Practical HTN Planning

1
Practical HTN Planning

• Putting HTN Planning
• into Use

2
Literature

• Human Planning
• Klein, G. (1998) Sources of Power How People
  Make Decisions, MIT Press.
• Refinement Search
• Kambhampati, S., Knoblock, C.A. and Yang, Q.
  (1995) Planning as Refinement Search A Unified
  Framework for Evaluating Design Tradeoffs in
  Partial-Order Planning, Artificial Intelligence,
  Vol. 76, No. 1-2, pp. 167-238, Elsevier.
• Nonlin
• http//www.aiai.ed.ac.uk/project/nonlin/
• Tate, A. (1977) Generating Project Networks,
  Proceedings of the Fifth International Joint
  Conference on Artificial Intelligence (IJCAI-77)
  pp. 888-893, Boston, Mass. USA, August 1977.
• O-Plan
• http//www.aiai.ed.ac.uk/project/oplan/
• Currie, K and Tate, A. (1991) O-Plan the Open
  Planning Architecture, Artificial Intelligence
  Vol. 52, No. 1, pp 49-86, Elsevier.
• Other Practical Planners
• Ghallab, M., Nau, D. and Traverso, P., Automated
  Planning Theory and Practice, chapters 19, 22
  and 23. Elsevier/Morgan Kaufmann, 2004.

3
Overview

• Human Approaches to Planning
• Practical HTN Planning
• Refinement Planning as a Unifying View
• Nonlin and O-Plan Features
• QA (Modal Truth Criterion)
• Time, Resource and Other Constraint Handling
• I-X/I-Plan Overview

4
Some Planning Features

• Expansion of a high level abstract plan into
  greater detail where necessary.
• High level chunks of procedural knowledge
  (Standard Operating Procedures, Best Practice
  Processes, Tactics Techniques and Procedures,
  etc.) at a human scale - typically 5-8 actions -
  can be manipulated within the system.
• Ability to establish that a feasible plan exists,
  perhaps for a range of assumptions about the
  situation, while retaining a high level overview.
• Analysis of potential interactions as plans are
  expanded or developed.
• Identification of problems, flaws and issues with
  the plan.
• Deliberative establishment of a space of
  alternative options, perhaps based on different
  assumptions about the situation involved, of
  especial use ahead of time, in training and
  rehearsal, and to those unfamiliar with the
  situation or utilising novel equipment.

5
More Planning Features

• Monitoring of the execution of events as they are
  expected to happen within the plan, watching for
  deviations that indicate a necessity to re-plan
  (often ahead of this becoming a serious problem).
• Represent the dynamic state of the world at
  points in the plan and use this for mental
  simulation of the execution of the plan.
• Pruning of choices according to given
  requirements or constraints.
• Situation dependent option filtering (sometime
  reducing the choices normally open to one
  obvious one.
• Satisficing search to find the first suitable
  plan that meets the essential criteria.
• Heuristic evaluation and prioritisation of
  multiple possible choices within the constrained
  search space.
• Uniform use of a common plan representation with
  embedded rationale to improve plan quality,
  shared understanding, etc.

6
Human Approach

• Previous slides describe aspects of problem
  solving behaviour observed in expert humans
  working in unusual or crisis situations.
• Gary Klein, Sources of Power, MIT Press, 1998.
• But they also describe the hierarchical and mixed
  initiative approach to planning in AI developed
  over the last 30 years.

7
HTN - Planning Approach

• HTN Planning is a useful paradigm
• Compose workflows/processes from requirements and
  component/template libraries
• Covers simple through to very complex
  (pre-planned) components
• Allows for execution support, reactive repair,
  recovery, etc.
• Suited to mixed initiative (people and systems)
  planning and execution
• Gives an understandable framework within which
  specialised constraint solvers, domain-specific
  planners (e.g. route finders), optimisers, plan
  analysers and simulators can work

8
HTN - Activity Composition
Introduce activities to achieve
preconditions Resolve interactions between
conditions and effects Handle constraints (e.g.
world state, resource, spatial, etc.)
9
HTN Initial Plan as Goals
Initial Plan can be any combination of Activities
and Constraints
10
Nonlin (1974-1977)

• Hierarchical Task Network Planning
• Partial Order Planner
• Plan Space Planner
• Goal structure-based plan development - considers
  alternative approaches only based on plan
  rationale
• QA/Modal Truth Criterion Condition Achievement
• Condition Types to limit search
• Compute Conditions for links to external data
  and systems (attached procedures)
• Time and Resource Constraint checks
• Nonlin core is basis for text book descriptions
  of HTN Planning

11
Nonlin Domain Language TF
x is a variable
12
QA/Modal Truth Criterion

• QA in a partially ordered network of nodes
• Way to establish value of a condition PV at some
  point in the plan
• Yes/no/maybe responses
• Alternative Terminology
• Contributors, deletors (Austin Tate, Nonlin, QA,
  Edinburgh, 1975-7)
• White nights and clobberers (David Chapman, MIT,
  MTC, 1987, 1st Formalisation)
• Producers, consumers (Some textbooks)
• Initially just allowed imposition of orderings on
  nodes for a condition, a ? b (ordering)
• Later also allowed variables within condition to
  be constrained (codesignation), ?
  (non-codesignation)
• Intuitively, a white knight is an activity which
  re-establishes a clobbered precondition p
• A clobberer in a plan can be "defeated" by
  imposing ordering or codesignation/non-codesignati
  on constraints on the plan, or by inserting a
  white knight between the clobberer and the point
  where a condition is needed

13
QA/Modal Truth Criterion
Need to ensure no deletor appears between a
chosen contributor and point of need
14
O-Plan (1983-1999) Features

• Hierarchical Task Network Planning
• Nonlin-like goal-structure, QA and Typed/Compute
  conditions
• Partial-Plan Refinement Approach
• Plan State has flaws/issues attached
• Agenda Architecture with Plan Modification
  Operations
• Opportunistic Search (agenda type,
  branch1/branch N)
• Multiple constraint managers with yes/no and
  maybe results
• Least Commitment Approach (on activity ordering,
  object/variable bindings and other constraints)
• Constraint Posting rather than explicit
  commitments (and/or trees with sets of before
  temporal constraints and variable binding ( and
  ?) constraints) as in MOLGEN
• Goal structure recording and monitoring to
  preserve plan rationale

15
O-Plan (1983-1999) Features
16
O-Plan Domain Language TF
?x is a variable
17
O-Plan Agent Architecture
18
O-Plan Agent Architecture
19
O-Plan Planning Workflow
20
A More CollaborativePlanning Framework

• Human relatable and presentable objectives,
  issues, sense-making, advice, multiple options,
  argumentation, discussions and outline plans for
  higher levels
• Detailed planners, search engines, constraint
  solvers, analyzers and simulators act in this
  framework in an understandable way to provide
  feasibility checks, detailed constraints and
  guidance
• Sharing of processes and information about
  process products between humans and systems
• Current status, context and environment
  sensitivity
• Links between informal/unstructured planning,
  more structured planning and methods for
  optimisation

21
I-X/I-Plan (2000- )

• Shared, intelligible, easily communicated and
  extendible conceptual model for objectives,
  processes, standard operating procedures and
  plans
• I Issues
• N Nodes/Activities
• C Constraints
• A Annotations
• Communication of dynamic status and presence for
  agents, and reports about their collaborative
  processes and process products
• Context sensitive presentation of options for
  action
• Intelligent activity planning, execution,
  monitoring, re-planning and plan repair via
  I-Plan and I-P2 (I-X Process Panels)

22
ltI-N-C-Agt Framework
Plan State
Nodes
A Annotations
23
ltI-N-C-Agt I-X
Plan State
Nodes
A Annotations
24
Anatomy of an I-X Process Panel
25
I-P2 aim is a Planning, Workflow and Task
Messaging Catch All

• Can take ANY requirement to
• Handle an issue
• Perform an activity
• Respect a constraint
• Note an annotation
• Deals with these via
• Manual activity
• Internal capabilities
• External capabilities
• Reroute or delegate to other panels or agents
• Plan and execute a composite of these
  capabilities (I-Plan)
• Receives reports and interprets them to
• Understand current status of issues, activities
  and constraints
• Understand current world state, especially status
  of process products
• Help user control the situation
• Copes with partial knowledge of processes and
  organisations

26
I-X Process Panel and Tools
Process Panel
27
I-X for Emergency Response
28
Planning Research Areas Techniques

• Domain Modelling HTN, SIPE
• Domain Description PDDL, NIST PSL
• Domain Analysis TIMS

• Plan Generalisation Macrops, EBL
• Case-Based Planning CHEF, PRODIGY
• Plan Learning SOAR, PRODIGY

• Search Methods Heuristics, A
• Graph Planning Algthms GraphPlan
• Partial-Order Planning Nonlin, UCPOP
• Hierarchical Planning NOAH, Nonlin, O-Plan
• Refinement Planning Kambhampati
• Opportunistic Search OPM
• Constraint Satisfaction CSP, OR, TMMS
• Optimisation Methods NN, GA, Ant Colony Opt.
• Issue/Flaw Handling O-Plan

• User Interfaces SIPE, O-Plan
• Plan Advice SRI/Myers
• Mixed-Initiative PlanS TRIPS/TRAINS

Problem is to make sense of all these techniques

• Plan Generalisation Macrops, EBL
• Case-Based Planning CHEF, PRODIGY
• Plan Learning SOAR, PRODIGY

• Planning Web Services O-Plan, SHOP2

• Plan Analysis NOAH, Critics
• Plan Simulation QinetiQ
• Plan Qualitative Mdling Excalibur

• Plan Sharing Comms I-X, ltI-N-C-Agt
• NL Generation
• Dialogue Management

• Plan Repair O-Plan
• Re-planning O-Plan
• Plan Monitoring O-Plan, IPEM

Deals with whole life cycle of plans
29
Summary

• Human Approaches to Planning
• Practical HTN Planning
• Refinement Planning as a Unifying View
• Nonlin and O-Plan Features
• QA (Modal Truth Criterion)
• Time, Resource and Other Constraint Handling
• I-X/I-Plan Overview