Many Multi-Agent Systems (MASs) have been built that represent different social structures, such as alliances, teams, coalitions, conventions and markets. Our primary objective in this work was to establish a fairly general-purpose market structure that can be reused in other market-based MASs. We have implemented different models of a deregulated electricity market. The models vary in the types of interaction between buyers and sellers, the mechanisms used to transact deals in the market, and in the role of the central agent. Extensive experiments were then conducted on the models to determine the type of market structure best suited for trading. The criteria of evaluation were based on resource consumption, number of deals completed, demand satisfied, supply used, excess-over-need bought, and prices paid. The conclusions should be applicable to other deregulated consumer markets of "perishable" commodities with strong time-dependence.

The danger of getting into a combinatorial explosion has remained probably the most serious impediment to using knowledge-based systems for real-life problems. Justifiable heuristics must cut down the size of potentially huge search spaces to a realistic and manageable size - without jeopardizing the success of finding satisfactory solutions. This paper describes some techniques that have proved to be effective in the operation of two different systems: the Quasi-Optimizer (QO) and the Generalized Production Rule System (GPRS). The QO is a domain-independent automatic knowledge acquisition tool that can obtain, verify, fuse and optimize human expertise. It generates computer models, descriptive theories, of human decision making strategies, and can also select and combine the best components of several such models into a single strategy which may be considered a normative theory in the statistical sense. The techniques developed to reduce the danger of combinatorial explosion with the QO include selecting the most independent (near-orthogonal) decision variables, chunking conceptually coherent decision variables, eliminating statistical outlier values, and using dynamically evolving experimental designs to result in a near-uniform reliability distribution of the strategy responses. The GPRS can estimate/predict values of hidden variables and can thus serve as a module of an expert system in need of numerical or functional estimates of hidden variable values. (Hidden variables can be observed and measured only intermittently and at irregular points of time and space – in contrast with open variables whose values can be identified at any time and location.) The estimation is based on generalized production rules expressing stochastic and potentially causal relations between known values of hidden variables and certain mathematical properties of the open variable distribution. A multi-dimensional learning process adds to, consolidates and optimizes the generalized rule base. It gradually merges “similar enough” production rules, and eliminates spurious and statistically not justifiable ones. Such processes reduce the danger of combinatorial explosions in knowledge acquisition. Finally, we note that human decision makers appear to resort to similar methods, although in a less systematic manner, when they are confronted with very large decision spaces.

Our work is based on distributed and knowledge-based methodology for real-time control of street and highway ramp signals. After a running prototype system that proved the feasibility of the approach, we have now completed the first part of a complex, two-phase learning program of the traffic control regime. It produces optimum local (intersection) control of signals in steady-state traffic scenarios. With appropriate communication between adjacent intersections, we should obtain an optimum, global, reactive and predictive control regime of the semi-autonomous, decentralized system.>

We contrast in this article the system demands imposed by brute force, complete replanning, and a new, goal-directed approach called dynamic plan revision (DPR). One of the two processes must be invoked when some aspects of the environment change and currently existing plans are likely to fail. Complete replanning is conceptually simpler but requires significantly more computation because it reproduces also those plan steps that have not been affected by the changes. DPR attempts to minimize the scope of replanning and makes to the current plans as small amount of modification as possible. the more rapidly the environment changes, the less practical complete replanning becomes. the issue is even more significant with time-critical applications DPR uses goal hierarchies produced at the same time when plans are generated. the planning systems has an “active attitude” toward the feedback from the environment and takes advantage of all information becoming available during plan execution. We have also developed and implemented a distributed version of this approach in the domain of Distributed Manufacturing Operations [N. V. Findler and J. Gao, Data and Know. Eng., 2, 285–301 (1987); Q. Ge and N. V. Findler, Proc. Austr. Joint Conf. on Art. Int., 439–455 (1988); N. Y. Findler and G. Ge, Int. J. Intell. Syst., 4, 459–470 (1989)]. the relevant algorithms are given in the Appendices A and B. © 1994 John Wiley & Sons, Inc.

Globally coherent behavior is essential for a distributed problem solving network. It is a characteristic of the whole problem solving process. We discuss in this article different forms of cooperation at different phases of the problem solving process, which have to be considered to increase global coherence. the connection problem and the timing problem are key issues for distributed problem solving. the “perceive-plan-act” loop is introduced for each problem solving node. It means that a node has to perceive the network state and plan its near-future activities before it takes an action. Different approaches to network perception are discussed. the experimental results show significant improvement of system performance.

The task of credit assignment is interpreted as the ability of a system to: identify and distinguish strategy components; associated with such components, in different regions of the domain of confrontation, good and poor outcomes of a sequence of actions prescribed by the strategy; improve the boundaries of these regions in minimizing the errors of misclassification of strategy actions; estimate the overall quality of a strategy; and provide information for a metastrategy that shifts the domain of confrontation to those regions in which the strategy studied is most proficient and thereby raises the effective quality of the strategy. The program QO-4, the fourth module of the large system QA (the quasi-optimizer) which can accomplish the above tasks within certain limitations, is described.

We present a class of similarity measures for quantitatively comparing two strings, that is, two linearly ordered sets of elements. The strings can be of different lengths, the elements come from a single alphabet, and an element may appear any number of times. The limiting values of each measure are 0, when two completely different strings are compared, and 1, when the two strings are identical. Applications of similarity measures are numerous in nonnumerical computations, such as in heuristic search processes in associative networks, in pattern recognition and classification, in game playing programs, and in music and text analysis. We offer a number of feasible measures from among which some are discarded on plausibility grounds. One can select the measure most adequate for one's needs on the basis of a few characteristic examples of strings compared and by considering the specific requirements of the application at hand.

The popularity of intelligence-simulating programs is increasing. Several parlour-games have been programmed for computing machines, with the idea of investigating machines learning techniques. This paper describes the strategies employed in a program for playing the game of dama on the SILLIAC. A learning process is suggested by means of which an optimal grand strategy can be achieved.