23
Interactive Plan Recognition (Formerly Sequential Plan Recognition) Reuth Mirsky, Roni Stern, Ya’akov (Kobi) Gal, Meir Kalech Department of informaon systems engineering Ben-Gurion University of the Negev 1

Interactive Plan Recognition

Embed Size (px)

Citation preview

Page 1: Interactive Plan Recognition

Interactive Plan Recognition

(Formerly Sequential Plan Recognition)

Reuth Mirsky, Roni Stern, Ya’akov (Kobi) Gal, Meir KalechDepartment of information systems engineering

Ben-Gurion University of the Negev

1

Page 2: Interactive Plan Recognition

Plan recognition“Infer the plan of theagent given observationsand a data base of recipes”Kautz (1986)

2

Page 3: Interactive Plan Recognition

Hypothesis representation• Plan = tree

• Hypothesis = set of plans

• Other representations (e.g. Ramírez and Geffner ‘09)3

Page 4: Interactive Plan Recognition

Hypothesis representation

The output of a PR algorithm is a set of hypotheses

Mix all

Pair-wise Mix

4

Page 5: Interactive Plan Recognition

Number of hypotheses

51 2 3 4 5 6 7 8 9 101

10

100

1000

10000

100000

VirtualLabsTinkerPlotsSimulated

# observations

Hyp

othe

ses

Hypothesis number grows exponentially (Geib and Goldam ‘09)

Page 6: Interactive Plan Recognition

The disambiguation problem

6

Page 7: Interactive Plan Recognition

Prior work

• Pruning • Kabanza and Filion. Controlling the hypothesis space in probabilistic plan

recognition. IJCAI. 2013. • Wiseman and Shieber. Discriminatively reranking abductive proofs for plan

recognition. Twenty-Fourth International Conference on Automated Planning and  Scheduling. 2014.

• Mirsky, Gal, Shieber. CRADLE: an online plan recognition algorithm for exploratory domains. [under submission].

• Less expressive representation• Avrahami-Zilberbrand and Kaminka. Incorporating observer biases in keyhole plan

recognition (efficiently!). AAAI. Vol. 7. 2007.• Geib. Delaying Commitment in Plan Recognition Using Combinatory Categorial

Grammars. IJCAI. 2009.  • Sukthankar and Sycara. Hypothesis pruning and ranking for large plan recognition

problems. AAAI. Vol. 8. 2008.   7

Page 8: Interactive Plan Recognition

Contributions

• The Interactive Plan Recognition Process (IPRP)• Soundness and Completeness proof• Policies for solving IPRP• Extensive empirical evaluation

8

Mirsky, Gal, Stern and Kalech. Sequential Plan Recognition. AAMAS (Extended Abstract). 2016.

Mirsky, Stern, Gal and Kalech. Sequential Plan Recognition. IJCAI. 2016.

Page 9: Interactive Plan Recognition

IPRP

9

Iterative process for disambiguating the hypothesis space

Coming up:•Which hypotheses to discard?•Which plan to choose?

Page 10: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

10

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

Page 11: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

11

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

Is plan A part of the correct hypothesis?

Page 12: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

12

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

Is plan A part of the correct hypothesis?

Yes, it is part of my plan

Then B and C might alsobe part of your plan

Page 13: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

13

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

Is plan A part of the correct hypothesis?

No, it is not part of my plan

Then neither B nor C can be part of your plan

Page 14: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

14

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

Is plan B part of the correct hypothesis?

Page 15: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

15

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

No, it is not part of my plan

Then B cannot be part of your plan, but C might

Is plan B part of the correct hypothesis?

Page 16: Interactive Plan Recognition

The Refinement Criteria

Solve Problem

Create Flasks

Perform Test

Write Results

16

Solve Problem

Create Flasks

Perform Test

Pair-wise Mix

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

A

B C

Yes, it is part of my plan

We Cannot discard hypotheses with A

Is plan B part of the correct hypothesis?

Page 17: Interactive Plan Recognition

The Match Criteria

Solve Problem

Create Flasks

Create All

Perform Test

Mix All

Write Results

17

Solve Problem

Create Flasks

Create All

Perform Test

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

B C

A

Page 18: Interactive Plan Recognition

The Match Criteria

Solve Problem

Create Flasks

Create All

Perform Test

Mix All

Write Results

18

Solve Problem

Create Flasks

Create All

Perform Test

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

B C

A

Is plan B part of thecorrect hypothesis?

No, Plan B is not a part of my plan

Then A and B can not be part of the

correct hypothesis

Page 19: Interactive Plan Recognition

The Match Criteria

Solve Problem

Create Flasks

Create All

Perform Test

Mix All

Write Results

19

Solve Problem

Create Flasks

Create All

Perform Test

Write Results

Solve Problem

Create Flasks

Perform Test

Mix All

Write Results

B C

AYes, Plan B is part of my plan

Then all plans can still be part of your

plan

Is plan B part of thecorrect hypothesis?

Page 20: Interactive Plan Recognition

Soundness and completeness

• Completeness: IPR does not remove any hypothesis that can be refined to the correct hypothesis• Soundness: Every hypothesis IPR keeps can be

refined to the correct hypothesis

20

All Hypotheses

Correct

Removed

Remaining

Page 21: Interactive Plan Recognition

Which plan to query next?• Random: choose a random plan• Most Probable Hypothesis (MPH): choose a plan from the

most probable hypothesis

• Most Probable Plan (MPP): choose the most probable plan

• Entropy: choose the most informative plan

21

Page 22: Interactive Plan Recognition

Reduction in Hypotheses by Strategy

22

Page 23: Interactive Plan Recognition

Discussion

• After 1 query• 39% reduction in hypotheses (simulated domain)• 46% reduction in hypotheses (VirtualLabs domain)

• Entropy strategy outperforms all other strategies

• Find an optimal querying policy

• Apply IPRP in real world domains

• Open source code on github:

https://github.com/ReuthMirsky/IPR23

[email protected]