1 of 18

The UK’s Intelligent Ship Project

26^th ICCRTS

Topic 4: C2 and human-AI/autonomy teaming

Joshua Cox

UK OFFICIAL

22/09/2021

2021 DSTL

DSTL/PUB134988

2 of 18

Intelligent Ship – The Challenge

The challenge

Complexity and volume of data available in the battlespace is ever growing.
Future threats present greater challenges. Faster, harder to detect and identify.
Increasingly congested & complex environment, picture may be degraded or denied.

The impact

Decisions will need to be faster and exploit this increasing volume of data
Potential strain & overload for operators and current information & communication systems
Need for command structures and systems to support increased information
Need for greater automation and utilisation of technologies such as Artificial intelligence
Changes to platform design to support the information and data flows required

22/09/21

3 of 18

Intelligent Ship – Project Vision and Aims

Project aims:

To develop AI and Autonomy concepts and technologies that could transform future military C2
Enable fuller exploitation of information advantage
Develop understanding of Human-Machine teaming challenges & opportunities
Investigate the system level design, initial use case of a naval platform but research will be applicable to other domains
Revolution not evolution - Not constrained by current design and processes

22/09/21

Looking at a 2040+ timescale

”Machine learning and Artificial Intelligence will be more closely integrated and teamed with humans, leading to timely, more informed and trusted decision making and planning, within complex operating environments”

4 of 18

Intelligent Ship - Project Phases

Phase 1 – Enablers

Development of Intelligent Ship AI Network (ISAIN) – an environment to explore human machine teaming
Development of Tactical Navigation (TACNAV) AI
Design and development of AI concepts (including the concept of ADeMs) to support Intelligent Ship
Initial interactions, potential command flows

Phase 2 – Integration

Further development of ISAIN
Further development of ADeMs (Agents for Decision Making)
Integration work
Evaluation and Demonstration of the Intelligent Ship concept focusing on human machine teaming

5 of 18

ADeMs and Enablers

22/09/2021

2021 DSTL

6 of 18

Phase 1

22/09/21

Initial development of supporting AI/ML agents & tools

Decision Aides

Human Interfaces

Enablers

Planners

Platform Systems

‘Sandpit’

Intelligent Ship AI Network [ISAIN]

A framework to support experimentation with AI and Human-Machine teaming

Background study

Platform Design Risks & Opportunities study [PeDRO]

Capturing the wider context, impacts, risks & opportunities of wider use of machine intelligence & automation

[VER] - Virtual Engine Room - Future, Intelligent Energy System for the Mission

The [Cobotic] Ops Room: Human-AI Dialogue in Ops Room

[IBIS(IS)] - Damage control strategy automation by deep learning

[FTEWA]

Force TEWA using Deep Reinforcement Learning

[TACNav]

Tactical Navigation

Optimising Interactions Between Humans & Intelligent Systems

[MATE]

Measuring Autonomy Team Effectiveness

[CIAO]

Compounded Intelligent Agents for Optimisation

[AMFP]

Alternative Maritime Futures Prediction

[AiDA] - Artificial Intelligence Decision Aid: A virtual ship persona

7 of 18

Phase 2

22/09/21

Continued work

TACNAV

SYCOIEA

AMFP2 - Montvieux

IBIS2 - Fraser Nash

CIAO2 - decisionLab

ISAIN - Intelligent Ship AI Network

Platform systems

ACE - Intelligent decision-making for vessel power & propulsion control - Rolls Royce

Human factors

HADES - Human-Agent Design & EvaluationS – Nottingham Trent Uni

Command process

GALILEO - GoAL based decomposition for InteLligent ship AI nEtwOrk – SeeByte

KNOT - Knowledge-based Naval Orders Toolset - Montvieux

TE2C

Red Mirror – Mimics and predicts red forces future actions – DIEM

Background

Evaluation plan & Metrics

Interactions work

Scenario development

8 of 18

ISAIN

22/09/21

An environment to develop and test AI-AI and AI-Human collaborative teams

[ISAIN] - Intelligent Ship AI Network

Capabilities

Provides a framework to support experimentation with multiple AI Agents and Human-Machine teaming
Aides evaluation of potential benefits and disadvantages of AI/AI-Human collaboration
Aides in the evaluation and demonstration of specific AI’s in a naval area & inform future naval design (But is applicable to other domains).
Enables the evaluation of team performance
Management, arbitration & de-confliction of outputs across the system to meet overall objectives
Decomposition & adaptive allocation of tasks

9 of 18

GALILEO – SeeByte

AIM - to offer a goal based mission decomposition tool into ISAIN
GALILEO is a mission focused ADeM that will integrate within ISAIN, and team with human operators and co-ordinate with other ADeMs
Aims to support decision making during pre-mission preparation, mission execution and post mission analysis.
GALILEO will be built around the existing capabilities of the Neptune AI Goal Based Autonomy and Agent Teaming product

22/09/21

Goal based decomposition
Estimates capabilities based on ADeMs
Decompose to individual tasks
Allocate tasks to ADeMs
Presents solutions to operators

Goal based decomposition for Intelligent Ship AI Network

10 of 18

KNOT – Montvieux

AIM - This will decompile human-written formal naval orders into a knowledge-graph, suitable for semantic querying by ADeMs supporting machine reasoning
KNOT will enable ADeMs to understand the meaning embedded within human-authored orders
KNOT will also have the ability to take outputs from ADeMs, such as recommended CoAs, and recompile this back into human readable formal orders

22/09/21

Knowledge-based Naval Orders Toolset

11 of 18

Evaluations and demonstrations

22/09/2021

2021 DSTL

12 of 18

Intelligent Ship – Interactions

22/09/21

Investigating potential interactions between ADeMs.

Make sure there is a sufficient system to enable a full demonstration.

Investigating information flows, command process, data exchange, task breakdown, human interaction, Decision points, COA formation, arbitration

13 of 18

Intelligent Ship – Demonstrations and evaluations

OFFICIAL

22/09/21

Scenario has been designed to test all of the different command threads currently present in the intelligent ship.

Building up from simpler demonstrations to full integration.

Choosing metrics to evaluate the overall system, the teaming and individual agents.

Also want to identify:

Are there bottlenecks?
Is there any emergent behaviour?
Are there any gaps in the command process?
Are the interfaces suitable for the operators?

14 of 18

Summary

22/09/2021

2021 DSTL

15 of 18

Intelligent Ship – Next steps

The next steps…

Finishing touches on integration
November 2021 – Full integration & scenario walkthrough
February 2022 – Final Evaluation
Analyse outcomes of evaluation
Plan the way forward
Potential exploitation

OFFICIAL

22/09/21

16 of 18

Intelligent Ship - Summary

Aims to:

Revolutionise ship design & operation
Offer enhanced autonomy & human-machine teaming
Efficiently & effectively use complex & growing, data & information sets
Enable commanders to make faster and higher quality decisions
Demonstrate concepts in 2021
Develop a platform agnostic methodology to explore flexible and adaptive human autonomy teaming

OFFICIAL

22/09/21

17 of 18

Intelligent Ship – Contacts and acknowledgements

ACKNOWLEDGMENTS

Dstl acknowledges the hard work done by our main suppliers:

Affect In
BMT
CGI IT UK
Daden
decisionLab
DIEM Analytics
Fraser Nash Consulting
GE Power Conversion
Montvieux
Nottingham Trent University
Roke Manor Research
Rolls-Royce
SeeByte
Thales

Contact details

Joshua Cox – Jcox@dstl.gov.uk

18 of 18

UK OFFICIAL

© Crown copyright (2021), Dstl. This material is licensed under the terms of the Open Government Licence except where otherwise stated. To view this licence, visit http://www.nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk

Discover more