Integrating Sensemaking

1
Integrating Sensemaking Response using Planning
Operator Induction

• Prof. dr. T.J. Grant
• Professor, Operational ICT
• T 31 (0)76 52 73261 (Mon-Wed)
• M 31 (0)638 193 749
• TJ.Grant_at_mindef.nl

2
Outline

• Introduction
• Sensemaking
• Planning Operator Induction
• Aircraft hijacking pre- post-9/11
• Applying POI to sensemaking
• Further work

3
Introduction (1)

• About myself
• Qualifications
• BSc, Aeronautical Engineering, U. Bristol (UK)
• Defence Fellowship (Masters), U. Brunel (UK)
• PhD (Artificial Intelligence), U. Maastricht (NL)
• Past experience
• Royal Air Force officer (20 years)
• Principal Consultant, Atos Origin (17 years)
• Current responsibilities
• Officer cadet education research
• ICT support for military operations

4
Introduction (2)

• CRAM emphasises planning
• BUT
• No plan survives 1st contact with enemy
• Fog of war uncertainty complexity
• What can we do after 1st contact?
• Make sense of situation, then (re-) plan
• My approach
• Integrate sensemaking planning using Planning
  Operator Induction (POI)
• N.B. Research in progress

5
Sensemaking (1)
Monitoring
6
Sensemaking (2)
7
Planning Operator Induction (1)
8
Planning Operator Induction (2)
From Fig 7.3 (p.283) in Nilsson, 1980
9
Aircraft hijacking (1)
Normal flight
Pilot
flies
Aircraft
10
Aircraft hijacking (2)
Hijacking pre-9/11
Pilot
flies
Aircraft
11
Aircraft hijacking (3)
Hijacking pre-9/11
9/11 hijackings
Pilot
Pilot
flies
Aircraft
Aircraft
12
Applying POI to sensemaking (1)

• Sensemaking changing domain model as result of
  observations
• Effect of ERM change
• Add entity - more states
• Add relationship - more states
• Relax constraint - more states
• Add constraint - less states
• More states
• Augmented STN

13
Applying POI to sensemaking (2)
Suppose original domain was just Blocks Table
14
Applying POI to sensemaking (3)
STN, original domain
From Fig 11.3 (p.265) in Genesereth Nilsson,
1987
15
Applying POI to sensemaking (4)
STN, augmented domain (original in green)
16
Further work

• Add functionality to POI algorithm to
• Monitor domain for changes
• Produce modified ERM
• Induce augmented STN
• Map augmented onto original STN
• Identify current state in augmented STN
• Select goal state in original STN
• Find response plan
• Embed tool in test environment
• Test using complex cases domains

17
Additional references

• Genesereth, M.R., Nilsson, N.J. 1987. Logical
  Foundations of Artificial Intelligence. Morgan
  Kaufman, Palo Alto, CA, USA. ISBN 0-934613-31-1
• Grant, T.J., van den Herik, H.J., Hudson,
  P.T.W. 1994. Which Blocks World is the Blocks
  World? In P. Prosser. (ed) Proceedings, 13th
  workshop, UK Planning Scheduling SIG,
  University of Strathclyde, Glasgow, UK, 14/15
  September 1994, ISSN 1368-5708
• Nilsson, N.J. 1980. Principles of Artificial
  Intelligence. Tioga Press / Morgan Kaufman, Los
  Altos, CA, USA. ISBN 3-540-11340-1

18
Any questions?