Basic Projects
 
 

This Project deals with the generation and organization of the interaction mechanisms between an artificial system and its natural environment, which is usually complex, partially known and dynamic. Then the information from the system and environment at hand is incomplete and fragmented in several representations and with different resolution degree, but in any of such situations decisions must be taken in order to accomplish the goal. Either in human or in artificial beings, perception mechanisms, planning, decision making, and learning are complementary faces in the generation of the right strategy; whenever the result is observable we call it behavior.

A hierarchical structure with different resolution levels, each one defined by means of behavioral - agents is proposed, to integrate expert knowledge and self - experience. In this organizational scheme, some of the decisions making processes at a level corresponds with the abstraction mechanisms. This abstraction is accomplished by generalization of the inference mechanisms that conform the agents of the lower adjacent level, broadening the data abstraction concept that holds in pattern recognition problems. The hierarchy of decision making mechanisms here suggested, is the guarantee to efficiently handling complexity. Same way, the Agent or consistent relational entity provide the robustness needed to reach the goal in spite of multiple alternatives available and uncertain knowledge. The architecture will be demonstrated on an autonomous mobile robot endowed with heterogeneous sensors (proximity, ultrasonic, infrared, compass, and vision subsystem) as it is becoming an standard for behavior based demostrations.
 
 

This proposal addresses the design, development, and validation of an integrated system for active vision (prototype of an electronic eye). To this end, the activities of three research groups with complementary experience in the artificial vision field are coordinated. These activities range from the design and realization of integrated circuits for optical image acquisition and analog processing (Si retinas) to the study and modeling of the optics, retina, and primary representations of the human visual system, to the design and construction of visually guided robots. Even though these complementary aspects have matured independently, integration proposals are scarce for a rapid, flexible, and low cost system that could immediately fulfill the requirements of many real-time academic or industrial applications. The reasons for this lie in the inherent complexity of this type of enterprise and in the heterogeneity of the theoretic and technological knowledge required, which demand the joint effort of multi-disciplinary research group.

The project aims to integrate the knowledge of the optical and nervous functions performed by different animal eyes, projecting them towards their possible artificial realization supported by the experience in the Robotics, Artificial Intelligence, and Microelectronics fields. Its integration in a specially designed programmable retina constitutes a breakthrough in size, power consumption, and cost, factors that usually restrict the scope of artificial vision systems. Many aspects of biological visual systems are goals to be reached through artificial vision systems, such as the adaptation to dynamics environments and goals, robustness, S/N ratio, speed in response, etc., and all of this significantly reduces resources and cost. Some of the well-known connectivity techniques used by the animal eye that enable such performances are: foveatization, detectors with different sensitivity and frequency response, lateral inhibition, differential response to the temporal integration, instantaneous control of visual axis, iris, and focus, stereovision, etc. Main points exhibited to be considered are its simplicity, modularity, and the parallelism of the elements that accomplish such functions, then its simulation being significantly simplified by the use of microelectronics circuits.

The profit of the work will be validated in clearly defined representative applications, with real vision environments and objectives. Besides, the economy of the resource requirements to obtain these goals will establish the practicality of industrialization and commercialization of the results. Two industrial-oriented applications are proposed as test-beds: 1) Visually based surveillance by mobile robots, and 2) Superficial defect detection.
 
 

This Project aims the insertion of a Distributed Control Architecture based upon multiple behavioral Agents in an autonomous mobile prototype endowed of multiple sensors to accomplish agricultural tasks. Sensing, perception and decision making processes will be framed in this Agents Architecture as a dynamic model able to hold for the active perception mechanisms needed in a rarely structured environment with a high degree of uncertainty and some unpredictability. In such a frame, the dynamic activation of a repertoire of actions or behaviors occurs jointly modulated by the objective and the interaction system-environment.

The Project focus its attention on knowledge organisation , from the perception and representation mechanisms that conform the model of the environment to the planning and actions selection, in order to reach the appropriate observable behavior needed to perform a task with energy optimization. A test-bed is proposed for the demostration of the degree of autonomy reached by the mobile robot in autonomous spraying of olive trees crops at the Instituto de Automática Industrial and fruit trees at the experimental crops of the Instituto Valenciano de Investigaciones Agrarias that will provide the equipment and the control of a new spray system actually under investigation. Final aims of the Project application will be from one side the reduction of the impact on the human being and on the ambience of a continuos flow of chemical products and from the other on the improvement of both eficciency and cost of the spraying.
 
 

European Projects

The increasing complexity of design in Aeronautic, Automotive and Energy Industries requires more and more robust optimization and control tools for solving simultaneously discrete, continuous and combinatorial difficult problems unresolved until now. Fluids Dynamics, Acoustics, Structures Mechanics, Electromagnetic, Automation control and Energy are featured multidisciplinary areas in Engineering and Applied Sciences targeted by INGENET where evolution approaches are becoming to show impressively well results.

Goals of INGENET are:

1. To bring together academic and industrial node partners to evaluate and compare in terms of accuracy and efficiency performances of recent evolution based methodologies and (parallel) software on selected test problems of industrial interest.
2. To widely and electronically stimulate and disseminate in Europe innovative evolution based methodologies and software, providing an interactive dynamic bridge between Artificial Intelligence, Computer Sciences and Engineering for solving real life problems.

The goal of this project is to provide a mechanism that allows the end-user monitor the location and status of a cargo throughout the whole logistic chain in an intermodal transport system comprising sea, rail and road transportation means. It will merge the existing technologies in the fields of cargo identification fixed and mobile data communication, database handling, directory based data relaying and electronic document interchange to build a value-added network that allows any user in the logistic chain to gain access to the required information in a reliable and user-friendly way. The IAI partner will be in charge of the development of a System that integrates both user defined relevant labels and sensor data under a fuzzy supervisor to make diagnosis and decisions on the goods state.
 
 

This project fits in the shoe manufacturing sector framework, and more specifically in Last Manufacturing (piece of high density plastic, with the shape of the final shoe used to make it by placing the different parts around it). There are specific machines which deal with each of the processes to make a Last. However, to problems can be pointed out: Deficiencies in the plastic injection machines (unnecessary oversizing of the mould, and complete lack of quality control), and the absence of electronic or computer systems which could allow a planning-coordination-control of the whole process. There are three SMEs involved in this project: two last production SMEs and one machinery SME which have great interest in solving these problems. The expected final results will have to be integrated and improved in a later trial and optimization phase, so that the programs, machines and methods become indeed an improvement for the industry. This proposal is enclosed in the objevtives of BRITE and CRAFT in the item of production technologies and materials and technologies for product innovation: Area 1 (1.1.1.2.1.3) and Area 2 (2.2.2.3).
 
 

Precompetitive and Industrial Projects

 Abstract

This project sets up the research on a multi-sensor integrated architecture for decision making in critic time and in presence of uncertainty. It promotes the processing of complex sensor information such is vision, by means of systems able to support multi-processing. The project also encourages the study of an adequate representation, to hold the global and local image processing and proposes the study of bayesian, neural nets, and fuzzy rule based systems as good candidates for classification purposes. Finally, it proposes the design, implementation and experimentation of a prototype of color artificial system for image classification. Demonstration will be based upon a multi-sensor system dealing with eggs-shell defect detection at the farm-packing industry.
 
 

 Abstract

Present work aims at the development of surveillance net for continuous environment surveillance of toxic atmospheric emissions from an industrial site located in Arganda del Rey (Madrid). The digital cartography of Arganda del Rey will allow an ease development of a GIS able to give information on: industries products and location, possible contaminants and such data will be associated with the features extracted from continuos measurements of atmospheric conditions through the acquisition information Units to be developed within this project to elaborate regional regulations to prevent air pollution emissions.
The main objective of the project is the development and installation of an open net of information Units (acquisition/processing system microcontroller based with high performance and low cost). Such Units, connected to a set of sensors, acquire and process data from a set of sensors, check System State (temperature, airflow, self-calibration), and generate alarm messages on a distributed processing net. Communication between Local Units and Central Unit will be performed via GSM or Radio-Ethernet through TCP/IP protocols) to allow for a wider bandwidth or via satellite as far as needs and economy allow for.
Arganda del Rey City Council will update the industries (about 1500 SME's) census, and will give personnel and economic support to the project. The product delivered will be easily portable and scalable to different applications domains dealing either with air, water or solids, all concerning remote sensing surveillance tasks for systems information and control. Finally, some comparative tests will be performed with silicon solid state sensors developed at the IFA-CSIC Institute.