📄 http:^^www.cs.utexas.edu^users^qr^abstracts.html
字号:
MIME-Version: 1.0
Server: CERN/3.0
Date: Tuesday, 07-Jan-97 15:51:07 GMT
Content-Type: text/html
Content-Length: 35995
Last-Modified: Monday, 18-Nov-96 16:14:57 GMT
<TITLE>Dissertation Abstracts</TITLE><!-- Changed by: Benjamin J. Kuipers, 22-Mar-1996 --><body bgcolor="#ffffff" text="#000000" link="#0000ee" vlink="551a8b" alink="ff0000"><H1>Dissertation Abstracts</H1><hr><ul> <li> <a href="#Berleant">Daniel Berleant</a>. 1991. <li> <a href="#Byun">Yung-Tai Byun</a>. 1990. <li> <a href="#Crawford">James Crawford</a>. 1990. <li> <a href="#Dvorak">Daniel Dvorak</a>. 1992. <li> <a href="#Farquhar">Adam Farquhar</a>. 1993. <li> <a href="#Franke">David Franke</a>. 1993. <li> <a href="#Froom">Richard Froom</a>. 1995. <li> <a href="#Hartman">John Hartman</a>. 1991. <li> <a href="#Akira">Akira Hayashi</a>. 1991. <li> <a href="#Kay">Herbert Kay</a>. 1996. New! <li> <a href="#WYLee">Wan Yik Lee</a>. 1996. New! <li> <a href="#WoodWLee">Wood Wai Lee</a>. 1993. <li> <a href="#Pierce">David Pierce</a>. 1995. <li> <a href="#Raman">Raman Rajagopalan</a>. 1995. <li> <a href="#Throop">David Throop</a>. 1991.</ul><hr><H2><a name="Berleant">The Use of Partial Quantitative Information with Qualitative Reasoning</a></H2>Daniel Berleant. 1991.<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Berleant-PhD-91.ps.Z"><b>The Use of Partial Quantitative Information with Qualitative Reasoning</b></a>Doctoral dissertation, Department of Computer Sciences, The University of Texas at Austin. <p><H3>Abstract</H3>There is a need for combining qualitative and quantitativesimulations, to do simulation tasks that would be difficult usingeither alone. This task is made more difficult by the fact thatavailable quantitative information may be incomplete, bounding valueswith intervals or describing them with probability distributionfunctions. This research demonstrates the combination of qualitativeand quantitative simulation in an implemented system, Q3. Q3 utilizespartial or complete quantitative information, to gradually refine aqualitative simulation into a simulation that has properties andadvantages of both qualitative simulations and quantitative ones.<p>The technique exemplified by Q3 is shown to possess properties oftenused in analyzing both qualitative and quantitative simulators.Qualitative and quantitative inferences are correct. Theoreticalconvergence to the true solution and stability in the presence ofpartial model inputs are also shown. <p>Q3 has been applied to the problem of finding probabilities ofqualitative behaviors, an important problem. Partial quantitativecharacterization of model inputs, in the form of intervals andprobability distributions, may be used to bound the probabilities ofdifferent behaviors. This is demonstrated for simple models includingone in the dependability analysis application domain. <p><hr><H2><a name="Byun">Spatial Learning Mobile Robots with a Spatial Semantic Hierarchical Model</a></H2>Yung-Tai Byun. 1990.<b>Spatial Learning Mobile Robots with a Spatial Semantic Hierarchical Model</b>Doctoral dissertation, Department of Computer Sciences,The University of Texas at Austin. <p><H3>Abstract</H3>The goal of this dissertation is to develop a spatial exploration andmap-learning strategy for a mobile robot to use in unknown,large-scale environments. Traditional approaches aim at buildingpurely metrically accurate maps. Because of sensorimotor errors, itis hard to construct accurately such maps. However, in spite ofsensory and computation limitation, humans explore environments, buildcognitive maps from exploration, and successfully path-plan, navigate,and place-find. Based on the study of human cognitive maps, wedevelop a spatial semantic hierarchical model to replace the globalabsolute coordinate frame used in traditional approaches. Thesemantic hierarchical model consists of three levels: control level,topological level, and geometrical level. The topological levelprovides the basic structure of the hierarchy. <p>At the control level, a robot finds places or follows travel edgeswhich can be described by qualitatively definable features. Thedistinctive features allow development of distinctiveness measures.The robot uses these measures to find, with negative feedback control,the distinctive places by hill-climbing search algorithms, and thetravel edges by edge-following algorithms. Distinctive places andtravel edges are connected to build a topological model. This modelis created prior to the construction of a global geometrical map.Cumulative location error is essentially eliminated while travelingamong distinctive places and travel edges by alternating between thehill-climbing search control algorithms and the edge-following controlalgorithms. On top of the topological model, metrical information isaccumulated first locally and then globally. <p>Using a simulation package with a robot instance, NX, we demonstratethe robustness of our method against sensorimotor errors. The controlknowledge for distinctive places and travel edges, the topologicalmatching process, and the metrical matching process with localgeometry make our approach robust in the face of metrical errors. Inaddition to robust navigation at the control and topological levels,our framework can incorporate certain metrically-based methods andthus provide the best of both approaches. <p><hr><H2><a name="Crawford">Access-Limited Logic -- A Language for Knowledge Representation</a></H2>James Crawford. 1990.<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Crawford-PhD-91.ps.Z"><b>Access-Limited Logic -- A Language for Knowledge Representation</b></a>Doctoral dissertation, Department of Computer Sciences,The University of Texas at Austin. <p><H3>Abstract</H3> Access-Limited Logic (ALL) is a language for knowledge representationwhich formalizes the access limitations inherent in a networkstructured knowledge-base. Where a deductive method such asresolution would retrieve all assertions that satisfy a given pattern,an access-limited logic retrieves all assertions reachable byfollowing an available access path. The time complexity of inferenceis thus a polynomial function of the size of the accessible portion ofthe knowledge-base, rather than the size of the entire knowledge-base.Access-Limited Logic, though incomplete, still has a well definedsemantics and a weakened form of completeness, Socratic Completeness,which guarantees that for any query which is a logical consequence ofthe knowledge-base, there exists a series of queries after which theoriginal query will succeed. We have implemented ALL in Lisp and ithas been used to build several non-trivial systems, including versionsof Qualitative Process Theory and Pearl's probability networks. ALLis a step toward providing the properties - clean semantics, efficientinference, expressive power - which will be necessary to build large,effective knowledge bases. <p><hr><H2><a name="Dvorak">Monitoring and Diagnosis of Continuous Dynamic Systems Using Semiquantitative Simulation</a></H2>Daniel Louis Dvorak. 1992.<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Dvorak-PhD-92.ps.Z"><b>Monitoring and Diagnosis of Continuous Dynamic Systems Using Semiquantitative Simulation</b></a>Doctoral dissertation, Department of Computer Sciences,The University of Texas at Austin. <p><H3>Abstract</H3> Operative diagnosis or diagnosis of a physical system in operation,is essential for systems that cannot be stopped every time an anomalyis detected, such as in the process industries, space missions, andmedicine. Compared to maintenance diagnosis where the system isoffline and arbitrary points can be probed, operative diagnosis islimited mainly to sensor readings, and diagnosis begins while theeffects of a fault are still propagating. Symptoms change as thesystem's dynamic behavior unfolds. <p>This paper presents a design for monitoring and diagnosis ofdeterministic continuous dynamic systems based on the paradigms of"monitoring as model corroboration" and "diagnosis as modelmodification" in which a semiquantitative model of aphysical system issimulated in synchrony with incoming sensor readings. When sensorreadings disagree with predictions, variant models are createdrepresenting different fault hypotheses. These models are thensimulated and either corroborated or refuted as new readings arrive.The set of models changes as new hypotheses are generated and as oldhypotheses are exonerated. In contrast to methods that base diagnosison a snapshot of behavior, this simulation-based approach exploits thesystem's time-varying behavior for diagnostic clues and exploits thepredictive power of the model to forewarn of imminent hazards. <p>The design holds several other advantages over existing methods: 1)semiquantitative models provide greater expressive power for states ofincomplete knowledge than differential equations, thus eliminatingcertain modeling compromises; 2) semiquantitative simulation generatesguaranteed bounds on variables, thus providing dynamic alarmthresholds and thus fewer fault detection errors than withfixed-threshold alarms; 3) the guaranteed prediction of all validbehaviors eliminates the "missing prediction bug" in diagnosis; 4) thebranching-time descruption of behavior permits recognition of allvalid manifestations of a fault (and of interacting faults); 5)hypotheses based on predictive semiquantitative models are moreinformative because they show the values of unseen variables and canpredict future consequences; and 6) fault detection degradesgracefully as multiple faults are diagnosed over time. <p><hr><H2><a name="Farquhar">Automated Modeling of Physical Systems in the Presence of Incomplete Knowledge</a></H2> Adam Farquhar. 1993.<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Farquhar-PhD-93.ps.Z"><b>Automated Modeling of Physical Systems in the Presence of Incomplete Knowledge</b></a>Doctoral dissertation, Department of Computer Sciences,The University of Texas at Austin. <p><H3>Abstract</H3>This dissertation presents an approach to automated reasoning aboutphysical systems in the presence of {\em incomplete knowledge} whichsupports formal analysis, proof of guarantees, has been fullyimplemented, and applied to substantial domain modeling problems.Predicting and reasoning about the behavior of physical systems is adifficult and important task that is essential to everyday commonsensereasoning and to complex engineering tasks such as design, monitoring,control, or diagnosis. <p>A capability for automated modeling and simulation requires <ul> <li> expressiveness to represent incomplete knowledge, <li> algorithms to draw useful inferences about non-trivial systems,and <li> precise semantics to support meaningful guarantees ofcorrectness. </ul>In order to clarify the structure of the knowledge required forreasoning about the behavior of physical systems, we distinguishbetween the <i>model building</i> task which builds a model to describethe system, and the <i>simulation</i> task which uses the model togenerate a description of the possible behaviors of the system. <p>This dissertation describes QPC, an implemented approach to reasoningabout physical systems that builds on the expressiveness ofQualitative Process Theory [Forbus, 1984] and the mathematicalrigor of the QSIM qualitative simulation algorithm [Kuipers, 1986]. <p>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -