1 of 28

Hans-Christian Schmitz, Jonathan Cawalla

Graph-based Representation of Military Action Spaces

Formatvorlagen des Textmasters bearbeiten

Zweite Ebene

      • Dritte Ebene
        • Vierte Ebene
          • Fünfte Ebene

2 of 28

Outline

Graph-based Representation of Military Action Spaces

  1. Motivation
  2. Concept
  3. Applications
  4. Summary

© Fraunhofer FKIE

Page 2

Formatvorlagen des Textmasters bearbeiten

Zweite Ebene

      • Dritte Ebene
        • Vierte Ebene
          • Fünfte Ebene

3 of 28

Command and Control (C2) Support

[Motivation]

Support of the C2 process

        • Cognitive relief for operators and higher quality of decisions
          • Timely processing of available information
          • Avoidance of errors
        • Reduction of resource requirements
          • Create conditions for smaller, more mobile command posts that can be well protected
          • Improve C2 capability also at lower tactical levels

​

​

© Fraunhofer FKIE

Page 3

Situation assessment and control

Estimation of the situation and decision seeking

Decision

Planning

Issuance of order

4 of 28

Command and Control (C2) Support

[Motivation]

Support of the C2 process

        • Cognitive relief for operators and higher quality of decisions
          • Timely processing of available information
          • Avoidance of errors
        • Reduction of resource requirements
          • Create conditions for smaller, more mobile command posts that can be well protected
          • Improve C2 capability also at lower tactical levels

​

Further development of C2IS

        • Enhancement of usefulness and usability

© Fraunhofer FKIE

Page 4

5 of 28

Artificial Intelligence for C2IS

[Motivation]

        • Mastering of strategy games like Chess, Go, Star Craft II: similar success possible in the military domain?
        • Experiments in Deep Reinforcement Learning

​

​

© Fraunhofer FKIE

Page 5

6 of 28

Artificial Intelligence for C2IS: Challenges

[Motivation]

        • The complexity of military scenarios is very high.
        • Environments change rapidly:
          • terrain
          • resources
          • doctrine
          • mission goals
        • Military operations cannot be repeated as often as required for training a system. Training data is not easily available.

​

​

​

​

​

​

© Fraunhofer FKIE

Page 6

7 of 28

Artificial Intelligence for C2IS: Challenges

[Motivation]

        • The complexity of military scenarios is very high.
        • Environments change rapidly:
          • terrain
          • resources
          • doctrine
          • mission goals
        • Military operations cannot be repeated as often as required for training a system. Training data is not easily available.
        • To do:
          • Further research in technological domain.
          • Meaningful reduction of complexity of C2 task definitions.
          • Identification of classes of solvable C2 decision problems.

© Fraunhofer FKIE

Page 7

8 of 28

Artificial Intelligence for C2IS: Challenges

[Motivation]

        • The complexity of military scenarios is very high.
        • Environments change rapidly:
          • terrain
          • resources
          • doctrine
          • mission goals
        • Military operations cannot be repeated as often as required for training a system. Training data is not easily available.
        • To do:
          • Further research in technological domain.
          • Meaningful reduction of complexity of C2 task definitions.
          • Identification of classes of solvable C2 decision problems.

© Fraunhofer FKIE

Page 8

9 of 28

Decision Making

[Motivation]

        • Terrain Assessment
        • Assessment of the civilian situation
        • Assessment of the opponent’s situation
        • Assessment of the own situation
        • Assessment of possibilities of action (red and blue)
        • Comparison of forces
        • Evaluation of possibilities of action
        • Decision

​

​

© Fraunhofer FKIE

Page 9

Situation assessment and control

Estimation of the situation and decision seeking

Decision

Planning

Issuance of order

10 of 28

Decision Making

[Motivation]

        • Terrain Assessment
        • Assessment of the civilian situation
        • Assessment of the opponent’s situation
        • Assessment of the own situation
        • Assessment of possibilities of action (red and blue)
        • Comparison of forces
        • Evaluation of possibilities of action
        • Decision

​

​

© Fraunhofer FKIE

Page 10

Situation assessment and control

Estimation of the situation and decision seeking

Decision

Planning

Issuance of order

11 of 28

Graph-based Representation of Action Spaces

[Concept]

© Fraunhofer FKIE

Page 11

12 of 28

Attributes

[Concept]

        • Edges of the graph encode
          • march time: discrete time it takes a resource/battle space object (BSO) to move from one node to another
          • delay suitability: number of resources needed to delay march for one time step
          • capacity: how may resources of which type can pass the edge simultaneously
          • other factors

​

​

© Fraunhofer FKIE

Page 12

13 of 28

Resources, Courses of Action (CoAs), and Interventions

[Concept]

        • Representation of red and blue BSOs:
          • “resource” as an abstraction, that can be represented as a moveable token
          • different types of resources can be defined
          • attributes of edges can refer to different types of resources

© Fraunhofer FKIE

Page 13

14 of 28

Resources, Courses of Action (CoAs), and Interventions

[Concept]

 

© Fraunhofer FKIE

Page 14

15 of 28

Resources, Courses of Action (CoAs), and Interventions

[Concept]

 

© Fraunhofer FKIE

Page 15

16 of 28

Resources, Courses of Action (CoAs), and Interventions

[Concept]

 

© Fraunhofer FKIE

Page 16

17 of 28

Related Work

[Concept]

        • COAST (Course of Action Scheduling Tool)
          • Determine a sequence of fundamental actions to achieve a defined goal
          • Check for consistency with given constraints
          • Coloured petri nets as an underlying representation, not visible to the user
        • CADET (Course of Action Development and Evaluation Tool)
          • Create detailed operational plans by expanding and synchronizing initial set of high-level activities
          • Declarative, rule-based approach
          • Application of scheduling heuristics
        • WCCAAM (Wargame Commodity Course of Action Automated Analysis Method)
          • Based on a war game scenario, a directed graph representing a multi commodity flow problem (MCFP) is created
          • An optimal blue CoA is determined by solving the MCFP
        • FOX-GA (Fox Genetic Algorithm)
          • Bit string representation of blue CoAs
          • Genetic Algorithm to modify initial CoAs and find optimal CoA
          • Fitness function based on the outcome of a rule-based war gamer

© Fraunhofer FKIE

Page 17

18 of 28

Assessing Measures and Countermeasures

[Applications]

        • Determination of shortest, least time-consuming paths
          • most likely CoA
          • most dangerous CoA
        • Planning of blue forces for optimal delaying action
          • definition of target function
          • optimization by means of
            • search algorithms
            • genetic algorithms
            • reinforcement learning
            • …

​

​

© Fraunhofer FKIE

Page 18

19 of 28

Expectation Management

[Applications]

        • Control:
          • Compare situation updates with assumed CoA/possible CoAs
          • Report critical developments and deviations from CoA

​

© Fraunhofer FKIE

Page 19

20 of 28

Representation of Multi-Dimensional Structures

[Applications]

        • Challenge of creating manageable representations of multi-dimensional structures
        • Example: three-dimensional structures (buildings, tunnels, …)
        • After a graph has been derived from a two-dimensional situational picture, further nodes and edges can be added.

​

​

​

​

​

© Fraunhofer FKIE

Page 20

21 of 28

Representation of Vertical Structures

[Applications]

© Fraunhofer FKIE

Page 21

22 of 28

Representation of Vertical Structures

[Applications]

© Fraunhofer FKIE

Page 22

23 of 28

Assessing the Value of Information for Reconnaissance Planning

[Applications]

Excursus: Aggregation and Enrichment of Situational Pictures

        • Situational pictures are usually incomplete.
        • Battle field information has to be aggregated and enriched.
        • Aggregation is the assignment of detected battle space objects (BSOs) to units.
        • By enrichment, BSOs that have not yet been detected but can be assumed to be present are added to the situational pictures.
        • Alternative enrichments can be possible.

​

​

​

© Fraunhofer FKIE

Page 23

24 of 28

Assessing the Value of Information for Reconnaissance Planning

[Applications]

Specification of Reconnaissance Targets

        • Sweet spots: locations where undetected BSOs can be expected.
        • Targets with a high Value of Information (VoI): reconnaissance leads to the falsification of as many alternative enriched situational pictures as possible
        • Targets with a high CoA-related VoI: reconnaissance leads to the conformation or falsification of assumed CoAs
          • test for actual conduct of a CoA
          • test for preconditions

​

​

​

​

© Fraunhofer FKIE

Page 24

25 of 28

Graph Modelling and Analysis

[Summary]

        • We have proposed a model for the graph-based representation of military action spaces.
        • The model is close to practice, at least for land-based operations.
        • Demonstrators of useful supporting services are to be developed.

​

© Fraunhofer FKIE

Page 25

26 of 28

The Value of Graph-based Representations for Courses of Action Analysis�

[Summary]

We state the following hypotheses, which are to be evaluated:

        • A graph-based representation of the action space can be a plausible and effective simplification of the situational picture that renders the application of AI-technology for C2 possible.
        • Automatic analyses based on graph-based representations can be made transparent and understandable for human operators.
        • Graph-based representation are also suitable for human analyses of action spaces and, therefore, a means for supporting situational assessment.

​

For the testing of these hypotheses, we plan to implement demonstrators and evaluate them with military subject matter experts, ideally during exercises.

© Fraunhofer FKIE

Page 26

27 of 28

Thank you.

28 of 28

Hans-Christian Schmitz, Jonathan Cawalla

Information Technology for Command and Control

Tel. +49 821 90 678-386, -640

{hans-christian.schmitz, jonathan.cawalla}@fkie.fraunhofer.de

​

Fraunhofer Institute for Communication, �Information Processing and Ergonomics FKIE

Fraunhoferstr. 20 | 53343 Wachtberg | Germany

​

www.fkie.fraunhofer.de/en

Contact

—