Recent developments in artificial intelligence (AI) have led to the use of knowledge-based techniques for solving scheduling problems. The authors survey several existing intelligent planning and scheduling systems with the aim of providing a guide to the main AI techniques used. In view of the prevailing difference is usage of the terms planning and scheduling between AI and operations research (OR), a taxonomy of planning and scheduling problems is presented. The modeling of real world problems from closed deterministic worlds to complex real worlds is illustrated with a project scheduling example. Some of the more successful planning and scheduling systems are surveyed, and their features are highlighted. The AI approaches are consolidated into knowledge representation and problem solving in the project management context.>

A knowledge-based approach and a reasoning system for multisensor data fusion is presented. The scenario taken for the example is an air-land battlefield situation. A data fusion system obtains data from a variety of sensors. A Dempster-Shafer approach to representing and combining data is found appropriate for combining uncertain information from these disparate sensor sources at different levels of abstraction. Evidential reasoning allows confidences to be assigned to sets of propositions rather than to just N mutually exclusive propositions. The software has been developed in the Lisp language and tested. The results illustrate the advantages of using multiple sensors in terms of increase in detection probability, increased spatial and temporal coverage, and increased reliability.>

An adaptive learning scheme, based on a fuzzy approximation to the gradient descent method for training a pattern classifier using unlabeled samples, is described. The objective function defined for the fuzzy ISODATA clustering procedure is used as the loss function for computing the gradient. Learning is based on simultaneous fuzzy decisionmaking and estimation. It uses conditional fuzzy measures on unlabeled samples. An exponential membership function is assumed for each class, and the parameters constituting these membership functions are estimated, using the gradient, in a recursive fashion. The induced possibility of occurrence of each class is useful for estimation and is computed using 1) the membership of the new sample in that class and 2) the previously computed average possibility of occurrence of the same class. An inductive entropy measure is defined in terms of induced possibility distribution to measure the extent of learning. The method is illustrated with relevant examples.

Simultaneous consideration of both performance and reliability issues is important in the choice of computer architectures for real-time aerospace applications. One of the requirements for such a fault-tolerant computer system is the characteristic of graceful degradation. A shared and replicated resources computing system represents such an architecture. In this paper, a combinatorial model is used for the evaluation of the instruction execution rate of a degradable, replicated resources computing system such as a modular multiprocessor system. Next, a method is presented to evaluate the computation reliability of such a system utilizing a reliability graph model and the instruction execution rate. Finally, this computation reliability measure, which simultaneously describes both performance and reliability, is applied as a constraint in an architecture optimization model for such computing systems.

The minimum cost classifier when general cost functions are associated with the tasks of feature measurement and classification is formulated as a decision graph which does not reject class labels at intermediate stages. Noting its complexities, a heuristic procedure to simplify this scheme to a binary decision tree is presented. The optimization of the binary tree in this context is carried out using dynamic programming. This technique is applied to the voiced-unvoiced-silence classification in speech processing.

Speaker recognition schemes which work satisfactorily for small populations often fail when the number of classes is very large. One way of solving such problems is to go in for multistage classification schemes. The basic technique is to successively reduce the number of classes in several stages using one feature at each stage and when the number of classes is less than a predetermined value then the final decision is made. The whole scheme is designed so that the probability of error is fixed at an acceptable level. The computational cost of such a multistage scheme depends on the features used at each stage and the cost of measurement of each feature. The features to be used at each stage are determined so as to reduce the average computational cost for making a decision. This procedure is formulated as a stochastic optimal control problem and is illustrated by designing speaker recognition system for 60 speakers. The overall accuracy of the system is 97.2 %

A study is presented which is aimed at developing techniques suitable for effective planning and efficient operation of fleets of aircraft typical of the air force of a developing country. An important aspect of fleet management, the problem of resource allocation for achieving prescribed operational effectiveness of the fleet, is considered. For analysis purposes, it is assumed that the planes operate in a single flying-base repair-depot environment. The perennial problem of resource allocation for fleet and facility buildup that faces planners is modeled and solved as an optimal control problem. These models contain two "policy" variables representing investments in aircraft and repair facilities. The feasibility of decentralized control is explored by assuming the two policy variables are under the control of two independent decisionmakers guided by different and not often well coordinated objectives.

Design of speaker identification systems for a small number of speakers (around 10) with a high degree of accuracy has evolved over the past few years. A sequential identification technique gives better results when the number of speakers is large. This scheme is implemented as a decision tree classifier in which the final decision is made only after a predetermined number of stages. The error rate could be controlled consistent with the features selected. This paper describes a 2-stage decision tree classifier implemented on a HP Fourier analyser system for identification of 30 speakers. The scheme proceeds as follows. At the first stage of the decision tree the population under consideration is reduced by a factor of 3 with high degree of accuracy, and at the second stage, the scheme available for ten speaker identification is used, The computational savings and performance achieved are compared with that obtained in a single stage classifier.

A pattern recognition approach for deciding whether a given segment of speech should be classified as voiced speech, unvoiced speech or silence based on a set of five measurements of the signal is given by Atal and Rabiner [1]. In this paper, we demonstrate that it is possible to achieve this classification with much less computational effort. These computational savings are mainly achieved by adopting a scheme based on the concept of variable decision space, using only three features and by avoiding the time consuming linear prediction analysis.

Quest for new speaker dependent features is a constant problem in the design of automatic speaker recognition systems. In speech, information about the speaker usually arises along with the semantic information which makes its independent use difficult. In this paper, a method based on linear prediction (LP) analysis is described which yields features that are more speaker dependent than the usual linear predictor coefficients (LPC). In this method the LPC contours are obtained through cascade realization of digital inverse filtering (DIF) for speech signals. A low order (2-4) DIF removes the gross spectral characteristics such as the large dynamic range and some significant peaks which tend to mask the weaker formants. Visual comparison of the contours and a preliminary statistical analysis indicate that the LPC contours obtained by processing the output signal of the first stage contain better features for speaker dependency than the direct LPC contours.

Optimal preventive maintenance policies, for a machine subject to deterioration with age and intermittent breakdowns and repairs, are derived using optimal control theory. The optimal policies are shown to be of bang-bang nature. The extension to the case when there are a large number of identical machines and several repairmen in the system is considered next. This model takes into account the waiting line formed at the repair facility and establishes a link between this problem and the classical ``repairmen problem.''

Optimal maintenance policies for a machine with degradation in performance with age and subject to failure are considered. The optimal policies are shown to be generally of a bang-coast nature, except in the case where the probability of machine failure is a function of maintenance. It is also shown, in the case where machine failure is not taken into account, that a high depreciation rate tends to reverse this policy to coast-bang. The results obtained here extend and unify the recent results for this problem.