October 17-19, 2001ESA Workshop “On-Board Autonomy” Efficiency Issues in Model-Based Approaches to On-Board Diagnosis P.Torasso, C.Picardi and L. Console.

Slides:



Advertisements
Similar presentations
2 Introduction A central issue in supporting interoperability is achieving type compatibility. Type compatibility allows (a) entities developed by various.
Advertisements

Heuristic Search techniques
Comparative Succinctness of KR Formalisms Paolo Liberatore.
Consistency-Based Diagnosis
1 Christophe S. Jelger, Michael Kleis, Burak Simsek, Rolf Stadler, Ralf König, Danny Raz Theories/formal methods in support of autonomic management Dagstuhl.
We build generic tools for automated diagnosis & reconfiguration of distributed dynamic systems Model-Based Supervision of Composite Systems Modularity.
Supporting Business Decisions Expert Systems. Expert system definition Possible working definition of an expert system: –“A computer system with a knowledge.
Ai in game programming it university of copenhagen Reinforcement Learning [Outro] Marco Loog.
AeroSense, April System Health Tracking and Safe Testing André Bos, Arjan van Gemund Jonne Zutt Delft University of Technology.
Display of Information for Time-Critical Decision Making Eric Horvitz Decision Theory Group Microsoft Research Redmond, Washington 98025
© 2005 Prentice Hall6-1 Stumpf and Teague Object-Oriented Systems Analysis and Design with UML.
Luigi Portinale, Pietro Torasso and Diego Margo Selecting Most Adaptable Diagnostic Solutions through Pivoting-Based Retrieval Teacher : C.S. Ho Student.
Knowledge Acquisitioning. Definition The transfer and transformation of potential problem solving expertise from some knowledge source to a program.
Models -1 Scientists often describe what they do as constructing models. Understanding scientific reasoning requires understanding something about models.
Scientific Thinking - 1 A. It is not what the man of science believes that distinguishes him, but how and why he believes it. B. A hypothesis is scientific.
Models of Human Performance Dr. Chris Baber. 2 Objectives Introduce theory-based models for predicting human performance Introduce competence-based models.
Developing Ideas for Research and Evaluating Theories of Behavior
Sheila McIlraith, Knowledge Systems Lab, Stanford University DX’00, 06/2000 Diagnosing Hybrid Systems: A Bayesian Model Selection Approach Sheila McIlraith.
Introduction to Model- Based Diagnosis Meir Kalech Partially based on the slides of Peter Struss.
1 Software Testing Techniques CIS 375 Bruce R. Maxim UM-Dearborn.
Leroy Garcia 1.  Artificial Intelligence is the branch of computer science that is concerned with the automation of intelligent behavior (Luger, 2008).
Romaric GUILLERM Hamid DEMMOU LAAS-CNRS Nabil SADOU SUPELEC/IETR ESM'2009, October 26-28, 2009, Holiday Inn Leicester, Leicester, United Kingdom.
LÊ QU Ố C HUY ID: QLU OUTLINE  What is data mining ?  Major issues in data mining 2.
A learning-based transportation oriented simulation system Theo A. Arentze, Harry J.P. Timmermans.
1 Performance Evaluation of Computer Networks: Part II Objectives r Simulation Modeling r Classification of Simulation Modeling r Discrete-Event Simulation.
Design Science Method By Temtim Assefa.
Leslie Luyt Supervisor: Dr. Karen Bradshaw 2 November 2009.
11 C H A P T E R Artificial Intelligence and Expert Systems.
Christian Heinzemann 11. Oktober 2015 Modeling Behavior of Self-Adaptive Systems Seminar Software Quality and Safety.
1 USC INFORMATION SCIENCES INSTITUTE CALO, 8/8/03 Acquiring advice (that may use complex expressions) and action specifications Acquiring planning advice,
Fuzzy Genetic Algorithm
1 Introduction to Software Engineering Lecture 1.
Copyright © 2011 by The McGraw-Hill Companies, Inc. All rights reserved. McGraw-Hill/Irwin Developing and Evaluating Theories of Behavior.
ACS'08, November, Venice, ITALY Designing organic reaction simulation engine using qualitative reasoning approach Y.C. Alicia Tang Tenaga Nasional.
Approximate Dynamic Programming Methods for Resource Constrained Sensor Management John W. Fisher III, Jason L. Williams and Alan S. Willsky MIT CSAIL.
Model Checking and Model-Based Design Bruce H. Krogh Carnegie Mellon University.
I Robot.
Indirect Supervision Protocols for Learning in Natural Language Processing II. Learning by Inventing Binary Labels This work is supported by DARPA funding.
Intelligent Agents อาจารย์อุทัย เซี่ยงเจ็น สำนักเทคโนโลยีสารสนเทศและการ สื่อสาร มหาวิทยาลัยนเรศวร วิทยาเขต สารสนเทศพะเยา.
Agents that Reduce Work and Information Overload and Beyond Intelligent Interfaces Presented by Maulik Oza Department of Information and Computer Science.
Artificial intelligence
Algorithmic, Game-theoretic and Logical Foundations
2nd Meeting of Young Researchers on MULTIPLE CRITERIA DECISION AIDING Iryna Yevseyeva Niilo Mäki Instituutti University of Jyväskylä, Finland
MODEL-BASED SOFTWARE ARCHITECTURES.  Models of software are used in an increasing number of projects to handle the complexity of application domains.
Distributed Models for Decision Support Jose Cuena & Sascha Ossowski Pesented by: Gal Moshitch & Rica Gonen.
Towards a Reference Quality Model for Digital Libraries Maristella Agosti Nicola Ferro Edward A. Fox Marcos André Gonçalves Bárbara Lagoeiro Moreira.
Outline Introduction Research Project Findings / Results
Discovery and Systems Health Technical Area NASA Ames Research Center - Computational Sciences Division Automated Diagnosis Sriram Narasimhan University.
SAFEWARE System Safety and Computers Chap18:Verification of Safety Author : Nancy G. Leveson University of Washington 1995 by Addison-Wesley Publishing.
Artificial Intelligence: Research and Collaborative Possibilities a presentation by: Dr. Ernest L. McDuffie, Assistant Professor Department of Computer.
Scientific Method 1.Observe 2.Ask a question 3.Form a hypothesis 4.Test hypothesis (experiment) 5.Record and analyze data 6.Form a conclusion 7.Repeat.
Testing Overview Software Reliability Techniques Testing Concepts CEN 4010 Class 24 – 11/17.
Introduction: What is AI? CMSC Introduction to Artificial Intelligence January 7, 2003.
Control-Theoretic Approaches for Dynamic Information Assurance George Vachtsevanos Georgia Tech Working Meeting U. C. Berkeley February 5, 2003.
Science and Engineering Practices K–2 Condensed Practices3–5 Condensed Practices6–8 Condensed Practices9–12 Condensed Practices Developing and Using Models.
Survey on Expert System Seung Jun Lee Dept. of Nuclear and Quantum Engineering KAIST Mar 3, 2003.
D10A Metode Penelitian MP-04b Metodologi Penelitian di dalam Ilmu Komputer/Informatika Program Studi S-1 Teknik Informatika FMIPA Universitas.
On the Relation Between Simulation-based and SAT-based Diagnosis CMPE 58Q Giray Kömürcü Boğaziçi University.
Wolfgang Runte Slide University of Osnabrueck, Software Engineering Research Group Wolfgang Runte Software Engineering Research Group Institute.
Introduction to Machine Learning, its potential usage in network area,
CS 4700: Foundations of Artificial Intelligence
Model-based Diagnosis: The Single Fault Case
Intelligent Systems (AI-2) Computer Science cpsc422, Lecture 7
It is great that we automate our tests, but why are they so bad?
Toward a Reliable Evaluation of Mixed-Initiative Systems
Systems Engineering for Mission-Driven Modeling
Market-based Dynamic Task Allocation in Mobile Surveillance Systems
Decision Support Systems
Test-Driven Ontology Development in Protégé
Intelligent Systems (AI-2) Computer Science cpsc422, Lecture 7
Presentation transcript:

October 17-19, 2001ESA Workshop “On-Board Autonomy” Efficiency Issues in Model-Based Approaches to On-Board Diagnosis P.Torasso, C.Picardi and L. Console Dipartimento di Informatica- Universita’ di Torino Italy

October 17-19, 2001ESA Workshop “On-Board Autonomy” Autonomy and Diagnosis Onboard autonomy requires several components: planning and scheduling are very crucial Autonomy requires also the ability of detecting what is going wrong and reacting Diagnostic reasoning is an important step Usually several competing diagnostic hypotheses Need of informing the human operator about diagnostic conclusions? Need of informing other software components

October 17-19, 2001ESA Workshop “On-Board Autonomy” Knowledge-based approaches to diagnosis Develop a domain theory for the system to be diagnosed Logical/qualitative models Correct and/or faulty behavior Usually component oriented Reasoning mechanisms for prediction and post- diction In-depth understanding of diagnostic reasoning (theory of diagnosis)

October 17-19, 2001ESA Workshop “On-Board Autonomy” Challenges of On-board Diagnosis (I) Off-board diagnosis is an iterative process Available observations are used for making hypotheses about faults Model-based approaches to diagnosis are able to suggest probes Revision of the diagnostic hypotheses on the basis of new measurements Strong interaction with the human agent On-board diagnosis has different requirements

October 17-19, 2001ESA Workshop “On-Board Autonomy” Challenges of On-board Diagnosis (II) Observations provided just via sensors Strict time constraints for producing diagnostic conclusions Difficulty/impossibility of acquiring extra data Usually a large number of potential diagnoses Need of complex reasoning mechanisms, very expensive from a computational point of view Sometimes, on-board computation resources are limited

October 17-19, 2001ESA Workshop “On-Board Autonomy” Efficiency issues Decomposing static and time-varying aspects in diagnostic problem solving [Console et al., Ann.Math&AI, 94] [Williams&Nayak, AAAI96] [Struss et al., DX96] Exploiting observations for decomposing diagnostic problems [e.g. Darwiche IJCAI 99] Exploiting hierarchical models for focussing diagnostic reasoning [Mozetic, IJMMS 91], [Out et al., Ann.Math&AI 94, Renon, IJCAI01]

October 17-19, 2001ESA Workshop “On-Board Autonomy” Towards On-board Diagnosis At Università di Torino work on on-board diagnosis in the automotive sector: European projects VMBD and IDD for the automotive sector in the space sector ASI sponsored basic research project on “An Intelligent System for Supervising Autonomous Space Robots.” in co-operation with other Italian universities, CNR-IP

October 17-19, 2001ESA Workshop “On-Board Autonomy” The compilation approach to on-board diagnosis (I) Developed inside VMBD project [Cascio et al. AI Comm 99, Console et al, IJCAI 01] Use of quite expressive qualitative models taking into account temporal and dynamic aspects of the system to be diagnosed Diagnostic reasoning on such models too expensive for on-board computing Off-line qualitative simulation of interesting diagnostic cases

October 17-19, 2001ESA Workshop “On-Board Autonomy” The compilation approach to on-board diagnosis (II) Automatic learning of decision trees starting from diagnostic cases obtained via simulation and fault injection Use of the decision tree for on-board diagnosis ==> fast and small, suitable for current ECU The nodes of the decision tree contain measurements provided by sensors in different time points and leaves contain (repair) actions Discrimination among faults done as long as different actions can be done Discriminative power limited by lack of sensors

October 17-19, 2001ESA Workshop “On-Board Autonomy” Exploring diagnostic problem solving: the SPIDER case study Investigated within a ASI supported project [Portinale&Torasso ISAIRAS 99, IJCAI 99] A logical model relating behavioral modes of components of the SPIDER arm and contextual information with qualitative manifestations (based on FMECA documents available) Innovative diagnostic strategies based on the notion of “Variable assignment problems” Exploiting analogies with constraint satisfaction problem for a compact representations of a set of diagnoses (“scenario”)

October 17-19, 2001ESA Workshop “On-Board Autonomy” Diagnostic Agent Diagnosis as explanation of observations DT behavioral model of SPIDER (nominal and faulty modes) COMP the set of the components in SPIDER CXT ground atoms modeling contextual information OBS ground atoms representing observations

October 17-19, 2001ESA Workshop “On-Board Autonomy” Diagnostic Agent (II) For most diagnostic problems a large number of diagnoses Preference criterion: based on a measure related to MDL Preferred solutions involving OK mode for most components In many cases no sufficient information for discriminating among alternative (preferred) diagnoses

October 17-19, 2001ESA Workshop “On-Board Autonomy” A diagnostic case where only qualitative measurements provided sensors are available

October 17-19, 2001ESA Workshop “On-Board Autonomy” Compact representation of multiple-fault diagnoses for the diagnostic case (only sensor data)

October 17-19, 2001ESA Workshop “On-Board Autonomy” The same diagnostic case where all qualitative measurements are available

October 17-19, 2001ESA Workshop “On-Board Autonomy” One multiple-fault diagnosis for the previous diagnostic case (all observations are available)

October 17-19, 2001ESA Workshop “On-Board Autonomy” Discriminability among diagnoses Goal: Formalize the notion that two diagnoses cannot be discriminated using a given type of measurements Classes of observations: sensorized vs not sensorized In SPIDER domain several classes:joint positions, temperature sensors, observations under human control

October 17-19, 2001ESA Workshop “On-Board Autonomy” Discriminability among diagnoses (II) Single out the behavioral modes of the components which cannot be discriminated using a class of manifestations Independent on the specific diagnostic problem at hand

October 17-19, 2001ESA Workshop “On-Board Autonomy” Finding indistinguishable behavioral modes Injecting fault bm r (and bm s ) for component C i Evaluate the transitive closure for all possible contexts Check if any difference is observable with meaurements in class CL j

October 17-19, 2001ESA Workshop “On-Board Autonomy” Automatic generation of abstract models If diagnostic problem DP involves just manifestations of class i, some faults may be indistinguishable ==> Generation of alternative diagnoses that cannot be discriminated Computation time is consumed Human or artificial supervisor may be confused Proposed solution: automatic generation of abstract models where indistinguishable modes are collapsed

October 17-19, 2001ESA Workshop “On-Board Autonomy” Automatic generation of abstract models (II) A simplied form of the algorithm for: collapsing behavioral modes revising domain theory

October 17-19, 2001ESA Workshop “On-Board Autonomy” Automatic generation of abstract models (III) Off-line: generate an abstract model for each class of measurements (if there are indistinguishable modes) On-line: for each diagnostic problem DP, select (from the model library) the model which fits available measurements Benefits: saving in computation time reduction in the number of diagnoses (if bm r and bm s for component C i are indistinguishable instead of two diagnoses just one involving bm r+s ) no loss of information

October 17-19, 2001ESA Workshop “On-Board Autonomy” Abstract models and abstraction of manifestations

October 17-19, 2001ESA Workshop “On-Board Autonomy” using the abstract model just single fault diagnoses

October 17-19, 2001ESA Workshop “On-Board Autonomy” Without abstract model, generation of multiple faults diagnoses with very low plausibility

October 17-19, 2001ESA Workshop “On-Board Autonomy” Conclusions (I) On-board diagnosis is an interesting and challenging problem Need of multiple approaches: efficient diagnostic strategies partitioning the diagnostic problem compilation techniques automatic generation of abstract models if modes are indistinguishable taking recovery into consideration

October 17-19, 2001ESA Workshop “On-Board Autonomy” Conclusions (II) Interesting results obtained in the automotive sector compilation techniques allow to put on-board diagnostic procedures which are automatically generated starting from a model of the artifact diagnosability is an important aspect

October 17-19, 2001ESA Workshop “On-Board Autonomy” Conclusions (III) Efficiency is just one aspect How to involve the human supervisor in the loop? ==> Interactive autonomy How to make the result of the diagnostic agent understandable to the human supervisor? Some work in: compact graphical representation of diagnoses preference criteria among diagnoses informing the supervisor about the impossibility of discriminating among diagnoses