214 2025-09-29T20:09:49Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) The Hough Transform is a robust algorithm, intended to detect lines, circles or even more complex shapes within an image. A weakness of the algorithm is that it requires an important processing time, in particular if the shape to be detected is not a straight line. In many practical applications this constraint could not be acceptable. A solution to this problem has been called the Fast Hough Transform (FHT) -[8] and [5]. The FHT approaches addresses the problem using specialized parallel hardware. Instead of this, this paper proposes an algorithmic approach for circle detection, which yields an acceptable processing time, without the need of any specialized hardware. To reach this goal the inversion transform [2] is used.

2003-01-01T00:00:00Z The Hough Transform is a robust algorithm, intended to detect lines, circles or even more complex shapes within an image. A weakness of the algorithm is that it requires an important processing time, in particular if the shape to be detected is not a straight line. In many practical applications this constraint could not be acceptable. A solution to this problem has been called the Fast Hough Transform (FHT) -[8] and [5]. The FHT approaches addresses the problem using specialized parallel hardware. Instead of this, this paper proposes an algorithmic approach for circle detection, which yields an acceptable processing time, without the need of any specialized hardware. To reach this goal the inversion transform [2] is used. Katz, Román Delrieux, Claudio http://sedici.unlp.edu.ar:80/handle/10915/185212 2025-09-29T20:09:49Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) In this work we present a robot path planning architecture based on Lee's routing algorithm. This algorithm was originally conceived to obtain minimal conductor nets among terminals on VLSI circuits. In our system it provides a simple and robust mechanism to compute near to shortest free paths in two-dimensional configuration environments. We incorporate this algorithm in a planning system, and the resulting architecture can be used to generate trajectories in order to guide a mobile agent from an initial point to a final (static or moving) position of the working space, avoiding not only static but also moving obstacles. The guidance engine is implemented as a rule-based version of Lee's algorithm, embedded in a multi-paradigm software system that performs logical inferences in a Prolog component, whose results are provided to an imperative and numerical processing compiled component. This allows a neat factorization of the functionality of the system: high level representation of the navigation algorithm is implemented through its logical formulation, and easy and natural means to acquire the layout of the environment with computer vision and image processing techniques is implemented via numerical programming. The same imperative approach is also natural to provide 2D or 3D visualization and simulation features to aid to understand the ongoing process. Thus, our architecture features both the speed and versatility of a visual language application, and the abstraction level and modularity of a logical description.

2003-01-01T00:00:00Z In this work we present a robot path planning architecture based on Lee's routing algorithm. This algorithm was originally conceived to obtain minimal conductor nets among terminals on VLSI circuits. In our system it provides a simple and robust mechanism to compute near to shortest free paths in two-dimensional configuration environments. We incorporate this algorithm in a planning system, and the resulting architecture can be used to generate trajectories in order to guide a mobile agent from an initial point to a final (static or moving) position of the working space, avoiding not only static but also moving obstacles. The guidance engine is implemented as a rule-based version of Lee's algorithm, embedded in a multi-paradigm software system that performs logical inferences in a Prolog component, whose results are provided to an imperative and numerical processing compiled component. This allows a neat factorization of the functionality of the system: high level representation of the navigation algorithm is implemented through its logical formulation, and easy and natural means to acquire the layout of the environment with computer vision and image processing techniques is implemented via numerical programming. The same imperative approach is also natural to provide 2D or 3D visualization and simulation features to aid to understand the ongoing process. Thus, our architecture features both the speed and versatility of a visual language application, and the abstraction level and modularity of a logical description. Tosini, Marcelo Alejandro http://sedici.unlp.edu.ar:80/handle/10915/185211 2025-09-29T20:09:50Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) A design and development tool to achieve artificial neural networks (ANN) implemented in a Field Programmable Gate Array (FPGA), is presented in this article. Its main components and functionality are thoroughly described. This tool, called NNGen, allows constructing digital ANN, which are easily programmable selecting different parameters. The output of such programming task is directly VHDL code to be ported to, in principle, any chip of the mentioned kind. A case study of a multilayer perceptron applied to weather forecast that was fully designed with NNGen, is also analyzed to obtain some conclusions.

2003-01-01T00:00:00Z A design and development tool to achieve artificial neural networks (ANN) implemented in a Field Programmable Gate Array (FPGA), is presented in this article. Its main components and functionality are thoroughly described. This tool, called NNGen, allows constructing digital ANN, which are easily programmable selecting different parameters. The output of such programming task is directly VHDL code to be ported to, in principle, any chip of the mentioned kind. A case study of a multilayer perceptron applied to weather forecast that was fully designed with NNGen, is also analyzed to obtain some conclusions. Levrini, Aldo Cipriano, Aldo http://sedici.unlp.edu.ar:80/handle/10915/185210 2025-09-29T20:09:50Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) This paper presents a comparison of three nonlinear models used for residual generation. The residual generation is part of a simple fault detection and diagnostics scheme applied to the Pendubot dynamic system operating in closed loop. The compared models are: Hammerstein, neural NARMAX, and Takagi-Sugeno Fuzzy models.

2003-01-01T00:00:00Z This paper presents a comparison of three nonlinear models used for residual generation. The residual generation is part of a simple fault detection and diagnostics scheme applied to the Pendubot dynamic system operating in closed loop. The compared models are: Hammerstein, neural NARMAX, and Takagi-Sugeno Fuzzy models. Katz, Román Delrieux, Claudio http://sedici.unlp.edu.ar:80/handle/10915/185209 2025-09-29T20:09:51Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) Border extraction is an important procedure associated with recognition and interpretation tasks in digital image processing and computer vision. Since local processing schemes (i.e., spatial filltering) are not appropriate for border extraction in noisy images, global and intelligent mechanisms are required to deal with this situation. Some heuristic algorithms that apply jointly local operators and some kind of optimization criterion have been successfully applied to treat images corrupted by additive noise. However, the behavior of these mechanisms is considerably undermined when managing images with multiplicative (speckle) noise. In this paper we present a gradient-based evolutionary algorithm that can achieve convenient boundary extraction in digital images with additive noise, and that can be successfully extended to operate with non-additive noisy images as well.

2003-01-01T00:00:00Z Border extraction is an important procedure associated with recognition and interpretation tasks in digital image processing and computer vision. Since local processing schemes (i.e., spatial filltering) are not appropriate for border extraction in noisy images, global and intelligent mechanisms are required to deal with this situation. Some heuristic algorithms that apply jointly local operators and some kind of optimization criterion have been successfully applied to treat images corrupted by additive noise. However, the behavior of these mechanisms is considerably undermined when managing images with multiplicative (speckle) noise. In this paper we present a gradient-based evolutionary algorithm that can achieve convenient boundary extraction in digital images with additive noise, and that can be successfully extended to operate with non-additive noisy images as well. Pérez Santángelo, Hugo http://sedici.unlp.edu.ar:80/handle/10915/185208 2025-09-29T20:09:51Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) The agents metaphor, expressed theoretically in the Distributed Artificial Intelligence field of Multi-Agent Systems, has been awakening the interest of the commercial software companies, because new emergent technologies like e-commerce follow the same metaphor from the consumer's perspective. Researchers in that field has been developing useful models that can be used to construct commercial servers, but the computational complexity problem and the lack of general models and architectures, sometimes make the technological transfer between researchers and commercial companies impracticable. State of the art is to put the focus on ontologies in order to create more general business models, but there are some software engineering issues related to the software construction that ontology does not cover. This paper offers a proposed solution to that problem, including conceptual aspects, and necessaries methods to model commercial Multi-Agent Systems, using fuzzy logic and evolutionary algorithms.

2003-01-01T00:00:00Z The agents metaphor, expressed theoretically in the Distributed Artificial Intelligence field of Multi-Agent Systems, has been awakening the interest of the commercial software companies, because new emergent technologies like e-commerce follow the same metaphor from the consumer's perspective. Researchers in that field has been developing useful models that can be used to construct commercial servers, but the computational complexity problem and the lack of general models and architectures, sometimes make the technological transfer between researchers and commercial companies impracticable. State of the art is to put the focus on ontologies in order to create more general business models, but there are some software engineering issues related to the software construction that ontology does not cover. This paper offers a proposed solution to that problem, including conceptual aspects, and necessaries methods to model commercial Multi-Agent Systems, using fuzzy logic and evolutionary algorithms. Soares Xavier, Maria Ester Varejão, Flávio Miguel http://sedici.unlp.edu.ar:80/handle/10915/185108 2025-09-25T20:10:21Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) In this work we present an extension of the parametric network representation language used by Active Documents models. The extension incorporates a new kind of parameter to the representation language. This new kind of parameter enables the reuse of a task-especific library of problem solving components implementing automatic search in Active Documents Systems. In order to illustrate the use of this new parameter, the Hill Climbing, Propose and Backtracking, Complete the Model Then Revise and Extend The Model Then Revise methods were implemented. These methods were used for solving an office allocation problem. The knowledge modeling technique used here allows the knowledge engineer to configure new methods using the library components. The knowledge engineer can also include new components to the library

2003-01-01T00:00:00Z In this work we present an extension of the parametric network representation language used by Active Documents models. The extension incorporates a new kind of parameter to the representation language. This new kind of parameter enables the reuse of a task-especific library of problem solving components implementing automatic search in Active Documents Systems. In order to illustrate the use of this new parameter, the Hill Climbing, Propose and Backtracking, Complete the Model Then Revise and Extend The Model Then Revise methods were implemented. These methods were used for solving an office allocation problem. The knowledge modeling technique used here allows the knowledge engineer to configure new methods using the library components. The knowledge engineer can also include new components to the library Granitto, Pablo Miguel Verdes, Pablo Fabián Ceccatto, Hermenegildo Alejandro http://sedici.unlp.edu.ar:80/handle/10915/185106 2025-09-25T20:10:21Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) We explore the feasibility of implementing fast and reliable computer-based systems for the automatic identification of weed seeds from color and black and white images. Seeds size, shape, color and texture characteristics are obtained by standard image-processing techniques, and their discriminating power as classification features is assessed. These investigations are performed on a database much larger than those used in previous studies, containing 10,310 images of 236 different weed species. We consider the implementation of a simple Bayesian approach (naïve Bayes classifier) and (single and bagged) artificial neural networks for seed identification. Our results indicate that the naïve Bayes classifier based on an adequately selected set of classification features has an excellent performance, competitive with that of the comparatively more sophisticated neural network approach. In addition, we discuss the possibility of using only morphological and textural characteristics as classification features, which would reduce the operational complexity and hardware cost of a commercial system since they can be obtained from black and white images. According to our results, under particular operational conditions this would result in a relatively small loss in performance when compared to the implementation based on color images.

2003-01-01T00:00:00Z We explore the feasibility of implementing fast and reliable computer-based systems for the automatic identification of weed seeds from color and black and white images. Seeds size, shape, color and texture characteristics are obtained by standard image-processing techniques, and their discriminating power as classification features is assessed. These investigations are performed on a database much larger than those used in previous studies, containing 10,310 images of 236 different weed species. We consider the implementation of a simple Bayesian approach (naïve Bayes classifier) and (single and bagged) artificial neural networks for seed identification. Our results indicate that the naïve Bayes classifier based on an adequately selected set of classification features has an excellent performance, competitive with that of the comparatively more sophisticated neural network approach. In addition, we discuss the possibility of using only morphological and textural characteristics as classification features, which would reduce the operational complexity and hardware cost of a commercial system since they can be obtained from black and white images. According to our results, under particular operational conditions this would result in a relatively small loss in performance when compared to the implementation based on color images. Cofre, Patricio Cipriano, Aldo http://sedici.unlp.edu.ar:80/handle/10915/185094 2025-09-25T20:10:22Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) This work compares three nonlinear models for a three-tank hydraulic system. From each model, nonlinear predictors are designed for the generation of residuals representing the fault conditions. The residuals feed three different statistical inference modules to be compared in search for the best solution for fault detection and diagnosis.

2003-01-01T00:00:00Z This work compares three nonlinear models for a three-tank hydraulic system. From each model, nonlinear predictors are designed for the generation of residuals representing the fault conditions. The residuals feed three different statistical inference modules to be compared in search for the best solution for fault detection and diagnosis. Acosta, Hector N. Fernández León, José A. http://sedici.unlp.edu.ar:80/handle/10915/185086 2025-09-25T20:10:22Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) RobotCup is an international competition designed to promote Artificial Intelligence (AI) and intelligent robotic research through a standard problem: a soccer game where a wide range of technologies can be integrated. This article shows, in a general way, an architecture proposed for controlling a robot soccer team called INCASoT. The team has been designed for its presentation in the CAFR-2003 competition (UBA) in Middle League Simurosot category. A brief description of control’s architecture is presented together with definitions of modules.

2003-01-01T00:00:00Z RobotCup is an international competition designed to promote Artificial Intelligence (AI) and intelligent robotic research through a standard problem: a soccer game where a wide range of technologies can be integrated. This article shows, in a general way, an architecture proposed for controlling a robot soccer team called INCASoT. The team has been designed for its presentation in the CAFR-2003 competition (UBA) in Middle League Simurosot category. A brief description of control’s architecture is presented together with definitions of modules. Acosta, Gerardo Gabriel Curti, Hugo Calvo, Oscar Mochnacs, Javier http://sedici.unlp.edu.ar:80/handle/10915/184933 2025-09-24T20:10:18Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) This article describes considerations and design hypothesis in the implementation of the strategy to generate trajectories for an autonomous vehicle. The application problem consists of an autonomous underwater vehicle (AUV) tracking a pipeline in the seabed. To solve this problem, a real time expert system (named EN4AUV) was proposed to be included in the on-board AUV central processing unit. EN4AUV takes trajectory control decisions based on a number of variables, arranged around the concept of scenarios. For different scenarios, the expert system is able to suggest trajectories. Although this work is still undergoing, in this paper some incipient results over computer simulation are shown. The article is concluded with some conclusions and a snapshot of future work.

2003-01-01T00:00:00Z This article describes considerations and design hypothesis in the implementation of the strategy to generate trajectories for an autonomous vehicle. The application problem consists of an autonomous underwater vehicle (AUV) tracking a pipeline in the seabed. To solve this problem, a real time expert system (named EN4AUV) was proposed to be included in the on-board AUV central processing unit. EN4AUV takes trajectory control decisions based on a number of variables, arranged around the concept of scenarios. For different scenarios, the expert system is able to suggest trajectories. Although this work is still undergoing, in this paper some incipient results over computer simulation are shown. The article is concluded with some conclusions and a snapshot of future work. Prade Nadaletti, Leandro Varejão, Flávio Miguel http://sedici.unlp.edu.ar:80/handle/10915/184932 2025-09-24T20:10:19Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) Several computational approaches have been applied to support problem solving on oil exploration activities in deep waters. While the focus of these works consists of finding effective methods that bring solutions to the problem, this article describes a new research perspective. It is based on the idea that the adoption of principles that improve the interaction between the user and the computational system can lead to a more efficient final result than the ones obtained by any of those entities (user or computational system) alone.

2003-01-01T00:00:00Z Several computational approaches have been applied to support problem solving on oil exploration activities in deep waters. While the focus of these works consists of finding effective methods that bring solutions to the problem, this article describes a new research perspective. It is based on the idea that the adoption of principles that improve the interaction between the user and the computational system can lead to a more efficient final result than the ones obtained by any of those entities (user or computational system) alone. Drozdowicz, B. Hadad, A. Evin, D. Böhm, C. Chiott, O. http://sedici.unlp.edu.ar:80/handle/10915/184931 2025-09-24T20:10:19Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) In this paper aspects related with the methodological design of a multi-agent system to support both supervision of anesthesiological processes and anesthesiologists decisions are approached. This work describes how it was arrived to a multi-agent architectural design complementing the Gaia methodology with an approach for roles identification. Then, it shows the ActionPlanSupervisor Role development using the Situation Calculus as a basis to model the domain evolution by actions (strategies performed by anesthesiologists). In this model the use of fuzzy relationships to represent preconditions, actions and post-conditions are proposed. The advantage of this representation is based on reaching a better interpretation and more flexibility as regards to the characteristics of each patient and the interaction with the anesthesiologist. In this way, this representation adds clearness to the domain analysis and a familiar language to the application field.

2003-01-01T00:00:00Z In this paper aspects related with the methodological design of a multi-agent system to support both supervision of anesthesiological processes and anesthesiologists decisions are approached. This work describes how it was arrived to a multi-agent architectural design complementing the Gaia methodology with an approach for roles identification. Then, it shows the ActionPlanSupervisor Role development using the Situation Calculus as a basis to model the domain evolution by actions (strategies performed by anesthesiologists). In this model the use of fuzzy relationships to represent preconditions, actions and post-conditions are proposed. The advantage of this representation is based on reaching a better interpretation and more flexibility as regards to the characteristics of each patient and the interaction with the anesthesiologist. In this way, this representation adds clearness to the domain analysis and a familiar language to the application field. Prati, Ronaldo Cristiano Monard, Maria Carolina de Carvalho, André C. P. L. F. http://sedici.unlp.edu.ar:80/handle/10915/184930 2025-09-24T20:10:19Z 2003-01-01T00:00:00Z

Objeto de conferencia Simposio Argentino de Inteligencia Artificial (ASAI 2003)- JAIIO 32 (CABA, 1 al 5 de septiembre de 2003) The search for patterns in data sets is a fundamental task in Data Mining, where Machine Learning algorithms are generally used. However, Machine Learning algorithms have biases that strengthen the classification task, not taking into consideration exceptions. Exceptions contradict common sense rules. They are generally unknown, unexpected and contradictory to the user believes. For this reason, exceptions may be interesting. In this work we propose a method to find exceptions out from common sense rules. Besides, we apply the proposed method in a real world data set, to discover rules and exceptions in the HIV virus protein cleavage process.

2003-01-01T00:00:00Z The search for patterns in data sets is a fundamental task in Data Mining, where Machine Learning algorithms are generally used. However, Machine Learning algorithms have biases that strengthen the classification task, not taking into consideration exceptions. Exceptions contradict common sense rules. They are generally unknown, unexpected and contradictory to the user believes. For this reason, exceptions may be interesting. In this work we propose a method to find exceptions out from common sense rules. Besides, we apply the proposed method in a real world data set, to discover rules and exceptions in the HIV virus protein cleavage process.