Probabilistic Reasoning for Robust Plan Execution Steve Schaffer, Brad Clement, Steve Chien Artificial Intelligence.

Slides:



Advertisements
Similar presentations
Simulation - An Introduction Simulation:- The technique of imitating the behaviour of some situation or system (economic, military, mechanical, etc.) by.
Advertisements

Dialogue Policy Optimisation
Jose-Luis Blanco, Javier González, Juan-Antonio Fernández-Madrigal University of Málaga (Spain) Dpt. of System Engineering and Automation May Pasadena,
1 ECE 776 Project Information-theoretic Approaches for Sensor Selection and Placement in Sensor Networks for Target Localization and Tracking Renita Machado.
Structural reliability analysis with probability- boxes Hao Zhang School of Civil Engineering, University of Sydney, NSW 2006, Australia Michael Beer Institute.
Partially Observable Markov Decision Process (POMDP)
SA-1 Probabilistic Robotics Planning and Control: Partially Observable Markov Decision Processes.
Planning with Resources at Multiple Levels of Abstraction Brad Clement, Tony Barrett, Gregg Rabideau Artificial Intelligence Group Jet Propulsion Laboratory.
Sensitivity Analysis In deterministic analysis, single fixed values (typically, mean values) of representative samples or strength parameters or slope.
Probabilistic Re-Analysis Using Monte Carlo Simulation
Southwest Research Institute, San Antonio, Texas A Framework to Estimate Uncertain Random Variables Luc Huyse and Ben H. Thacker Reliability and Materials.
GoldSim 2006 User Conference Slide 1 Vancouver, B.C. The Submodel Element.
Effectively Indexing Uncertain Moving Objects for Predictive Queries School of Computing National University of Singapore Department of Computer Science.
©GoldSim Technology Group LLC., 2004 Probabilistic Simulation “Uncertainty is a sign of humility, and humility is just the ability or the willingness to.
Artificial Learning Approaches for Multi-target Tracking Jesse McCrosky Nikki Hu.
Sam Pfister, Stergios Roumeliotis, Joel Burdick
FIN 685: Risk Management Topic 5: Simulation Larry Schrenk, Instructor.
Resampling techniques Why resampling? Jacknife Cross-validation Bootstrap Examples of application of bootstrap.
A Principled Information Valuation for Communications During Multi-Agent Coordination Simon A. Williamson, Enrico H. Gerding, Nicholas R. Jennings School.
Robust Monte Carlo Localization for Mobile Robots
Improving the Accuracy of Continuous Aggregates & Mining Queries Under Load Shedding Yan-Nei Law* and Carlo Zaniolo Computer Science Dept. UCLA * Bioinformatics.
Maximum Likelihood (ML), Expectation Maximization (EM)
Planning operation start times for the manufacture of capital products with uncertain processing times and resource constraints D.P. Song, Dr. C.Hicks.
Using ranking and DCE data to value health states on the QALY scale using conventional and Bayesian methods Theresa Cain.
Computer vision: models, learning and inference Chapter 10 Graphical Models.
Dynamic Bayesian Networks CSE 473. © Daniel S. Weld Topics Agency Problem Spaces Search Knowledge Representation Reinforcement Learning InferencePlanningLearning.
Lecture II-2: Probability Review
Particle Filtering. Sensors and Uncertainty Real world sensors are noisy and suffer from missing data (e.g., occlusions, GPS blackouts) Use sensor models.
1 Assessment of Imprecise Reliability Using Efficient Probabilistic Reanalysis Farizal Efstratios Nikolaidis SAE 2007 World Congress.
Computer Simulation A Laboratory to Evaluate “What-if” Questions.
Bayesian Filtering for Robot Localization
1 CE 530 Molecular Simulation Lecture 7 David A. Kofke Department of Chemical Engineering SUNY Buffalo
Markov Localization & Bayes Filtering
Error Analysis Accuracy Closeness to the true value Measurement Accuracy – determines the closeness of the measured value to the true value Instrument.
Michael Ernst, page 1 Collaborative Learning for Security and Repair in Application Communities Performers: MIT and Determina Michael Ernst MIT Computer.
Lecture 12 Statistical Inference (Estimation) Point and Interval estimation By Aziza Munir.
Monte Carlo Simulation CWR 6536 Stochastic Subsurface Hydrology.
Copyright © 2010 Lumina Decision Systems, Inc. Monte Carlo Simulation Analytica User Group Modeling Uncertainty Series #3 13 May 2010 Lonnie Chrisman,
General Principle of Monte Carlo Fall 2013 By Yaohang Li, Ph.D.
IV&V Facility PI: Katerina Goseva – Popstojanova Students: Sunil Kamavaram & Olaolu Adekunle Lane Department of Computer Science and Electrical Engineering.
Using Abstraction in Multi-Rover Scheduling Bradley J. Clement and Anthony C. Barrett Artificial Intelligence Group Jet Propulsion Laboratory {bclement,
Inferring High-Level Behavior from Low-Level Sensors Don Peterson, Lin Liao, Dieter Fox, Henry Kautz Published in UBICOMP 2003 ICS 280.
The Deep Space Network Scheduling Problem Brad Clement, Mark Johnston Artificial Intelligence Group Jet Propulsion Laboratory California Institute of Technology.
TKK | Automation Technology Laboratory Partially Observable Markov Decision Process (Chapter 15 & 16) José Luis Peralta.
“WHY ARE PROJECTS ALWAYS LATE?” (“and what can the Project Manager DO about that?) Craig Henderson, MBA, PMP ARVEST Bank Operations.
1 6. Reliability computations Objectives Learn how to compute reliability of a component given the probability distributions on the stress,S, and the strength,
Computer Science, Software Engineering & Robotics Workshop, FGCU, April 27-28, 2012 Fault Prediction with Particle Filters by David Hatfield mentors: Dr.
Stochastic DAG Scheduling using Monte Carlo Approach Heterogeneous Computing Workshop (at IPDPS) 2012 Extended version: Elsevier JPDC (accepted July 2013,
Mixture Models, Monte Carlo, Bayesian Updating and Dynamic Models Mike West Computing Science and Statistics, Vol. 24, pp , 1993.
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation 1 Dynamic Sensor Resource Management for ATE MURI.
Michael Isard and Andrew Blake, IJCV 1998 Presented by Wen Li Department of Computer Science & Engineering Texas A&M University.
1 / 12 Michael Beer, Vladik Kreinovich COMPARING INTERVALS AND MOMENTS FOR THE QUANTIFICATION OF COARSE INFORMATION M. Beer University of Liverpool V.
Machine Design Under Uncertainty. Outline Uncertainty in mechanical components Why consider uncertainty Basics of uncertainty Uncertainty analysis for.
OPERATING SYSTEMS CS 3530 Summer 2014 Systems and Models Chapter 03.
Monte-Carlo based Expertise A powerful Tool for System Evaluation & Optimization  Introduction  Features  System Performance.
Systems Realization Laboratory The Role and Limitations of Modeling and Simulation in Systems Design Jason Aughenbaugh & Chris Paredis The Systems Realization.
6. Population Codes Presented by Rhee, Je-Keun © 2008, SNU Biointelligence Lab,
Particle Filtering. Sensors and Uncertainty Real world sensors are noisy and suffer from missing data (e.g., occlusions, GPS blackouts) Use sensor models.
Uncertainty and Reliability Analysis D Nagesh Kumar, IISc Water Resources Planning and Management: M6L2 Stochastic Optimization.
G. Cowan Lectures on Statistical Data Analysis Lecture 9 page 1 Statistical Data Analysis: Lecture 9 1Probability, Bayes’ theorem 2Random variables and.
Håkan L. S. YounesDavid J. Musliner Carnegie Mellon UniversityHoneywell Laboratories Probabilistic Plan Verification through Acceptance Sampling.
Density Estimation in R Ha Le and Nikolaos Sarafianos COSC 7362 – Advanced Machine Learning Professor: Dr. Christoph F. Eick 1.
1 VaR Models VaR Models for Energy Commodities Parametric VaR Historical Simulation VaR Monte Carlo VaR VaR based on Volatility Adjusted.
Ocean Observatories Initiative OOI CI Kick-Off Meeting Devils Thumb Ranch, Colorado September 9-11, 2009 Observation Planning and Autonomous Mission Execution.
Using Abstraction to Coordinate Multiple Robotic Spacecraft Brad Clement, Tony Barrett, Gregg Rabideau Artificial Intelligence Group Jet Propulsion Laboratory.
Unified Adaptivity Optimization of Clock and Logic Signals Shiyan Hu and Jiang Hu Dept of Electrical and Computer Engineering Texas A&M University.
OPERATING SYSTEMS CS 3502 Fall 2017
Probabilistic Robotics
Professor S K Dubey,VSM Amity School of Business
Probabilistic Surrogate Models
Presentation transcript:

Probabilistic Reasoning for Robust Plan Execution Steve Schaffer, Brad Clement, Steve Chien Artificial Intelligence Group Jet Propulsion Laboratory California Institute of Technology

Handling Uncertainty Why? –Better account for competing requirements of safety, science, and engineering concerns –Sequence generation and validation rely on the accuracy of models –Current process limits the capabilities by improperly evaluating alternative courses of action restricting the time horizon of operations between command uplinks –Fully exploit the capabilities of a spacecraft and enable greater autonomous operation Capabilities –Avoid constraint violations –Identify sources of greatest risk –Balance risk and efficiency –Allow for varying user risk preferences –Make execution more reliable Approaches –Conservative estimates of effects –Bound resource capacities –Slack padding –Re-planning –Handle in execution system –Contingency planning

Main Idea Represent uncertainty of action effects and durations as parametric, continuous probability distribution functions Propagate distributions through plan to project states/resources Score plan based on risk –risk = probability outside limits Plan to reduce risk

“Full Probabilistic” Modeling Durations and resource usage normally distributed

Modeling Approximations Full probabilistic Means only –Track only the expected value (mean) –Same as non-probabilistic risk-ignorant Pessimistic –Track only “worst” possible value –“Worst” depends on domain / resource Single peak –For time-dependent multimodals –Track only one “average” Gaussian Chebyshev bound –Distribution-free limit on probability density –Only track the mean and standard deviation single peak

Evaluation Domains Abstract testbed –one resource –various consumers, replenishers –schedule within time horizon –conflicts resolvable Orbiter domain –image planet, process, downlink –~10 resources, ~10 actions –schedule goals within horizon –conflicts not all resolvable – must minimize

Evaluation Methodology Generate plan (batch mode) –Use different approximations –Planner is not allowed to remove goals Run plan on stochastic simulator Score execution by # errors caused –error = resource oversubscribed

Planning in ASPEN Start (if conflicts exist and user time-limit not exceeded)... Select probable conflict Select a repair method... move... Select an activity Select a start time

Results: Abstract Domain –Full probabilistic performs best –Single peak performs well when variance low –Chebyshev worst

Results: Runtime Full probabilistic Means only PessimisticSingle peak Chebyshev consumable35s2s 30s25s consumable 2x std dev 35s2s 35s25s non- consumable 3s 400s5s600s Mean run times on the abstract domain problems Runs were terminated after 2000 iterations

Results: Orbiter Domain –Full probabilistic performs best –Single peak performs well –Chebyshev worst

Results: Problem Size –Means only worse on average for larger problems

Results: Problem Difficulty

Results: User Risk Metric –Means only worse on average for low risk tolerance

Result: Time Behavior

Conclusions Alternative methods for handling uncertain continuous variables Full probabilistic reasoning is most robust –superior plans fewer errors tailored to user risk attitudes –but requires modeling overhead –and computationally expensive –suited for high risk-averseness / cost of failure Naive approximations do almost as well

Future Directions Direct temporal constraints Domain-specific pessimistic approximation Need to also evaluate –bounded distributions –particle filter representation Integration with execution system –observations update distributions –dynamic replanning

Overcoming Normal Representation Inaccuracies Normals give probability to values from -∞ to ∞ Variable domain inaccuracy –duration must be greater than zero –usage is either > 0 or < 0 (if replenishing) –more problematic with small means and high variance Timeline domain inaccuracy –Resources often have only one bound of conflict (e.g. can’t have an overfull battery) –Becomes a problem for mixture of consumers and replenishers μ0

Solution to variable domain inaccuracy –redistribute impossible value probability into normal Timeline domain inaccuracy –move impossible value probability into a spike with same integral Overcoming Normal Representation Inaccuracies μ0 μ0 μ0 μ0

Particle Filter Representation Commonly used for robot localization A Monte Carlo simulation draws sample values (particles) from source random variables to derive likelihoods of alternative states In a planner, the particles approximate state/resource projections; the more particles, the more precise the estimate Gets around exponential peak computations of normal representation by trading precision and time

Handling Temporal Constraints A good execution system can issue a command when a preceding activity finishes. When activities are given temporal constraints (e.g. back-to-back), there should be no probability of overlap. To handle this, a Bayes net can be constructed based on temporal constraints to calculate the resource distribution resulting from different possible usage contribution combinations. Prob. Non.