Catálogo de publicaciones - libros

The aim of image segmentation is the partition of the image in homogeneous regions. In this paper we propose an approximation based on Markov Random Fields (MRF) able to perform correct segmentation in real time using colour information. In a first approximation a simulated annealing approach is used to obtain the optimal segmentation. This segmentation will be improved using an adaptive threshold algorithm, to achieve real time. The experiment results using the proposed segmentation prove its correctness, both for the obtained labelling and for the response time.

In many realistic planning domains, the exact duration of actions is only known at the instant of executing the action. This is the case, for instance, of temporal domains where it is common to find external factors that cause a delay in the execution of actions. In this paper we present an approach to obtain a plan for a temporal domain with delays. Our approach consists in combining a planning process, from which a temporal plan is obtained, and a scheduling process to allocate (instantiate) such a temporal plan over a time line.

In this work a sliding mode control carried out by neural networks and applied to a wastewater treatment plant is proposed. The controller has two modules: the first one performs the plant control when its dynamics lies inside an optimal working region and is carried out by a neural network trained to reproduce the behavior of the technician who controls an actual plant, while the second one drives the system dynamics towards that region when it works outside it and is carried out by another neural network trained to perform that task. Both controllers are combined with a two layers neural network where the synaptic weights of the only neuron in the second one is adjusted by those in the previous layer in order to balance the contribution of each controller to the total control action.

After defining what is understood by textual entailment and semantic equivalence, the present state and the desirable future of the systems aimed at recognizing them is shown. A compilation of the currently implemented techniques in the main Recognizing Textual Entailment and Semantic Equivalence systems is given.

This paper presents some enhancements in the temporal reasoning of a Hierarchical Task Network (HTN) planner, named SIADEX, that, up to authors knowledge, no other HTN planner has. These new features include a sound partial order metric structure, deadlines, temporal landmarking or synchronization capabilities built on top of a Simple Temporal Network [3].

We formulate the multi-team formation (M-TF) domain-independent problem and describe a generic solution for the problem. We illustrate the M-TF derogation component in the domain of an urban fire disaster.. The M-TF problem is the precursor of teamwork that explicitly addresses the achievement of several short time period goals, where the work to achieve the complete set of goals overwhelms the working capacity of the team formation space (all teams formed from the finite set of available agents). Decisions regarding team formation are made considering that team reformation is the means to counteract possible deviations from a desirable teamwork behavioral performance. The RoboCupRescue large-scale disaster environment is used to illustrate the design of the derogation domain-specific M-TF component.

Junk e-mail detection and filtering can be considered a cost-sensitive classification problem. Nevertheless, preprocessing methods and noise reduction strategies used to enhance the computational efficiency in text classification cannot be so efficient in e-mail filtering. This fact is demonstrated here where a comparative study of the use of stopword removal, stemming and different tokenising schemes is presented. The final goal is to preprocess the training e-mail corpora of several content-based techniques for spam filtering (machine approaches and case-based systems). Soundness conclusions are extracted from the experiments carried out where different scenarios are taken into consideration.

In this paper we discuss how to apply the Mental State Framework to model an agent that uses different motivations to make decisions. We employ several mental states attributes to characterize the mechanism of decision and we add a set of motivation-derived desires to the top of a BDI-like architecture. We show that an agent can change her behavior if motives behind the decisions change. Finally we explore reasons behind reasons, arguing that we can expect to identify different agent types. Endowed with such operational extensions, the Mental State Framework can provide a tool with which to program flexible agents in a principled way.

In this paper we describe a new procedure to obtain the control code for a mobile robot, based on system identification: Initially, the robot is controlled by a human operator, who manually guides it through a desired sensor-motor task. The robot’s motion is then “identified” using the NARMAX system identification technique. The resulting transparent model can subsequently be used to control the movement of the robot.

Using a transparent mathematical model for robot control furthermore has the advantage that the robot’s motion can be analysed and characterised quantitatively, resulting in a better understanding of robot-environment interaction.

We demonstrate this approach to robot programming in experiments with a Magellan Pro mobile robot, using the task of door traversal as a testbed.