Jeffrey Van Baalen

Automated design of specialized representations

Abstract We present a general notion of specialized representation and a methodology that automates the design of such representations. These representations are combinations of data structure and procedure that efficiently perform inferences required in solving a problem. Our implementation of the methodology begins with a problem stated in a sorted first-order logic and designs a representation for the problem by combining building blocks, called representation schemes . It has a library of representation schemes. A scheme is identified for use in designing a representation by matching a description of the inferences it performs against one or more statements in a problem. Once a scheme is designed into a representation, the general problem solver need no longer consider the problem statements matched in the scheme identification process because the scheme computes the required inferences from those statements. The machinery to perform the required inferences has been moved out of the problem solving cycle and into the representation. Experience with the implementation has shown that a theorem prover reasoning about a smaller set of statements in a specialized representation is dramatically more efficient than reasoning about the original set of statements in the original representation.

The role of reversible grammars in translating between representation languages

Abstract A capability for translating between representation languages is critical for effective knowledge base reuse. We describe a translation technology for knowledge representation languages based on the use of an interlingua for communicating knowledge. The interlingua-based translation process can be thought of as consisting of three major steps: (1) translation from the source language into a subset of the interlingua, (2) translation between subsets of the interlingua, and (3) translation from a subset of the interlingua into the target language. The first translation step into the interlingua can typically be specified in the form of a grammar that describes how each top-level form in the source language translates into the interlingua. We observe that in cases where the source language does not have a declarative semantics, such a grammar is also a specification of a declarative semantics for the language. We describe the conditions under which such a grammar is reversible so that the grammar can also be used to translate out of the interlingua. In particular, we formally describe the translation process into and out of an interlingua, present a method for determining whether a given grammar in fact specifies how to construct a translation for every top-level form in a given source language; and present a method for determining whether a given grammar is reversible so that it can be used to translate both into and out of an interlingua.

Specifying and Checking Fault-Tolerant Agent-Based Protocols Using Maude

Fault tolerance is an important issue in mobile, agent-based computing systems. However, most research in this area has focused on security and mobility issues. The DaAgent (Dependable Agent) system [8] is similar to several other agentbased computingsystems includingAg ent-Tcl [6], Messengers [5], and Ajanta [9]. However, unlike these systems DaAgent is being designed to address fault tolerance issues. Within the DaAgent system several fault tolerant protocols are being investigated. These protocols have been speci.ed in natural language, English prose, and a Java implementation within the DaAgent system is being tested. This approach has proved to be extremely time-consumingand in.exible, for example, it is di.cult to rapidly change test conditions and fault injection is extremely primitive involvingph ysically haltingor resetinga machine, pulling network connections, or sending a kill message to the agent process. Testing the implementation of course serves as a weak form of evidence of correctness, but o.ers little real assurance of the correctness of the system.

Automated Procedure Construction for Deductive Synthesis

Deductive program synthesis systems based on automated theorem proving offer the promise of software that is correct by construction. However, the difficulty encountered in constructing usable deductive synthesis systems has prevented their widespread use. Amphion is a real-world, domain-independent, completely automated program synthesis system. It is specialized to specific applications through the creation of an operational domain theory and a specialized deductive engine. This paper describes an experiment aimed at making the construction of usable Amphion applications easier.The software system Theory Operationalization for Program Synthesis (TOPS) has a library of decision procedure templates with a theory schema for each procedure. TOPS identifies sets of axioms in the domain theory that are instances of theory schema associated with library procedures. For each procedure instance, TOPS uses iterated partial deduction to augment the procedure with the capability to construct ground terms for deductive synthesis. Synthesized procedures are interfaced to a resolution theorem prover. Axioms in the original domain theory that are implied by the synthesized procedures are removed.The inference rules of the theorem prover have been extended so that during deductive synthesis, each procedure is invoked to test conjunctions of literals in the language of the theory of that procedure. When possible, the procedure generates ground terms and binds them to variables in a problem specification. These terms are program fragments. Experiments show that the procedures synthesized by TOPS can reduce theorem proving search at least as much as hand tuning of the deductive synthesis system.

JOSIE: an integration of specialized representation and reasoning tools

We present an overview of an implemented knowledge system development tool, called JOSIE, that provides a general framework for integrating specialized representation and reasoning facilities. In our presentation we emphasize the problematic issues that arise when integrating specialized representations and the distinctive features of the facilities included in the system. The current system includes assertion, retraction, and query facilities using a predicate-calculus style interface language, a default sentential representation for the interface language, an inference rule language and forward chaining interpreter based on the interface language, a justification-based truth maintenance system that supports restricted nonmonotonic proofs, a frame-based specialized representation system that provides arbitrarily nested slot descriptions and default slot values, and a constraint language and reasoner within the frame system that provides both constraint propagation and symbolic solution of linear equations via Gaussian elimination.

Automated Protocol Analysis in Maude

We present an approach to automated protocol analysis using the Maude rewrite system. We describe the process of translating high-level protocol specifications from the Common Authentication Protocol Specification Language (CAPSL) into an executable Maude specification. After this translation step, we dynamically extend the specification by adding an intruder agent. Based on this representation of a protocol environment, different model checkers can be applied to evaluate whether all protocol properties hold in the presence of the intruder.

An Evolutionary Algorithm for Autonomous Robot Navigation

This paper presents an implementation of an evolutionary algorithm to control a robot with autonomous navigation in avoiding obstacles. The paper describes how the evolutionary system controls the sensors and motors in order to complete this task. A simulator was developed to test the algorithm and its configurations. The tests were performed in a simulated environment containing a set of barriers that were observed by means of a set of sensors. The solution obtained in the simulator was embedded in a real robot, which was tested in an arena containing obstacles. The robot was able to navigate and avoid the obstacles in this environment.

Node-depth phylogenetic-based encoding, a spanning-tree representation for evolutionary algorithms. part I: Proposal and properties analysis

Abstract Representation choice and the development of search operators are crucial aspects of the efficiency of Evolutionary Algorithms (EAs) in combinatorial problems. Several researchers have proposed representations and operators for EAs that manipulate spanning trees. This paper proposes a new encoding called Node-depth Phylogenetic-based Encoding (NPE). NPE represents spanning trees by the relation between nodes and their depths using a relatively simple codification/decodification process. The proposed NPE operators are based on methods used for tree rearrangement in phylogenetic tree reconstruction: subtree prune and regraft; and tree bisection and reconstruction. NPE and its operators are designed to have high locality, feasibility, low time complexity, be unbiased, and have independent weight. Therefore, NPE is a good choice of data structure for EAs applied to network design problems.