An Introduction to Artificial Intelligence CE 40417

47
An Introduction to Artificial Intelligence CE 40417 Chapter 12 – Planning and Acting in Real World Ramin Halavati ([email protected]) In which we see how more expressive representations and more interactive agent architectures lead to planners that are useful in real world.

description

An Introduction to Artificial Intelligence CE 40417. Chapter 12 – Planning and Acting in Real World Ramin Halavati ([email protected]). In which we see how more expressive representations and more interactive agent architectures lead to planners that are useful in real world. Outline. - PowerPoint PPT Presentation

Transcript of An Introduction to Artificial Intelligence CE 40417

Page 1: An Introduction to  Artificial Intelligence CE 40417

An Introduction to Artificial Intelligence

CE 40417

Chapter 12 – Planning and Acting in Real World

Ramin Halavati ([email protected])

In which we see how more expressive representations and more interactive

agent architectures lead to planners that are useful in real world.

Page 2: An Introduction to  Artificial Intelligence CE 40417

Outline

• Time, Schedules, and Resources

• Hierarchical Task Network Planning

• Planning and Acting in Nondeterministic

Domains

• Multi Agent Planning

Page 3: An Introduction to  Artificial Intelligence CE 40417

Time, Schedules, & Resources

• Basic Planning:

– What to do and in which order?

• Real World:

– What an When to do? + Limited Resources.

– JOB SHOP SCHEDULING

Page 4: An Introduction to  Artificial Intelligence CE 40417

Job Shop Scheduling

Page 5: An Introduction to  Artificial Intelligence CE 40417

Job Shop Scheduling

• How to assign time to a partial order

plan?

Page 6: An Introduction to  Artificial Intelligence CE 40417

Critical Path Method (CPM)

• Forward March:

– Set Earliest Start (ES)

Page 7: An Introduction to  Artificial Intelligence CE 40417

Critical Path Method (CPM)

• Backward March:

– Set Latest Start (LS)

Page 8: An Introduction to  Artificial Intelligence CE 40417

Critical Path Method (CPM)

Page 9: An Introduction to  Artificial Intelligence CE 40417

Limited Resources

• Resources:

– Consumable vs. Reusable.

• Notation:

– Aggregation

– Immediate Effect

– Resource:R(k)

• Requirement / Temporary Effect

Page 10: An Introduction to  Artificial Intelligence CE 40417

Limited Resources

• No General Approach (NP-Hard)

• Just Order the task so that the

requirements are met.

• Heuristic:

– Minimum Slack Algorithm:

• Give more priority to the task with least remaining

slack.

Page 11: An Introduction to  Artificial Intelligence CE 40417

Job Shop Scheduling, One Last Word.

• Separated / Integrated Planning and

Scheduling.

• Semi Automatic

Page 12: An Introduction to  Artificial Intelligence CE 40417

Hierarchical Planning

• Hierarchical Task Network:

– At each “level,” only a small number of

individual planning actions, then descend to

lower levels to “solve these” for real.

– At higher levels, the planner ignores “internal

effects” of decompositions. But these have

to be resolved at some level…

Page 13: An Introduction to  Artificial Intelligence CE 40417

HTN Sample

• Construction Domain:

– Actions:

• Buy Land: Money Land

• Get Load: Good Credit Money

• Get Permit: Land Permit

• Hire Builder: Contract

• Construction: Permit Contract House Built

• Pay Builder: Money House Built House

• …

Page 14: An Introduction to  Artificial Intelligence CE 40417

HTN Sample (cont)

• Macro Action in Library:

– Build House:

Page 15: An Introduction to  Artificial Intelligence CE 40417

HTN Sample (cont)

Page 16: An Introduction to  Artificial Intelligence CE 40417

HTN Sample (cont)

Page 17: An Introduction to  Artificial Intelligence CE 40417

HTN Cons and Pros

• What’s Bad?

– Recursion?

– Sub Task Sharing:

• Enjoy honey moon in Hawaii and raise a family.

• Library: – Enjoy Honey moon in Hawaaii: Get Married , Go to

Hawaii.

– Raise Family: Get Married, Have two children.

Page 18: An Introduction to  Artificial Intelligence CE 40417

HTN Cons and Pros

• What’s Good:

– Almost all real applications are HTN + some

thing else.

– It’s a heuristic to decrease the branching

factor by a great level.

Page 19: An Introduction to  Artificial Intelligence CE 40417

NonDeterministic Domains

• What if we don’t know all about situations

and effects.

• E.g.

– Init: A table and a chair of unknown colors.

– Goal: A table and a chair of the same colors.

– Condition: Painting may have flaws.

Page 20: An Introduction to  Artificial Intelligence CE 40417

Sensorless Planning

• We don’t know all beforehand and we

can’t find it out, even when it is done.

– Plan so that to reach the goal state,

regardless of everything. (Coercion)

– Not always possible.

Page 21: An Introduction to  Artificial Intelligence CE 40417

Conditional Planning

• We can check the state ahead, then

perform the pre-planned program.

– Sense Actions

– Conditional Branches

Page 22: An Introduction to  Artificial Intelligence CE 40417

Conditional Planning in Fully Observable Domains

• Vacuum World:

– Left: AtRight AtLeft AtRight

– Left: AtRight

(AtLeft AtRight) (AtLeft AtRight)

– Suck: when AtLeftCleanLeft

when AtRightCleanRight

– Left: when AtLeft CleanLeft

when AtRightAtLeft AtRight

Page 23: An Introduction to  Artificial Intelligence CE 40417

Notation Expantion:

• Expanding Plan Notation:

– If (state) Then (…) else (…)

– If (AtLeftCleanLeft CleanRight) Then {}

else Suck.

Page 24: An Introduction to  Artificial Intelligence CE 40417

State Space:

Page 25: An Introduction to  Artificial Intelligence CE 40417

Conditional Planner:

Page 26: An Introduction to  Artificial Intelligence CE 40417

Unavoidable Loops in Conditional Planner

• New Notation:

– Instead of just Left : while (AtRight) Left

Page 27: An Introduction to  Artificial Intelligence CE 40417

Partially Observable Domains

Page 28: An Introduction to  Artificial Intelligence CE 40417

Partially Observable Domains

• Easiest Approach:

– Assume set of current states and the next

state sets are created, quite similar to non-

deterministic actions case.

Page 29: An Introduction to  Artificial Intelligence CE 40417

Execution Monitoring and Replanning

• Check if the plan is going on is pre-

decided? If not, replan based on current

situation.

Page 30: An Introduction to  Artificial Intelligence CE 40417

Execution Monitoring & Replanning

• Action Monitoring:• See if current state is as it was supposed, if not,

find a solution to return it to what it was (repair).

Page 31: An Introduction to  Artificial Intelligence CE 40417

• Plan Monitoring:

– See if the previous plan is still wise?

– Serendipity!

– A precondition of future actions has failed

and can not be recovered.

Execution Monitoring & Replanning

Page 32: An Introduction to  Artificial Intelligence CE 40417

Execution Monitoring in Partially Observable Domains

• Things may fail and we don’t know.

• Sensing actions may be required

– And they may need extra-planning.

• We may stuck in futile attempts:

– The electronic key is incorrect, but we think

it might be due to incorrect pushing in.

Page 33: An Introduction to  Artificial Intelligence CE 40417

Continues Planner

• Keep planning, sensing and executing…

– Which is not unlikely, such as maintenance

planning, auto-pilot, plant control, …

Page 34: An Introduction to  Artificial Intelligence CE 40417

Continues Planner

Page 35: An Introduction to  Artificial Intelligence CE 40417

Continues Planner

Page 36: An Introduction to  Artificial Intelligence CE 40417

Continues Planner

Page 37: An Introduction to  Artificial Intelligence CE 40417

Continuous Planner

• POP + …

– Missing Goal:

• A new goal has erupted. Just add it.

– Open precondition:

• An action has lost its support links. Add a new

causal link.

– Causal Conflicts:

• A causal link is suddenly threatened. Choose an

appropriate ordering.

Page 38: An Introduction to  Artificial Intelligence CE 40417

Continuous Planner

• POP + …

– Unsupported Link:

• A link from start to something has suddenly last

its true value. Remove it.

– Redundant Action:

• An action no more produces something needed.

Remove it.

Page 39: An Introduction to  Artificial Intelligence CE 40417

Uncertainty is Over.

Page 40: An Introduction to  Artificial Intelligence CE 40417

Multi Agent Planning

• When there is more than one agent in the

scene.

– Competitive

– Cooperative

• Coordination– Communication

Page 41: An Introduction to  Artificial Intelligence CE 40417

Cooperation

• Multi Body Planning

– One is in charge of all decisions…

• Having the agent as one of parameters:

– Go(R2D3, Right) ^ Go(C3PO,Left).

• Synchronization and Timing…

Page 42: An Introduction to  Artificial Intelligence CE 40417

Cooperation – Multi Body

• Joint Planning:

– Planning using action pairs:

• Exponentially Many Actions: Actions Agents

– Having Concurrent Actions List

• Which actions happen together and which not,

such as orders in POP.

Page 43: An Introduction to  Artificial Intelligence CE 40417

Cooperation - Coordination

• Accepting a prior Convention.

– Everyone drive on his/her right side of the

road.

– Domain Independent:

• Choosing the first feasible action.

• Producing all possible feasible actions and

choosing the one which stands first in alphabetic

order!

Page 44: An Introduction to  Artificial Intelligence CE 40417

Cooperation – Emergence

• Evolutionary Emergent Behavior

– Birds Flocking:

• Separation

• Cohesion

• Alignment

– Ants.

Page 45: An Introduction to  Artificial Intelligence CE 40417

Coop. - Communication

• A short message expressing

– the plan / next step.

• A message expressing the next step.

• Plan Recognition!

Page 46: An Introduction to  Artificial Intelligence CE 40417

Competition

• Minimax + Conditional Planning

Page 47: An Introduction to  Artificial Intelligence CE 40417

Essey & Project Proposals

• To Do.