Automated Scenario Generation Toward Tailored and Optimized Military - - PowerPoint PPT Presentation

▶
automated scenario generation
SMART_READER_LITE
LIVE PREVIEW

Automated Scenario Generation Toward Tailored and Optimized Military - - PowerPoint PPT Presentation

Jan 18, 2024 168 likes •407 views

Entertainment Intelligence Lab Automated Scenario Generation Toward Tailored and Optimized Military Training in Virtual Environments Alex Zook Stephen Lee-Urban, Mark Riedl, Heather Holden, Robert Sottilare, Keith Brawner Scenario-based

slide-1
SLIDE 1

Entertainment Intelligence Lab

Automated Scenario Generation

Toward Tailored and Optimized Military Training in Virtual Environments

Alex Zook Stephen Lee-Urban, Mark Riedl, Heather Holden, Robert Sottilare, Keith Brawner

slide-2
SLIDE 2

Scenario-based Training

  • Scenario – script of

events for training purposes

patrol(market) make-friends(private) bullied(sergeant) ambush() get-shot(private, leg) get-shot(sergeant, chest) enemy-retreat() give-care(sergeant, patch) get-thanked(sergeant) die(private)

…

slide-3
SLIDE 3

Scenario-based Training Challenges

  • Repeat to learn

– Many contexts for same skill

patrol(market) make-friends(private) bullied(sergeant) ambush() get-shot(private, leg) get-shot(sergeant, chest) enemy-retreat() give-care(sergeant, patch) get-thanked(sergeant) die(private)

…

drive-to(village) investigate(house) attack(villager) subdue(villager)

slide-4
SLIDE 4

Scenario-based Training

  • Repeat to learn

– Many contexts for same skill

  • Varying learner

needs

– Tailoring to user abilities

patrol(market) make-friends(private) bullied(sergeant) ambush() get-shot(private, leg) get-shot(sergeant, chest) enemy-retreat() give-care(sergeant, patch) get-thanked(sergeant) die(private)

…

get-shot(sergeant, arm)

give-care(sergeant, tourniquet)

slide-5
SLIDE 5

Scenario-based Training

  • Repeat to learn

– Many contexts for same skill

  • Varying learner needs

– Tailoring to user abilities

  • Changing deployment

contexts

– Reauthoring content

patrol(market) make-friends(private) bullied(sergeant) ambush() get-shot(private, leg) get-shot(sergeant, chest) enemy-retreat() give-care(sergeant, patch) get-thanked(sergeant) die(private)

…

patrol(jungle)

slide-6
SLIDE 6

Scenario Generation Goals

1.

  • 1. Augment authoring volume with

automated generation 2.

  • 2. Tail

ilor scenarios to individual differences

  • 3. Generate content on
  • n-demand
slide-7
SLIDE 7

Automated Scenario Generation

  • Automated generation of training scenarios

given knowledge of:

– learning objectives – learner attributes – domain knowledge

  • domain content
  • scenario quality evaluation
slide-8
SLIDE 8

Automated Scenario Generation

author domain knowledge learner scenario generator scenario learning

  • bjectives

learner attributes

slide-9
SLIDE 9

Automated Scenario Generation

author domain knowledge learner scenario generator scenario learning

  • bjectives

learner attributes authoring augmentation content tailoring

slide-10
SLIDE 10

Generation Methods

  • planning vs genetic algorithms

– causal coherence vs evaluation optimality – result construction vs iterative modification – construction knowledge vs result evaluation knowledge

  • incremental vs final result criteria
slide-11
SLIDE 11

Generation Methods

patrol(market) make-friends(private) bullied(sergeant)

…

ambush() ambush() give-care(private, arm)

PLANNING

slide-12
SLIDE 12

Generation Methods

patrol(market) make-friends(private) bullied(sergeant)

…

PLANNING

patrol(market) make-friends(private) bullied(sergeant)

…

GENETIC ALGORITHM

ambush() give-care(private, arm) make-friends(private) bullied(sergeant)

slide-13
SLIDE 13

Genetic Algorithms

  • Inputs:

– Event templates – Event ordering constraints – Evaluation grammar

  • Output:

– Scenarios with fitness values

14

Initialization Selection Reproduction Termination

slide-14
SLIDE 14

Generation

  • Event templates

give-care(?character, ?care-type) get-shot(?character, ?injury type) make-friends(?character)

slide-15
SLIDE 15

Evaluation

  • Evaluation

– evaluation functions

  • character use
  • event use
  • scenario length
  • …

– evaluation grammar – learner model

slide-16
SLIDE 16

Evaluation Functions

  • example: character use

+ few characters + character reuse across events

slide-17
SLIDE 17

Evaluation Grammar

give-care(?character, ?care-type) get-shot(?character, ?injury type) make-friends(?character)

hurt-friend

get-shot(?character, ?injury type)

injury-care care-friend hurt-friend injury-care

slide-18
SLIDE 18

Learner Model

  • Match predicted to desired performance

Performance Events

slide-19
SLIDE 19

Scenario Generator Evaluation

  • How do you compare generation systems?
  • System dynamics

– Quality over time – Diversity over time

  • Human evaluation
slide-20
SLIDE 20

System Dynamics

  • Scenario Quality

– evaluation functions + evaluation grammar

slide-21
SLIDE 21

System Dynamics

  • Scenario Diversity

– scenario population edit distance

slide-22
SLIDE 22

Human Evaluation

  • Generator measures

– actual vs predicted performance

  • Subjective measures

– difficulty – enjoyment

  • Outside validation

– paper test of learning – on-field performance

slide-23
SLIDE 23

Thanks!

Questions?