New Bulgarian University > | Center for Cognitive Science > | Preparatory Program > | Course Description |
COG504 Planning
Advanced Topics in AI, Berlin: Springer Verlag (1988).
Wilkins, D.,
Practical Planning: Extending the Classical AI Planning Paradigm, San Mateo, CA: Morgan Kaufmann (1988).
Allen, J., Kautz, H., Pelavin, R., Tenenberg, J.,
Reasoning about Plans, Los Altos, CA: Morgan Kaufmann (1991).
Introduction to Planning
Topic 1: The task for planning of actions. Planning as a dynamic component of
the rational behaviour. Taxonomy of actions - mental, verbal, motor actions, and time
changes; active and reactive agents. AI planning problems: representation, inference,
control. Frame problem, qualifica-tion problem, prediction, persistence. Complexity.
Topic 2: Classical AI Planning Problem. Basic concepts of the state-space ap-proach
to planning: goal state, initial state, state changes and action se-quencing.
Precondition-effects model of actions. Planning as search. Means-ends analisys. GPS.
Methods for Planning of Change
Topic 3:Linear planning. Deductive planning. Green algorithm.Situation
calcu-lus. Goal regression, progression of knowledge and digression methods.
Non-monotonicity of knowledge and its representations. Frame axioms and complexity of the
changes. STRIPS approach to coping with non-monotonicity.
Topic 4: Hierarchical Planning. Abstraction in planning. Sequencing and
macrooperators. Abstraction levels and abstraction barriers. Control over granularity.
AbSTRIPS, TWEAK.
Topic 5: Non-linear planning. Goal interaction and action dependencies. Resolving
of conflicts. Least commitment principle. Labelling and propagation of markers. NONLIN,
NOAH, and SIPE.
Lab work:Experiments with the computer simulation: AbSTRIPS-like planner.
Topic 6: Planning with constraints. Linear, nonlinear and hierarchical
con-straints. Constraint satisfaction and constraint propagation. Stepwise refinement.
Topic 7: Meta-planning. Hierarchy of actions, hierarchy of changes and hierar-chy
of plans. Meta-planning and reflection. MOLGEN.
Seminar:Discussing of the classical dynamic planning algorithms.
Topic 8: Logical Models of Dynamics. Situation calculus vs. Dynamic logics vs.
Hoare logics. Complexity of inference. Classroom Exercises: Formal evaluation of the
complexity of planning tasks in differ-ent specificational languages.
Methods for Planning in Time.
Topic 9:Planning within Time Limits. Time indexation scheme and time win-dowing.
DEVISOR. Planning with satisfaction of temporal constraints. Dean's approach.
Topic 10: Logical Models of Time.Point-based, interval-based, and event-based
models. Linear and branching time. Complexity of reasoning about time.
Classroom exercises: Formal evaluation of the complexity of the temporal planning
task in different specificational languages.
Topic 11: Temporal Planning with Time Prediction. Changes in the future, in the
next moment and continuing changes. Temporal necessity and temporal possibility.
McDermott's planner.
Topic 12: Interval-Based Temporal Planning. Events and processes. Duration of
actions. Continuing possibility and continuing necessity. Allen's plan-ner.
Topic 13: Event-Based Temporal Planning. Events and situations. History of events
and event triggering. Cooperative events and parallel actions. Lansky's planner.
Seminar:Discussing of the temporal methods for planning.Methods for Non-Classical
Planning.
Topic 14: Distributed Planning. Action blackboard and plan agenda. Independent and
dependent knowledge sources. Control over the planning process. Critics and plan
corrections.
Topic 15: Reactive planning. Pros and cons of deterministic planning. Replanning
during execution. Prediction vs. experimentation in planning. Genetic, evolutionary and
chaotic planning.
Lab work: Experimental evaluation of the planning strategies using planning
workbench (Truckworld, Tileworld, Mice).