Lecturer: Austin Tate Date Prepared: 6-Nov-2009 1 Overview Deep - PDF document

AI Planner Applications Practical Applications of AI Planners Lecturer: Austin Tate Date Prepared: 6-Nov-2009 1

Overview � Deep Space 1 � Other Practical Applications of AI Planners � Common Themes AI Planner Applications 2 2

Literature Deep Space 1 Papers � Ghallab, M., Nau, D. and Traverso, P., Automated Planning – Theory and � Practice , chapter 19,. Elsevier/Morgan Kaufmann, 2004. Bernard, D.E., Dorais, G.A., Fry, C., Gamble Jr., E.B., Kanfesky, B., Kurien, J., � Millar, W., Muscettola, N., Nayak, P.P., Pell, B., Rajan, K., Rouquette, N., Smith, B., and Williams, B.C. Design of the Remote Agent experiment for spacecraft autonomy . Procs. of the IEEEAerospace Conf., Snowmass, CO, 1998. � http://nmp.jpl.nasa.gov/ds1/papers.html Other Practical Planners � Ghallab, M., Nau, D. and Traverso, P., Automated Planning – Theory and � Practice , chapter 22 and 23. Elsevier/Morgan Kaufmann, 2004 Tate, A. and Dalton, J. (2003) O-Plan: a Common Lisp Planning Web Service, � invited paper, in Proceedings of the International Lisp Conference 2003, October 12-25, 2003, New York, NY, USA, October 12-15, 2003. http://www.aiai.ed.ac.uk/project/ix/documents/2003/2003-luc-tate-oplan-web.doc � AI Planner Applications 3 Literature Deep Space 1 is described in Chapter 19 in the course textbook: Malik Ghallab, Dana Nau, and Paolo Traverso. Automated Planning – Theory and Practice . Elsevier/Morgan Kaufmann, 2004 . Further practical planning systems are described in several chapters of the course textbook… e.g. chapters 22 and 23. Malik Ghallab, Dana Nau, and Paolo Traverso. Automated Planning – Theory and Practice . Elsevier/Morgan Kaufmann, 2004 . 3

Deep Space 1 – 1998-2001 http://nmp.jpl.nasa.gov/ds1/ AI Planner Applications 4 NASA’s Deep Space 1 (DS1) http://nmp.jpl.nasa.gov/ds1/ Deep Space 1 launched from Cape Canaveral on October 24, 1998. During a highly successful primary mission, it tested 12 advanced, high-risk technologies in space. In an extremely successful extended mission, it encountered Comet Borrelly and returned the best images and other science data ever from a comet. During its fully successful hyperextended mission, it conducted further technology tests. The spacecraft was retired on December 18, 2001. IPS = Ion Propulsion System (one of the advanced technologies being demonstrated on DS1) 4

DS 1 – Comet Borrelly http://nmp.jpl.nasa.gov/ds1/ AI Planner Applications 5 NASA’s Deep Space 1 (DS1) at Comet Borrelly http://nmp.jpl.nasa.gov/ds1/ 5

DS1 Domain Requirements Achieve diverse goals on real spacecraft � High Reliability • single point failures • multiple sequential failures � Tight resource constraints • resource contention • conflicting goals � Hard-time deadlines � Limited Observability � Concurrent Activity AI Planner Applications 6 Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf 6

DS1 Remote Agent Approach � Constraint-based planning and scheduling • supports goal achievement, resource constraints, deadlines, concurrency � Robust multi-threaded execution • supports reliability, concurrency, deadlines � Model-based fault diagnosis and reconfiguration • supports limited observability, reliability, concurrency � Real-time control and monitoring AI Planner Applications 7 Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf 7

DS1 Levels of Autonomy Listed from least to most autonomous mode: single low-level real-time command execution 1. time-stamped command sequence execution 2. single goal achievement with auto-recovery 3. model-based state estimation & error detection 4. scripted plan with dynamic task decomposition 5. on-board back-to-back plan generation, 6. execution, & plan recovery AI Planner Applications 8 Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf 8

DS 1 Levels of Autonomy Slides from DS1 Final Report http://nmp.jpl.nasa.gov/ds1/papers.html http://nmp-techval- reports.jpl.nasa.gov/DS1/Remote_Integrated_Report.pdf 9

DS 1 Systems Planning Execution Monitoring Slides from DS1 Final Report http://nmp.jpl.nasa.gov/ds1/papers.html http://nmp-techval- reports.jpl.nasa.gov/DS1/Remote_Integrated_Report.pdf 10

DS1 RAX Functionality PS/MM generate plans on-board the spacecraft � reject low-priority unachievable goals � � replan following a simulated failure enable modification of mission goals from ground � EXEC provide a low-level commanding interface � initiate on-board planning � � execute plans generated both on-board and on the ground recognize and respond to plan failure � � maintain required properties in the face of failures MIR confirm executive command execution � � demonstrate model-based failure detection, isolation, and recovery demonstrate ability to update on-board state via ground commands � AI Planner Applications 11 Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf PS = Planner Scheduler MM = Mission Manager EXEC = Executor MIR = Mode Identification and Recovery 11

DS1 Remote Agent (RA) Architecture Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf MIR = Mode Identification and Recovery Module (Livingstone) MI = Mode Identification MR = Mode Recovery 12

DS1 Planner Architecture Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf ACS = Attitude Control System NAV = Navigation IRS = Incremental Refinement Scheduler HSTS = Heuristic Scheduling Testbed System TDB = Temporal Data Base DLL = Domain Description Language 13

DS1 Diversity of Goals Final state goals � • “Turn off the camera once you are done using it” Scheduled goals � • “Communicate to Earth at pre-specified times” Periodic goals � • “Take asteroid pictures for navigation every 2 days for 2 hours” Information-seeking goals � • “Ask the on-board navigation system for the thrusting profile” Continuous accumulation goals � • “Accumulate thrust with a 90% duty cycle” Default goals � • “When you have nothing else to do, point HGA to Earth” AI Planner Applications 14 Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf HGA = High Gain Antenna 14

DS1 Diversity of Constraints State/action constraints � • “To take a picture, the camera must be on.” Finite resources � • power True parallelism � • the ACS loops must work in parallel with the IPS controller Functional dependencies � • “The duration of a turn depends on its source and destination.” Continuously varying parameters � • amount of accumulated thrust Other software modules as specialized planners � • on-board navigator AI Planner Applications 15 Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf ACS = Attitude Control System IPS = Ion Propulsion System 15

DS1 Domain Description Language Temporal Constraints in DDL Command to EXEC in ESL Slides from report by Gregory Dorain and David Kortenkamp, NASA. http://www.traclabs.com/~korten/tutorial/arc.pdf DDL = Domain Description Language ESL = EXEC Support Language 16

DS1 Plan Fragment From Bernard et al. 1998 17

DS1 RA Exec Status Tool From Bernard et al. 1998 18

DS1 RA Ground Tools From Bernard et al. 1998 Timeline Applet – available real time over the web during the RAX flight experiment. 19

DS1 – Flight Experiments 17 th – 21 st 1999 RAX was activated and controlled the spacecraft � autonomously. Some issues and alarms did arise: Divergence of model predicted values of state of Ion � Propulsion System (IPS) and observed values – due to infrequency of real monitor updates. EXEC deadlocked in use. Problem diagnosed and fix � designed by not uploaded to DS1 for fears of safety of flight systems. Condition had not appeared in thousands of ground tests � indicating needs for formal verification methods for this type of safety/mission critical software. Following other experiments, RAX was deemed to have � achieved its aims and objectives. AI Planner Applications 20 From Bernard et al. 1998 The flight experiments were conducted from May 17 th 1999 over a 2 day period. 20

DS 1 Experiment 2 Day Scenario Slides from DS1 Final Report http://nmp.jpl.nasa.gov/ds1/papers.html http://nmp-techval- reports.jpl.nasa.gov/DS1/Remote_Integrated_Report.pdf 21

DS 1 Summary Objectives and Capabilities Slides from DS1 Final Report http://nmp.jpl.nasa.gov/ds1/papers.html http://nmp-techval- reports.jpl.nasa.gov/DS1/Remote_Integrated_Report.pdf 22

Earlier Spacecraft Planning Applications Deviser � NASA Jet Propulsion Lab � Steven Vere, JPL � First NASA AI Planner � 1982-3 � Based on Tate’s Nonlin � Added Time Windows � Voyager Mission Plans � Not used live � AI Planner Applications 23 Steven A. Vere. Planning in time: Windows and duration for activities and goals . IEEE Transactions on Pattern Analysis and Machine Intelligence, 5(3):246--267, 1983. 23