Pattern‐directed inference systems (P.D.I.S.) are among the most largely used tools in A.I. to‐day in order to represent and exploit knowledge. Generally, P.D.I.S.'s use production rules triggered by matching between rule patterns and elements of the data base. However, the lack of flexibility in the matching remains a drawback in this kind of system. In the framework of the communication in natural language with robots, approximate descriptions of real world situations and approximately specified rules are needed; furthermore, similarity in the matching process does not always need to be perfect. Thus, the pervading fuzziness of natural language can be taken into account. The following levels, belonging to the real interval [0,1], are evaluated: The possibility of similarity between referents designated in the data and in the pattern respectively; the necessity that a referent designated in the data is similar to a referent designated in the pattern. Designations are fuzzy when the pattern or the data are fuzzy, which is usual with words of a natural language.

Planners of the Graphplan family (Graphplan, IPP, STAN...) are currently considered to be the most efficient ones on numerous planning domains. Their partially ordered plans can be represented as sequences of sets of actions. The sets of actions generated by Graphplan satisfy a strong independence property which allows one to manipulate each set as a whole. We present a detailed formal analysis that demonstrates that the independence criterion can be partially relaxed in order to produce valid plans in the sense of Graphplan. Indeed, two actions at a same level of the planning-graph do not need to be marked as mutually exclusive if there exists a possible ordering between them that respects a criterion of "authorization", less constrained than the criterion of independence. The ordering between the actions can be set up after the plan has been generated, and the extraction of the solution plan needs an extra checking process that guarantees that an ordering can be found for actions considered simultaneously, at each level of the planning-graph. This study lead us to implement a modified Graphplan, LCGP (for "Least Committed GraphPlan"), which is still sound and complete and generally produces plans that have fewer levels than those of Graphplan (the same number in the worst cases). We present an experimental study which demonstrates that, in classical planning domains, LCGP solves more problems than planners from the family of Graphplan (Graphplan, IPP, STAN...). In most cases, LCGP also outperforms the other planners.

In CTW90], the ATMS resolution mechanisms have been formalized, using the concept of DK-tree. We propose in this paper a new and simple approach based on the xed point theory and an original demonstration of the completeness of these mechanisms. We rst introduce the basic de nitions of terms, sets and functions which are used in the paper and build a lattice. Then, the basic formal tools to describe the ATMS mechanisms are introduced: The notion of DK-Clash (in fact, the inference rule), and a monotonic function, which represents the deduction process and whose least xed point corresponds to the output of an ATMS. We prove the soundness and the completeness (for speci c clauses) of these mechanisms. This demonstration brings to light an original and interesting property, whose applications are discussed at the end of the paper.

The notion of Roi de rats corresponds to a functional association between values of variables. The causality relations between the more diversified properties of objects (concrete on abstract objects) can be easily represented and automatically dealt with. It was first introduced in the framework of the production rules system, ARGOS‐II; this notion is also implemented in TLISP. After preliminary remarks about variables, evaluation and assignment, we define direct variables and partitioned/tied variables and we give some examples of Roi de rats. Then the implementation of the Roi de rats by the TLISP interpreter is briefly described. We present three types of applications: Updating process of a data‐basis, checking of a variable and definition of a generator variable. In concluding remarks we observe that the RDR mechanism can be naturally expressed in PLASMA.

No abstract