This paper outlines the adaption of artificial intelligence techniques in spatial information systems. This adaption improves the analytical strength of spatial information systems. This paper includes a discussion of previously coupled techniques such as expert systems, fuzzy logic, hybrid connection systems and neural networks. It proposes a spatial reasoning prototype which is a based upon the integration of case-based reasoning and spatial information systems techniques.
The paper presents a general engineering model of a hybrid system for speech recognition and information retrieval. The speech recognition part consists of a low level neural network module for phoneme recognition and a higher level fuzzy reasoning module for word recognition and language modelling. There are several features such systems are characterised by, namely: adaptability, i.e. the system is adaptable to new speakers and accents; dealing with ambiguity during the recognition process; extendability, i.e. new modules can be introduced to the system at a later stage and the dictionary is extendable. A case study of building an experimental English to Maori talking dictionary for a small dictionary of words and speaker independent mode is reported.
We propose a neural network model for the processing of fuzzy data. The network parameters (weights) are standard real numbers and the spreads at the output level result exclusively from uncertainty in the input data. Our network model performs 'intelligent' inference calculations on the basis of fuzzy data and minimizes uncertainty in the final output. The number of free parameters (weights) in our network model coincides with the number of connections emanating from the various nodes. In this paper, we refer only to the single node case. As usual, the learning mechanism corresponds to a nonlinear regression over the network parameters. The local transfer functions associated with the nodes are more sophisticated than the standard ones. The difference is mainly in the preliminary integration module, in which the various local inputs contribute to the single total input to the node. Instead of the standard linear combination with the parameters /spl omega//sub i/, our model weights each local input in a way which is inversely proportional to its spread (uncertainty). As a result, the more precise data are dominant in the local network computation.
For computer aided design (CAD)/design automation (DA) for printed circuit board (PCB), automatic placement of parts has been a difficult problem. For realizing good placement, not only the structures but also the functions of the circuit should be grasped and some hierarchical methods for the placement should be done. This paper presents a new hierarchical method of placement of parts on the PCB with fuzzification of parts connections. The devices in a circuit are divided into several blocks based on the flows of main signals of the circuit. However it is difficult to judge which device belongs to which block based on only the information included in the CAD data. We propose a method to fuzzify the degrees of belonging of the devices to the blocks. The boundaries of the areas for the blocks are also fuzzified. Each device is placed based on fuzzy areas of blocks, The feasibility of the method is verified using an actual PCB.
In process automation the control of plants with uncertainty or unknown parameters normally requires either a costly identification phase or an intensive fine-tuning of the controller. An easy adaptive fuzzy control is proposed which allows automatic controller tuning during operation with respect to desired closed loop behaviour. This adaptive control was added to the fuzzy kernel of an existing process automation system. Finally, experimental results of a three fingered robot gripper using this adaptive fuzzy control are presented.
Case-based reasoning (CBR) systems rely on the conceptual ordering of entities called cases. If atomic case features are allowed to assume numeric as well as symbolic values, then a systematic comparison regime is needed to aggregate similarity scores. A common approach to deal with real-numbered features is normalisation. However, there are two conspicuous problems with this procedure: the similarity between two features is dependent on the corresponding values of all other cases to be ranked; and real-numbered features are often interpreted by human experts according to conceptual constraints associated with features. In such situations, a conceptual distance between two features should be determined rather than the length of a `gap' on a linear scale. Within the framework of a comprehensive case-knowledge architecture, the notion of a concept frame that can be associated with a case feature is proposed. Through this component it is possible to represent polymorphic atomic case features, and to systematically establish the concept distance between two real-numbered feature instances
This paper examines the applicability of intelligent information processing techniques for the analysis of vascular laboratory data associated with varicose vein disorders. In the first section a brief description of varicose disease is provided. Next, the notion of applying different types of neural network to learning the dynamics of the disease is examined in two experiments. Subsequent to these, a new approach to visualising the output of a Kohonen network is presented. A brief discussion then follows on an architecture for combining these networks into an intelligent hybrid decision making system. Finally, directions for future research are discussed
One of the shortcomings of artificial neural networks (ANNs) is the difficulty in predicting the best control parameters for a certain application. The number of combinations of parameters is very large. This makes it very inefficient and expensive to search manually by trial and error. Genetic Algorithms (GAs) are an excellent and effective search technique suitable for this task. This paper describes an investigation into the use of GAs to automate the choice of parameters in both a Standard Back Propagation (SBP) and a Fuzzy Back Propagation (FBP) network for different applications.
The work presented in this paper shows that fuzzy logic can be useful for the scheduling problems where the production data such as processing times is fuzzy. The chosen area of application of fuzzy logic is the hybrid control architecture of a Flexible Manufacturing System (FMS). Fuzzy logic is used for prioritization and the ordering of the jobs in the Factory Controller queue. The preliminary results indicate that implementing a fuzzy scheduling system to shop floor scheduling will ease the rather complicated scheduling problems.
Fuzzy logic has emerged as a superior control method for processes that are mathematically difficult to model. The paper investigates the potential of fuzzy logic for the control of a MIMO process with interactive process states. A fuzzy logic controller was designed and implemented for an industrial storage chamber where both temperature and humidity conditions are maintained during operation. Results from experiments and a comparison with PID control demonstrate the reliability and robustness of the controller under normal operating conditions and in the presence of disturbances
The development of knowledge based systems and decision support systems that provide advice on policy, law and legislation is still in its infancy. Such systems require a degree of expertise to be embedded within them. This paper describes the process of development of a compliance system that interprets a subset of the rules and regulations of the European Community's Common Agricultural Policy (CAP), which is one of the most contentious pieces of wide-ranging and complex legislation in the world. The CAP legislation in Northern Ireland is complex and farmers find it difficult to interpret. For this reason a knowledge based solution was developed and implemented. The finished system illustrates the usability of high-level object-oriented development tools in the development of flexible knowledge based software.
The paper addresses the problems involved in the control of advanced manufacturing systems. An intelligent auction based distributed decision making system is developed for the control of activities at the factory and cell level within the advanced manufacturing system. It utilises the hybrid control system, taking advantages offered by both hierarchical and heterarchical control structures. Such a system has proved to have greater fault tolerance of unforeseen events. Assignment of operations to various machines are not performed according to a prescribed schedule, rather decisions are made in real time as to what machine will process what task leading to the term real time scheduling
Work in the field of AI over the past twenty years has shown that many problems can be represented as constraint satisfaction problems and efficiently solved by constraint satisfaction algorithms. However, constraint satisfaction in its pure form isn't always suitable far real world problems, as they often tend to be inconsistent, which means the corresponding constraint satisfaction problems don't have solutions. A way to handle inconsistent constraint satisfaction problems is to make them fuzzy. The idea is to associate fuzzy values with the elements of the constraints, and to combine these fuzzy values in a reasonable way, i.e., a way that directly corresponds to the way in which crisp constraint problems are handled. The purpose of the paper is to briefly introduce a framework for fuzzy constraint satisfaction problems and to discuss some heuristics for solving then efficiently.
This paper describes an experiment in applying a standard supervised machine learning algorithm (C4.5) to the problem of developing subject classification rules for documents. This algorithm is found to produce surprisingly concise models of document classifications. While the models are highly accurate on the training sets, evaluation over test sets or through cross-validation shows a significant decrease in classification accuracy. Given the difficult nature of the experimental task, however, the results of this investigation are promising and merit further study. An additional algorithm, 1R, is shown to be highly effective in generating lists of candidate terms for subject descriptions
Typically, two sources of information about a system are available: some artisan knowledge and a sample of input-output data. This paper proposes a method for the amalgamation of these to synthesise a fuzzy model of the system. The artisan knowledge will likely be qualitative, of low resolution and accuracy whilst the data sample noisy and incomplete (not comprehensively covering the whole input space). A model derived from the union of these is potentially superior to one developed from either alone.
This paper describes a novel trainable controller which extends the feedback-error paradigm of Kawato (1987), called an adaptive trajectory generator. The controller can be adapted so that a controlled system shows a behavior which is specified using heuristic rules. The training algorithm is simple and computationally inexpensive. Its effectiveness is demonstrated with the control of a simulated two-link robot arm to repeatedly throw and catch a ball.
The importance of analysing moving scenes within the wide area of digital image processing is increasingly high. Although a simple detection of object velocity by biological models has been considered in previously published papers (A. Tsukamoto et al., 1993; S. Wimbauer et al., 1994; J. Hogden et al., 1993), an implementation of artificial neural networks using a priori information for motion analysis is still quite rare. The paper shows the benefits from artificial neural networks and from using a priori information about the contents of the history in the image sequence to improve accuracy and speed of estimating motion parameters in the cases of distorted or overlapped objects. Firstly, it introduces 3 dimensional self organizing maps (SOM) with 2 dimensional input layers.
This paper presents our ongoing work in building an architecture for instructable systems, along with an illustrative example. The architecture uses knowledge about instruction discourse, planning and induction, to create a task representation that is suitable for execution as well as for planning later tasks. Major components have been built; future work includes putting them together. Our aim is to provide end users with "expert assistant" tools for automating their repetitive computer tasks.
The 1R machine learning scheme (Holte, 1993) is a very simple one that proves surprisingly effective on the standard datasets commonly used for evaluation. This paper describes the method and discusses two aspects of the algorithm that bear further analysis: the way, that intervals are formed when discretizing continuously-valued attributes; and the way missing values are treated. We then show how the algorithm can be extended to avoid a problem endemic to most practical machine learning algorithms-their frequent dismissal of an attribute as irrelevant when in fact it is highly relevant when combined with other attributes.
The blackboard control architecture has proven to be qualified for tasks as complex as supervisory process control. The task structure framework permits a hierarchical decomposition of tasks into subtasks common in the domain of supervisory process control. Qualitative task descriptions based on fuzzy sets grant the integration of the task structure framework while preserving the inherent flexibility of the blackboard control architecture. Fuzzy sets are further qualified to improve the basic control cycle. Linguistic variables allow the specification of control knowledge in a more natural way according to human knowledge
