James G. Schmolze

ICAPS Conference 2000 Conference Paper

PSIPLAN: Open World Planning with y-Forms

• Tamara Babaian
• James G. Schmolze

Wepresent a new, partial order planner called PSIPLAN, which builds on SNLP. Wedrop the closed worldassumption, addsensingactions, "~ld a class of propositions about the agent’s knowledge, and add a class of universallyquantifiedpropositions. Thislatter class of propositions, whichwecall C-forms, distinguishesthis research. ~, -formsrepresentpartially closedworlds, suchas "’Block. 4 is clear", or "z. ps is the only postscript file in directory/tez. "Wepresent our theory"of planningwith sensingandshow, hob-partial orderplanningis performed usingI/~-forms. Noteworthyare the facts that lack of information canbe representedpreciselyandall quantifiedreasoninghas polynomialcomplexity. Thus, in finite domainswhere the maximum plan length is bounded, planning with PSIPLAN is NP-complete. with -forms. James G. Schmolze Dept. of Electrical Eng. and Computer Science Tufts University Medford, MA 02155USA schmolze~eecs. tufts. edu http: //www. eccs. tufts. edu/~schmolze sensing actions and uncertain effects. Wehave developed a planning formalismcalled PSIPLAN and a sound and complete partial order plmmer (POP) called PSIPOPthat uses PSIPLAN to plan open worlds without sensing. Moreover, we have extended both PSIPLANand PSIPOPto handle sensing actions, knowledgegoals, information loss and conditional effects. In this paper, we focus on demonstratingthe power of our ~-form-based language PSIPLAN, discussing the issues critical to its soundnessand completeness in open world planning, and extending the standard POPalgorithm to produce PSIPOP. Representing Open Worlds We consider theproblem ofopenworld planning where theagentdoesnothavecomplete information about theworld. We assume thattheworldevolves as a sequence ofstates, wherethetransitions occur only astheresult ofdeliberate action taken bythesingle agent. Since the agent’s modelof the world is incomplete, wemust distinguish betweenthe world state (or state of the world, or situation, in situation calculus terms (McCarthy& Hayes1969)) and the state of the agent’s knowledgeof the world, which we call SOK. Wefurther assumethat the agent’s knowledgeof the world is

AAAI Conference 1997 Conference Paper

Detecting Redundant Production Rules

• James G. Schmolze

AAAI Conference 1992 Conference Paper

Comparison of Three Algorithms for Ensuring Serializable Executions in Parallel Production Systems

• James G. Schmolze

To speed up production systems, researchers have developed parallel algorithms that execute multiple instantiations simultaneously. Unfortunately, without special controls, such systems can produce results that could not have been produced by any serial execution. We present and compare three different algorithms that guarantee a serializable result in such systems. Our goal is to analyze the overhead that serialization incurs. All three algorithms perform synchronization at the level of instantiations, not rules, and are targeted for shared-memory machines. One algorithm operates synchronously while the other two operate asynchronously. Of the latter two, one synchronizes instantiations using compiled tests that were determined from an offline analysis while the other uses a novel locking scheme that requires no such analysis. Our examination of performance shows that asynchronous execution is clearly faster than synchronous execution and that the locking method is somewhat faster than the method using compiled tests. Moreover, we predict that the synchronization and/or locking needed to guarantee serializability will limit speedup no matter how many processors are used.

AAAI Conference 1990 Conference Paper

A Parallel Asynchronous Distributed Production System

• James G. Schmolze

To speed up production systems, many researchers have turned to parallel implementations. We describe a system called PARS that executes production rules in parallel. PARS is novel because it (1) executes many rules simultaneously, (2) runs in a highly asynchronous fashion, and (3) runs on a distributed memory machine. Item (1) improves available concurrency over systems that only perform the MATCH step in parallel. Item (2) reduces bottlenecks over synchronous parallel production systems. Item (3) makes the techniques more available given the lower cost of distributed versus shared memory machines. The two main problems regarding correctness, namely serialization and the maintenance of consistent distributed databases, are addressed and solved. Estimates of the effectiveness of this approach are also given.

AAAI Conference 1986 Conference Paper

Physics for Robots

• James G. Schmolze

Robots that plan to perform everyday tasks need knowledge of everyday physics. Physics For Robots (PFR) is a representation of part of everyday physics directed towards this need. It includes general concepts and theories, and it has been applied to tasks in cooking. PFR goes beyond most AI planning representation schemes by including natural processes that the robot can control. It also includes a theory of material composition so robots can identify and reason about physical objects that break apart, come together, mix, or go out of existence. Following on Naive Physics (NP), issues about reasoning mechanisms are temporarily postponed, allowing a focus on the characterization of knowledge. However, PFR departs from NP in two ways. (1) PFR characterizes the robot’s capabilities to act and perceive, and (2) PFR replaces the NP goal of developing models of actual common sense knowledge. Instead, PFR includes all and only the knowledge that robots need for planning, which is determined by analyzing proofs showing the effectiveness of robot I/O programs.

IJCAI Conference 1983 Conference Paper

Classification in the KL-ONE Knowledge Representation System

• James G. Schmolze
• Thomas A. Lipkis