Dynamic Domain Architectures for Model Based Autonomy MoBIES Embedded Software Working Group Meeting April 10-11 2001 B. Williams, B. Laddaga, H. Shrobe,

Slides:



Advertisements
Similar presentations
Dialogue Policy Optimisation
Advertisements

MBD and CSP Meir Kalech Partially based on slides of Jia You and Brian Williams.
Autonomic Systems Justin Moles, Winter 2006 Enabling autonomic behavior in systems software with hot swapping Paper by: J. Appavoo, et al. Presentation.
Timed Automata.
MBD in real-world system… Self-Configuring Systems Meir Kalech Partially based on slides of Brian Williams.
Approved for Public Release, Distribution Unlimited Pervasive Self-Regeneration through Concurrent Model-Based Execution Brian Williams (PI) Paul Robertson.
ECE 720T5 Fall 2012 Cyber-Physical Systems Rodolfo Pellizzoni.
SSP Re-hosting System Development: CLBM Overview and Module Recognition SSP Team Department of ECE Stevens Institute of Technology Presented by Hongbing.
Integrating POMDP and RL for a Two Layer Simulated Robot Architecture Presented by Alp Sardağ.
Safeguarding the Three Corner Sat Constellation By Stephen Levin-Stankevich Stephen Nauman.
The Rare Glitch Project: Verification Tools for Embedded Systems Carnegie Mellon University Pittsburgh, PA Ed Clarke, David Garlan, Bruce Krogh, Reid Simmons,
Simulating A Satellite CSGC Mission Operations Team Cameron HatcherJames Burkert Brandon BobianAleks Jarosz.
Sheila McIlraith, Knowledge Systems Lab, Stanford University DX’00, 06/2000 Diagnosing Hybrid Systems: A Bayesian Model Selection Approach Sheila McIlraith.
System Integration Management (SIM)
Automation for System Safety Analysis: Executive Briefing Jane T. Malin, Principal Investigator Project: Automated Tool and Method for System Safety Analysis.
Model-based Programming of Fault Aware Systems Brian C. Williams CSAIL, MIT.
1 EVALUATING INTELLIGENT FLUID AUTOMATION SYSTEMS USING A FLUID NETWORK SIMULATION ENVIRONMENT Ron Esmao - Sr. Applications Engineer, Flowmaster USA.
Soft Computing Lecture 20 Review of HIS Combined Numerical and Linguistic Knowledge Representation and Its Application to Medical Diagnosis.
ECE 720T5 Winter 2014 Cyber-Physical Systems Rodolfo Pellizzoni.
A Hierarchical Approach to Model-based Reactive Planning in Large State Spaces Artificial Intelligence & Space Systems Laboratories Massachusetts Institute.
An Introduction to Software Architecture
20a - 1 NASA’s Goddard Space Flight Center Attitude Control System (ACS) Eric Holmes, Code 591 Joe Garrick, Code 595 Jim Simpson, Code 596 NASA/GSFC August.
Command and Data Handling (C&DH)
Streaming Predictions of User Behavior in Real- Time Ethan DereszynskiEthan Dereszynski (Webtrends) Eric ButlerEric Butler (Cedexis) OSCON 2014.
SOFTWARE DESIGN (SWD) Instructor: Dr. Hany H. Ammar
Plant Modeling for Powertrain Control Design Modelica Automotive Workshop Dearborn, MI November 19, 2002 Dr. Larry Michaels GM Powertrain Controls Engineering.
.1 RESEARCH & TECHNOLOGY DEVELOPMENT CENTER SYSTEM AND INFORMATION SCIENCES JHU/MIT Proprietary Titan MESSENGER Autonomy Experiment.
Model-Based Diagnosis of Hybrid Systems Papers by: Sriram Narasimhan and Gautam Biswas Presented by: John Ramirez.
Probabilistic Reasoning for Robust Plan Execution Steve Schaffer, Brad Clement, Steve Chien Artificial Intelligence.
MIT Dept of Aeronautics and Astronautics March 21, 2003 Graduate Open House Aero/Astro Open House MERS Research Group Model-based Embedded and Robotic.
1 Jillian Redfern Orbital Express Presentation TITAN All-Hands 07/08/2003.
1 Context-dependent Product Line Practice for Constructing Reliable Embedded Systems Naoyasu UbayashiKyushu University, Japan Shin NakajimaNational Institute.
Aero/Astro Open House MERS Research Group Model-based Embedded and Robotic Systems Group Space Systems Laboratory Massachusetts Institute of Technology.
DSL Distributed Systems Laboratory ATC 23 August Model Mission: Magnetospheric Multiscale (MMS) Mission Goal “To study the microphysics of three.
1 Distributed and Optimal Motion Planning for Multiple Mobile Robots Yi Guo and Lynne Parker Center for Engineering Science Advanced Research Computer.
10/16/02copyright Brian Williams, courtesy of JPL Diagnosing Multiple Faults Brian C. Williams J/6.834J October 16 th, 2002 Brian C. Williams,
University of Windsor School of Computer Science Topics in Artificial Intelligence Fall 2008 Sept 11, 2008.
MAPLD 2005/254C. Papachristou 1 Reconfigurable and Evolvable Hardware Fabric Chris Papachristou, Frank Wolff Robert Ewing Electrical Engineering & Computer.
Fall 2004EE 3563 Digital Systems Design EE 3563 VHSIC Hardware Description Language  Required Reading: –These Slides –VHDL Tutorial  Very High Speed.
Network UAV C3 Stage 1 Final Briefing Timothy X Brown University of Colorado at Boulder Interdisciplinary Telecommunications Program Electrical and Computer.
31 March 2009 MMI OntDev 1 Autonomous Mission Operations for Sensor Webs Al Underbrink, Sentar, Inc.
Massachusetts Institute of Technology September 7, 2005 A Tractable Approach to Probabilistically Accurate Mode Estimation Oliver B. Martin Seung H. Chung.
Model-based Programming of Cooperative Explorers Prof. Brian C. Williams Dept. of Aeronautics and Astronautics Artificial Intelligence Labs And Space Systems.
Aquarius Mission Simulation A realistic simulation is essential for mission readiness preparations This requires the ability to produce realistic data,
Final Version Kequan Luu May 13-17, 2002 Micro-Arcsecond Imaging Mission, Pathfinder (MAXIM-PF) Flight Software.
GLAST LAT Project LAT System Engineering 1 GLAST Large Area Telescope: LAT System Engineering Pat Hascall SLAC System Engineering Manager
Discovery and Systems Health Technical Area NASA Ames Research Center - Computational Sciences Division Automated Diagnosis Sriram Narasimhan University.
GLAST Large Area Telescope LAT Flight Software System Checkout TRR Test Suites (Backup) Stanford Linear Accelerator Center Gamma-ray Large Area Space Telescope.
Robotic Space Explorers: To Boldly Go Where No AI System Has Gone Before A Story of Survival J/6.834J September 19, 2001.
Outline Deep Space One and Remote Agent Model-based Execution OpSat and the ITMS Model-based Reactive Planning Space Robotics.
Space Systems Laboratory Massachusetts Institute of Technology AUTONOMY MIT Graduate Student Open House March 24, 2000.
SAS_05_Contingency_Lutz_Tal1 Contingency Software in Autonomous Systems Robyn Lutz, JPL/Caltech & ISU Doron Tal, USRA at NASA Ames Ann Patterson-Hine,
Copyright B. Williams J/6.834J, Fall 02 Lecture 3: Immobile Robots and Space Explorers Prof. Brian Williams Rm Wednesday, September 11 th,
SwCDR (Peer) Review 1 UCB MAVEN Particles and Fields Flight Software Critical Design Review Peter R. Harvey.
Control-Theoretic Approaches for Dynamic Information Assurance George Vachtsevanos Georgia Tech Working Meeting U. C. Berkeley February 5, 2003.
Autonomy: Executive and Instruments Life in the Atacama 2004 Science & Technology Workshop Nicola Muscettola NASA Ames Reid Simmons Carnegie Mellon.
Monitoring Dynamical Systems: Combining Hidden Markov Models and Logic
Chapter 11: Artificial Intelligence
Reading B. Williams and P. Nayak, “A Reactive Planner for a Model-based Executive,” International Joint Conference on Artificial Intelligence, 1997.
GLAST Large Area Telescope:
Model-based Diagnosis: The Single Fault Case
CS b659: Intelligent Robotics
J. Michael, M. Shing M. Miklaski, J. Babbitt Naval Postgraduate School
Vesa Klumpp, Knowtion Applications of Intelligent Control in Industry and Adaption to Space Missions Vesa Klumpp, Knowtion
Robust Belief-based Execution of Manipulation Programs
Autonomous Operations in Space
CubeSat vs. Science Instrument Complexity
MIT AI Lab: B. Williams, H. Shrobe, R. Laddaga
Presentation transcript:

Dynamic Domain Architectures for Model Based Autonomy MoBIES Embedded Software Working Group Meeting April B. Williams, B. Laddaga, H. Shrobe, M. Hofbaur MIT Artificial Intelligence Lab Space Systems Lab

Mode Estimation and Fault Diagnosis Karl Hedrick in Automotive Powertrain OEP Overview: “Powertrain control system software for automotive application is becoming more and more complex due to increasing functionality as well as ever more wide-ranging onboard diagnostic systems mandated by state and federal governments.” Challenge Problem: Model-based synthesis of an onboard monitoring and diagnostic engine that determines the (fault) mode of the powertrain system based on sensor data and the powertrain model. Model-based: Mode estimates (diagnoses) generated from component-based models. Handles multiple faults and multiple symptoms. Focuses on most likely modes (faults). Handles un-modeled failures / situations.

Fault Scenarios: Intake System Throttle/Intake Faults: (partially) clogged stuck-at intake leak … EGR Valve Faults stuck-at valve leak Intake Manifold Pressure Sensor binary faults: stuck-at soft faults: drifting, noise … VDL powertrain model

Diagnostic Approaches: 1.Livingstone (Mode Estimation and Repair) Models = probabilistic automata + discrete constraints. Performs extensive deduction online. 2.MiniMe (Miniature Mode Estimation) –Component models = discrete constraints –Precompiles diagnoses of individual symptoms. –Combines diagnoses on line. 3.HybridME (Hybrid Mode Estimation) –Models = probabilistic hybrid automata –detects degradation and incipient faults

Experiments for FY01 Milestones Implement MiniME and HybridME v1 modules. Validate Livingstone and MiniMe based on the simulated Intake Subsystem of the Powertrain OEP: –Intake/Throttle faults –Pressure and Airflow sensor faults Extend HybridME to concurrent automata. Validate HybridME based on the simulated Intake Subsystem of the Powertrain OEP model Validate intake subsystem diagnosis experiments on real-world data. FY02 Program Milestones

Success Criteria: Detection of incipient failures Detection of novel failures Real-time performance Use of both discrete and continuous models Demonstration on simulated data. Demonstration on real world data

Open Issues Need: sensor / actuator models Realistic fault scenarios and fault models Realistic sensor placement for power train Model of linkage from actuator to engine (e.g, by wire?) Model for EGR control Faults that may be injected within the real world OEP powertrain (e.g., valve and intake leaks)?

Temporal planner Livingstone Command goals Observations Flight System Control RT Control Layer State Thrust Goals Attitude Point(a) Engine Off Delta_V(direction=b, magnitude=200) Power Livingstone: Model-based Mode Estimation and Reconfiguration

Model Temporal planner Livingstone Command goals Observations Flight System Control RT Control Layer State Thrust Goals Attitude Point(a) Engine Off Delta_V(direction=b, magnitude=200) Power Closed Valve Open Stuckopen Stuckclosed OpenClose inflow = outflow = 0

Model Temporal planner Livingstone Command goals Observations Flight System Control RT Control Layer State Thrust Goals Attitude Point(a) Engine Off Delta_V(direction=b, magnitude=200) Power mode = open  high pressure = high flow; nominal pressure = nominal flow;... mode = closed  Zero flow

Model Temporal planner Livingstone Command goals Observations Flight System Control RT Control Layer State Thrust Goals Attitude Point(a) Engine Off Delta_V(direction=b, magnitude=200) Power State estimates Mode Reconfiguration Mode Estimation Configuration goals

Model Temporal planner Livingstone Command goals Observations Flight System Control RT Control Layer State Thrust Goals Attitude Point(a) Engine Off Delta_V(direction=b, magnitude=200) Power Reactive Model-based Programming (RMPL) Fire backup engine Valve fails stuck closed Open four valves

Model based Diagnosis (GDE) –Conflict Recognition A symptom is a discrepancy between the model and observations. A conflict is a minimal set of component modes that is inconsistent with the observations. Conflict Recognition produces ALL conflicts. –Candidate Generation A diagnosis is an assignment of component modes that resolves all conflicts. A kernel diagnosis is a minimal set of component modes that resolves all conflicts. Candidate generation produces the likely kernel diagnoses

Miniature Mode Estimation (MiniME) Compile time: - Given component models and interconnections. - Generates rules mapping observations to conflicts. Run time: - Rules select conflicts based on current observations. - Generate most-likely kernel diagnoses from conflicts. Idea: Generate Conflicts offline:

MiniME – Compile Time Model components specified in RMPL modeling language. –Initially will use Livingstone language to exploit heritage. Compiler receives models and outputs Concurrent Constraint Automata to the Rule Generator. Rule Generator compiles CCA to rules (called dissents) that map sensor values to conflicts, using an instance of a prime implicate algorithm. –Prime Implicates Relay everything that must be true about the model without being redundant. –Dissent Special case of a prime implicate that relates sensor values to component modes.

MiniME – Run Time Uses current observations and full set of dissent rules to extract the conflicts that currently hold. Candidate Generation is viewed as tree search: The branches at a level select component modes from some conflict. A kernel diagnosis is a tree path that selects a mode from all conflicts. The cost of a branch is the mode’s probability. The most likely diagnosis corresponds to the shortest path in the tree.

HybridME Compiler A Hybrid Probabilistic Automata (HPA) encodes a hidden Markov model. Modes exhibit a continuous dynamical behavior, expressed by differential or difference equations. Hybrid Probabilistic Automata model M … set of discrete modes x, u, y … continuous state, input, output F … continuous dynamics for each mode (ODE) T … probabilistic transitions between modes Probabilistic transitions: Go to a nominal or failure mode Are conditioned on continuous variables Compiler maps RMPL Model to concurrent Hybrid Probabilistic Automata.

HybridME Mode Estimator Hybrid State Estimator maintains a set of likely hybrid state estimates as a set of trajectories. Hybrid mode estimator determines possible mode transitions for each trajectory. selects candidate trajectories to be tracked by the continuous state estimators. HMM belief state update is used to determine the likelihood for each traced trajectory

Techsat-21 Mission 3 spacecraft reconfigurable constellation High relative positioning accuracy (3mm-1cm) 28.5 inclination orbit (+/- 20 deg lat) 90 minute orbit, 1-day ~repeat track Each spacecraft has an X-band radar Est. 1m resolution (range and x-range) Non-repeat pass interferometry possible due to simultaneous emission and receipt of all 3 s/c using orthogonal waveforms –Not used for ASC demonstration due to computational cost of onboard processing Backscatter of X-band wavelength can easily distinguish water, ice, land (fresh v. old lava)

Autonomous Science Constellation Mission Concept 1. Target Imaged 5. New Science Images taken (repeat track) 2. Science Event Detected 3. Re-Plan 4. Constellation Reconfigured

Key Technologies Cluster Management (PSS/AFRL) –Enables single interface to cluster Robust Execution (ICS/AFRL) –Enables plans to deal wit run-time uncertainties Model-based Mode Identification and Reconfiguration (MIT SSL) –Enables timely state estimation and low level control Onboard Science (JPL) Feature detection & pattern recognition Change detection Enables onboard science decision-making Onboard Replanning (JPL) –Enables onboard development of new plans to perform observations

Technology Impact Enable vast improvements in science for fixed downlink –Downlink only highest science value images Enable observation of short-lived science events Reduce downtime due to anomalies Reduce setup time via exploitation of execution feedback

ASC Demonstration Software Message Detail Diagram CASPER Solaris s/w bus OSE s/w bus SCL SIM OA 5. Cancelled goals to SCL (ACTIVITY_DELETE) 8. Activity commits to SCL (ACTIVITY_EXECUTE) 8.1 Request to OA (COMPUTE_FORMATION) 10. Time to all (TIME) 4. Goals from SCL (ACTIVITY_CREATE) 8.1 Response from OA (FORMATION) 8.4 Goals from SCL (UPDATE_GOALS) 9. r/s/a updates from SCL (RESOURCE_UPDATE, STATE_UPDATE, ACTIVITY_UPDATE) Bus mirroring provided by ICS 9. r/s/a updates to SCL 8.3 Commands from SCL (SAR_ON, CALIBRATE, TAKE_DATA) 2. Goals from ground (UPDATE_GOALS) 6. Cancelled goals from CASPER (ACTIVITY_DELETE) 8. Commits from CASPER (ACTIVITY_EXECUTE) 8.1 Maneuver commands from OA (TBD) 8.1 Response from OA (FORMATION_COMPLETE) 8.2 Radar point commands from OA (TBD) 8.2 Response from OA (RADAR_POINTED) 8.4 Response from Science (IMAGE_FORMED, IMAGE_PROCESSED) 8.4 Goals from Science (IMAGE, DOWNLINK, CACHE_IMAGE) 9. r/s/a updates from SIM SolarisOSE r/s/a = resource, state, activity 8.1 Responses to CASPER (FORMATION) 8.1 Responses to SCL (FORMATION_COMPLETE) 8.1 Maneuver commands to SCL (TBD) 8.2 Response to SCL (RADAR_POINTED) 8.2 Radar point commands to SCL (TBD) 8.1 Request from CASPER (COMPUTE_FORMATION) 8.1 Command from SCL (CHANGE_FORMATION) 8.2 Command from SCL (POINT_RADAR) Science 8.4 Response to SCL (IMAGE_FORMED, IMAGE_PROCESSED) 8.4. New goals to SCL (IMAGE, DOWNLINK, CACHE_IMAGE) 8.4. Commands from SCL (FORM_IMAGE, PROCESS_IMAGE) 3. Goals to CASPER (ACTIVITY_CREATE) 8.1 Command to OA (CHANGE_FORMATION) 8.1 Maneuver commands to Broadreach (TBD) 8.2 Command to OA (POINT_RADAR) 8.2 Radar point commands to Broadreach (TBD) 8.3 Commands to SIM (SAR_ON, CALIBRATE, TAKE_DATA, METROLOGY) 8.4 Commands to Science (TRANSFER_SAR_DATA, FORM_IMAGE, PROCESS_IMAGE) 8.4 Goals to CASPER (UPDATE_GOALS) 9. r/s/a updates to CASPER (RESOURCE_UPDATE, STATE_UPDATE, ACTIVITY_UPDATE) Dynamics SIM Burton

Validation Plan CASPER Planner SAR Image Formation Onboard Science SCL Execution Autonomy ObjectAgent TeamAgent Cluster Mgmt ASC Experiment Study Phase Refinement Phase Techsat-21 Flight Integration on Workstations System Testing on relevant data Unit Testing on relevant data ASC Transition to AFRL Testbed ASC Transition to Flight Hardware ASC Flight Demonstration Gates System Demonstration on AFRL Techsat-21 Testbed System Demonstration on Techsat-21 Flight Hardware ASC System Demonstration on Workstations Broadreach Low-level FSW Sep 2000 March 2002 September 2003 September April 2000 Burton FDIR