Abstract
A major public transport company, working in a urban environment, in Portugal, has adapted a telemetric system for automatic vehicle location, AVL, in order to optimise vehicle use, timetabling and scheduling. The system is able to provide real time location and load of the buses, to exchange voice messages with the driver and to propose preventive maintenance actions on the vehicles. The control centre exchanges messages with all the buses. These messages have two different components: the real-time control component, that is thrown away after use, and the off-line component that has its data stored sequentially on disk, far later analysis. Our main concern is to extract relevant information from this bulk of data to provide the bus management expert with a more accurate knowledge of the bus fleet performance, namely the bus operation efficiency, to improve vehicle use and scheduling. Several factors may corrupt the data that is stored on disk and make it impossible to automatically extract useful information. A pre-processing stage is needed to classify data as consistent or inconsistent. A strategy to implement this preprocess stage has been proposed in a previous paper [Cunha 1997], The idea is based on a virtual bus model. The virtual bus travels on a bus route and recreates the real bus service. It compares messages which have been received with those that could be expected in the model. The model is extended in this paper, in order to analyse inconsistent data and take automatic correction actions. In less common situations, control is passed to a human operator for him/her to make an appropriate correction.

Abstract
One of the advantages of logic programming in the fact that it offers many sources of implicit parallelism, such as and-parallelism and or-parallelism Recently, research has been concentrated on integrating the different forms of parallelism into a single combined system. In this work we concentrate on the problem of integrating or-parallelism and independent and-parallelism for parallel Prolog systems. We contend that previous data structures require pure recomputation and therefore do not allow for orthogonality between and parallelism and or-parallelism. In contrast, we submit that a simpler solution, the sparse binding array, does guarantee this goal, and explain in detail how independent and-parallelism and or-parallelism can thus be efficiently combined.

Abstract
The detection and location of faulty passive components is crucial in testing printed circuit boards. This issue is one of the objectives to be satisfied by the development of a mixed-signal test bus standard - IEEE P1149.4 - now on the final phase of evaluation. In this paper a method is presented that enables electronic in-circuit testing for presence and value of passive components, based on the use of analogue boundary modules like those developed under P1149.4, but making use of the positive power supply rail as an analogue test line. This method allows saving one analogue test line while profiting from the eventual presence of current sensors inserted for power supply current monitoring.

Abstract
The development of a mixed-signal test bus infrastructure - IEEE P1149.4 - is now in the final stages of the standardization process. Evaluating the test capabilities enabled by this infrastructure is an important step needed to support it as a well established standard. This paper presents experiments carried out with a test chip provided by the P1149.4 working group, which explore the architecture of the proposed analog boundary module to implement alternative testing methods. These include a method for parametric testing of passive components based on the monitoring of the power supply current, and a mixed current/voltage technique allowing the implementation of correlation for testing analog and mixed-signal macros.

Abstract
The technique of genetic algorithms is proposed as a means of auto-tuning PID controllers. The technique revolves firstly using on-line data and the genetic algorithm to identify a model of the process. Then the identified model, the genetic algorithm and simulation methods, are used to off-line tune the PID controller, so as to minimize a time-domain based cost function. Finally, the genetically tuned controller is implemented on-line on the real process. The results of the genetic auto-tuner are illustrated by auto-tuning a PID controller on a laboratory heat exchanger, and comparing the genetic auto-tuning technique with the Astrom-relay auto-tuning technique.

Abstract
This article presents an alternative approach to the problem of regression. The methodology we describe allows the use of classification algorithms in regression tasks. From a practical point of view this enables the use of a wide range of existing machine learning (ML) systems in regression problems. In effect, most of the widely available systems deal with classification. Our method works as a pre-processing step in which the continuous goal variable values are discretised into a set of intervals. We use misclassification costs as a means to reflect the implicit ordering among these intervals. We describe a set of alternative discretisation methods and, based on our experimental results, justify the need for a search-based approach to choose the best method. The discretisation process is isolated from the classification algorithm, thus being applicable to virtually any existing system. The implemented system (RECLA) can thus be seen as a generic pre-processing tool. We have tested RECLA with three different classification systems and evaluated it in several regression data sets. Our experimental results confirm the validity of our search-based approach to class discretisation, and reveal the accuracy benefits of adding misclassification costs. © 1997 Elsevier Science B.Y.