Group of Sensorial Systems and NC Machining:
Activities
Objectives:
Vision
In the context of Intelligent Buildings, first experiments were made with the tracking of moving objects. The global objective is the development of an autonomous system which is able to follow a non-rigid object in some restricted field of view. First, studies were pursued about existing techniques for achieving the envisaged goals. Optical flow methods were considered for a possibility to solve the problem of determining the path of a moving object. The computing requirements for the calculus of the optical flow however are too demanding to be used in a real-time environment. Simpler token-tracking approaches were implemented which are based on object-background segmentation. The results are still very elementary. Compromises have to be found in order to find a solution for a practicable system under a series of constraints. Current research activities are investigating this topic. A framework for a closed-loop system has been achieved, namely a pan-and-tilt camera, an image processing system and a control module for the movement of the camera.
Learning
The application field of machine tool supervision was used as a test bed for machine learning techniques. The concept of a Specific Monitor was elaborated. This is an entity that is responsible for the fault detection of a limited piece of machine operation. Traditional pattern recognition techniques are used to learn how to supervise a production process. A high degree of autonomy can be achieved with Specific Monitors. After a training phase, specific monitors are responsible for the classification of process situations. A basic philosophy for the construction of the supervision system is a model-free approach. No internal analytical model of the process has to be defined to perform the supervision task. Besides a global framework for the application of learning was defined (see Activity: Prognosis in NC machining).
Pattern Recognition techniques
Artificial Neural Networks seem to be the most promising perspective to solve complex recognition tasks or to learn many different kind of functions. This area was studied with more intensity. Supervised learning was applied with Learning Vector Quantization (LVQ) in the area of machine tool supervision. This non-parametric technique is especially well suited for the model-free classification of process situations.
Toolboxes
The "TOOLDIAG" pattern recognition software was extended considerably. Different classifier models can be defined. For the part of feature selection, search strategies and selection criteria can be combined. More cross validation techniques were provided.
People involved:
Thomas Rauber, Luis Gomes, Dirk Tilsner, Pedro Próspero Luis
List of projects:
Objectives:
Working activities have been carried in the study and development of a methodology to specify, simulate and implement the control of distributed sensorial systems architectures and take care of the associated controllers programming. A coloured and synchronized Petri net based model has been proposed as a common tool for multi-level specification and simulation of this kind of systems. In this model, the natural capabilities of Petri nets for discrete event systems modeling are associated with fuzzy reasoning. Aplications to programmable controllers control specification was developed.
Projects related with applications of distributed sensorial systems to specific applications has been prepared.
People involved:
A. Steiger_Garção, Luis Gomes, Pedro Próspero Luis, João Paulo Barros, Carlos Soares
List of projects:
CRI infrastructures, ISEnto
Objectives:
The objective of this project is develop and implement an integrated Information System inside ATECNIC company. This work is supported by the results achieved during CIMTOFI project, namely the development of the laboratorial prototype SIP (STEP-based Integration Platform), in the scope of European BRITE/EURAM program. Considering the set of industrial applications available in the company, we expect to integrat using SIP a CAD system, CNC machines, a shop-floor control, a MRPII, a CAPP system and several accounting and financial software tools.
People involved:
A. Steiger_Garção, Ricardo Gonçalves, Pedro Sousa, Miguel Vital, Helder Silva, João Paulo Pimentão
List of projects:
CRI-ATEC
Objectives:
The objective of this project is the development of a generic multipurpose control architecture for an integrated outdoor building site of mobile heavy-duty platforms. The work will concentrate on the automation of specific heavy-duty machines in order to obtain a manless or manpower reduced building site with a computer integrated enviroment including design, planning, production and control.
The whole building site will be controlled and surveyed by a central site controller, UNINOVA's aim within this project. The site controller will supervise and coordinate the work of the working cells. Each working cell has its own intelligence and could be driven by remote control from the site controller or manually by an operator with the help of a local user interface to secure functioning in unforseeable or unstable conditions.
The communication between the site controller and the working cells will be done via a bidirectional transmission module, which has to cope with problems associated to the environmental, atmospherical and electrical conditions of out-door building sites.
Within this project, development work for automation of two specific processes will be done, namely of a road paving machine and of an excavator.
People involved:
A. Steiger_Garção, Ricardo Gonçalves, João Paulo Pimentão, Teresa C. Gonçalves, Henrique Melo, Jorge SobrinhoList of projects:
RoadRobot
Objectives:
To develop Autonomous Vehicles for different applications in unstructured environments. The main approach followed has been centered on considering a vehicle as an artificial being, whose decision control uses a behavior based architecture. Each behavior is taken as an independent module which interacts with other modules in order to produce a resulting global behavior that the vehicle displays and which should allow it to perform some useful tasks. An architecture with distributed selection has been developed and applied in a small vehicle, allowing it to stash away film boxes. A simulator of an Autonomous Underwater Vehicle, using such a model, had already been developed and is also available. Applications in vision based autonomous vehicles are now under development, as well as an evolution of the model to encompass autonomous adaptation.
People involved:
A. Steiger_Garção, Luis Correia, Sérgio SantosList of projects:
Objectives:
There are two main requirements that a machine tool should satisfy: i) high functional reliability and ii) the ability to produce continually pieces presenting an acceptable level of quality. The accuracy of the machined part is a function of several parameters like cutting conditions, type of operations and tools used, age and condition of the machine, quality of the work piece material, etc. To ensure that the produced part is conformant with its quality specifications, the quality of the producing system must be determined and assured by measuring either the produced workpieces or/and the machine itself.
We can use the data obtained from workpiece measurement and machine calibration to compensate the static and dynamic errors of the machining system. However the dynamic errors due to machining forces, bending of the system, vibrations, tool wear, are very difficult to measure in a finished workpiece and might be neglected. Also determining the quality performance of a machine by measuring only the machined workpieces can be an expensive process. If the workpiece presents defects then it must be rejected and this might represent high unprofitable situations. In order to minimize these loses the feedback between the measuring system and the correction must be minimized as much as possible. This activity is therefore related to diagnosis and prognosis in NC machines. Machine learning techniques are being investigated in this context.
People involved:
Manuel Barata, Thomas Rauber, Ricardo Gonçalves, Helder Silva
List of projects:
B-LEARN II
Objectives:
The major goal for automated control of infrastructures such as industrial and institutional buildings has been a more adequate use of the available resources. It has become more and more important to consider the different systems present in the building as coherent parts of a global system for the integrated control, monitoring and management of the building.
The strategies for a rational management of the available resources have to consider the capability to integrate different systems and funcionalities, exploring the synergetic results obtained, more than the simple addition.
The work has been focused on the monitoring and access control sub-systems, based on different sensors and actuators, like cameras, presence detectors, video multiplexers and card readers.
Cooperative projects with national industries have been prepared.
People involved:
A. Steiger_Garção, Luis Gomes, Anikó Costa, Pedro Próspero Luis, Carlos Soares, João Paulo Barros, Luis Câmara Pires
List of projects:
CRI infrastructures, ISEnto, PEDIP training
please email comments to Mário Rui Sousa Updated: Out 18, 1995