By Robert Dickie¹and Rachel Horne²

¹Group Leader, Systems Safety & Assurance, Frazer Nash Consultancy Ltd 

²Assurance of Autonomy Activity Lead, Trusted Autonomous Systems (TAS)

Autonomous vessels of various sizes, forms and speeds are already at sea, on the surface and beneath it. The International Regulations for Preventing Collisions at Sea (COLREGs), published by the International Maritime Organization (IMO) in 1960 and updated in 1972, govern the ‘rules of the road’ at sea. COLREGs describe the features that vessels must have to facilitate being seen and identified, define means of communication between vessels for the purposes of signaling intent, and most importantly they describe the navigational behaviors expected of vessels in proximity to one another, for the purposes of avoiding collision. It’s clear from the terms and phrases used in COLREGs that the authors didn’t conceive of navigational or operational decisions being made by computers, and compliance for autonomous vessels is difficult and not well-understood.

Autonomous systems technology does not replicate humans, it emulates some of their skills using a different set of ‘senses’ and decision-making processes and brings new capabilities to operations. This means that humans and autonomous technologies have different weaknesses, strengths, risks and mitigations.

The autonomous maritime industry has been wrestling with the challenge of ‘compliance’ with COLREGs for years, in terms of both understanding how it applies, and how to demonstrate compliance. The challenge for the designer or operator of an autonomous vessel is that the regulations are phrased from the underlying assumption that a human is operating the vessel. Where an autonomous control system is performing some or all of the functions a human previously would have been, it can be difficult to work out what constitutes ‘compliance’, in a practical sense, and in a way that the regulator, the Australian Maritime Safety Authority (AMSA), would accept. This difficulty can lead to additional costs, delay, and operations which are subject to more limitations than may be reasonable based on the actual risks presented.

Developing one-off COLREGs compliance cases for a single autonomous vessel is onerous for the designer or operator, and also causes AMSA difficulty in terms of the resources required to assess the compliance case, and ensure consistency in regulatory decision making. There are significant efficiencies to be gained for designers, operators, and regulators of autonomous vessels in the development of a repeatable compliance framework designed to reduce these burdens.

New TAS project to address COLREGs challenge

The TAS Assurance of Autonomy team have commenced a new project aimed at addressing the COLREGs challenge by developing an enabling framework which supports a practical and appropriate level of compliance for autonomous vessels. The project, known as the ‘TAS COLREGs project’, will consider the specific risks posed by the full spectrum of autonomous vessels, thus allowing flexibility in the operator’s approach to compliance. In order to provide maximum usefulness and future proofing, the approach will be to offer a range of risk mitigation options which address the ‘spirit’ of COLREGs, rather than defining a set of performance requirements or specification with which all autonomous vessels much comply.

Figure 1: Guiding principles for the TAS COLREGs project

The guiding principles for the TAS COLREGs project (see Figure 1) are:

• The compliance methodology will be repeatable and scalable across a broad range of autonomous vessels.
• The underpinning philosophy of the compliance methodology will be logical, reasoned and justified by argument.
• The compliance methodology will be enabling rather than constraining whilst upholding the purpose and spirit of COLREGs.
• The process followed to use the compliance methodology will be simple to follow and supported by guidance.
• The project aims to develop a methodology which can be agreed to by the regulator.

Outputs of the TAS COLREGs project includes an operational tool and user guidance.

Stakeholder engagement will occur over the coming weeks to ensure the tool is fit for purpose and considered acceptable and appropriate by AMSA. This tool will be trialed with a variety of autonomous vessels in 2021, and then refined as necessary, before being released.

The TAS COLREGs project is being delivered by Frazer-Nash Consultancy and led by Robert Dickie, with the support of Marceline Overduin and Andrejs Jaudzems. Robert has extensive experience in the assurance of maritime autonomous systems, having been a member of the UK Maritime Autonomous Systems Regulatory Working Group (MASRWG) and having developed the initial draft of the Lloyd’s Register Code for Un-crewed Marine Systems.

Robert Dickie, Frazer-Nash Consultancy

The output of the TAS COLREGs project will be available on the TAS website as a useful resource for designers and operators in the form of a repeatable COLREGs compliance framework, supported by a tool and user companion guide.

The COLREGs resource will also be included in the Body of Knowledge Toolbox being developed by the TAS Assurance of Autonomy team under the National Accreditation Support Facility Pathfinder Project (NASF-P), which will assist designers, operators, regulators, and other stakeholders in the Australian autonomous systems ecosystem to navigate the assurance and accreditation process for autonomous systems efficiently and successfully.

If you would like to contact us in relation to the TAS COLREGs project, to offer feedback, suggestions, or request more information, please email us at NASFP@tasdcrc.com.au.

By Tom Putland – Director of Autonomy Accreditation – Air

The development of best practice policy, appropriate standards, and a strong assurance and accreditation culture has the potential to enhance innovation and support market growth for drones with autonomous abilities in the maritime, air and land domains.

The Trusted Autonomous Systems  (TAS) National Accreditation Support Facility Pathfinder Project (NASF-P), under the Assurance of Autonomy Activity, represents an opportunity to unlock Queensland’s, and by extension Australia’s, capacity for translating autonomous system innovation into deployments, given the existing test facilities operating in Queensland, the existing industry need identified and strong government backing.

The overarching purpose of the NASF-P is to:

• Make it easier to design, build, test, assure, accredit, and operate an autonomous system in Australia, without compromising safety; and
• Support and promote Queensland’s existing test ranges; and
• Encourage both domestic and international business to operate in, and use Queensland as a base for the purpose of testing, assuring, and accrediting autonomous systems; and
• Investigate, design, and facilitate the creation of an appropriate, independent third party entity that can continue to support the design, build, test, assurance, accreditation, and operation of autonomous system in Australia, by bridging the gap between industry, operators, and regulators.

New project: Development of a DAA DT&E guideline

In the air domain specifically, the largest impediment to the integration of unmanned aircraft (regardless of autonomy) into the National Airspace System (NAS) is complying with the intent[1] of the See and Avoid (SAA) requirements detailed within Regulations 163 and 163A of the Civil Aviation Regulations 1988 (Cth), particularly in Beyond Visual Line of Sight (BVLOS) operations. In lieu of a solution, operators and the regulator must go through a labour-intensive stakeholder engagement process with all aviation parties to prevent mid-air collisions. This approach will not scale to the projected numbers of Unmanned Aerial Systems (UAS) operations into the future.

An autonomous/highly automated DAA system that complies with the safety objectives of CAR 163 and CAR 163A is a key enabling technology for integration into complex Australian airspace and will form an integral part of the safety assurance framework for UAS operations into the future.

The TAS team have initiated a new project, led by Revolution Aerospace, to develop a new Detect and Avoid (DAA) Design, Test and Evaluation (DT&E) guideline for low-risk, uncontrolled airspace outside the airport environment. This is particularly relevant to Australian unmanned aircraft operations.

Dr Terry Martin (CEO) and the Revolution Aerospace team have a wealth of world-leading experience in Detect and Avoid, Machine Learning, Safety Assurance, Verification & Validation (V&V), and Test & Evaluation (T&E), representing Australia at international forums such as NATO, JARUS, and RTCA (to name a few).

Dr Terrence Martin, CEO of Revolution Aerospace

Terry is supported in this project by TAS’s Tom Putland – Director of Autonomy Assurance in the Air Domain. Tom has extensive experience in the regulation and safety assurance of UAS, having previously worked for CASA, and being CASA’s representative at JARUS and other international working groups. This team has the expertise, drive, and ability to solve this critical airspace problem for Australia.

Tom Putland, Director of Autonomy Assurance – Air at TAS

DT&E guideline will create a process acceptable to CASA[2] that allows:

• the derivation of high-level safety objectives; and
• development of Verification and Validation (V&V) requirements; and
• the conduct of relevant simulations and tests to demonstrate compliance with the safety objectives; and
• the collation of compliance process and data into a package to support regulators (e.g. CASA) in issuing an approval for the operation.

TAS will engage closely with key stakeholders, including CASA, the Australian Association for Unmanned Systems (AAUS) and other industry members to ensure the process reflects current best practice, and is appropriate and useful for the Australian aviation industry. The intent is for the new process to be available for testing by the end of the year.

Upon completion of this project a subsequent project will be undertaken to work with industry partners and the Queensland Flight Test Range, Australia’s first and only commercial flight test range, to utilise and comply with this guideline. Completion of these projects will increase the access and flexibility available for unmanned aircraft operations in Australian airspace.

If you are interested in learning more about these projects, be involved as an industry test partner, or discuss specifics (i.e. EO/IR sensing, classifiers/learning assurance, alerting and avoidance logic, formal verification, T&E), please email tom.putland@tasdcrc.com.au.

Other NASF-P projects underway

The NASF-P team have a number of projects underway, including:

• Preparation of a Body of Knowledge on the assurance and accreditation of autonomous systems;
• Maritime domain: development of a repeatable, regulator-accepted methodology to demonstrate compliance with COLREGS for autonomous and remotely operated vessels; and
• Preparation of a business case for a new, independent, National Accreditation Support Facility, based in Queensland, that will better connect operators and regulators to facilitate more efficient assurance and accreditation.

If you would like to find out more about our work, or provide feedback on where you see the key risks and opportunities for the autonomous systems industry in Australia, please contact us as NASFP@tasdcrc.com.au.

[1] Through an approval under Subregulation 101.073(2) and Regulation 101.029 of the Civil Aviation Safety Regulations 1998 (Cth).

[2] Regular engagement with CASA will assist to ensure the final process is acceptable to them, but note it has not been approved  at this point in time. In the future, the process is intended to form part of an acceptable means of compliance for a BVLOS approval.

by Dr Kate Devitt, Chief Scientist, Trusted Autonomous Systems

“Military ethics should be considered as a core competency that needs to be updated and refreshed if it is to be maintained”

Inspector General ADF, 2020, p.508

The Centre for Defence Leadership & Ethics Australian Defence College has commissioned Trusted Autonomous Systems to produce videos and discussion prompts on the ethics of robotics, autonomous systems and artificial intelligence.

These videos for Defence members are intended to build knowledge of theoretical frameworks relevant to potential uses of RAS-AI to improve ethical decision-making and autonomy across warfighting and rear-echelon contexts in Defence.

Major General Mick Ryan says that he can “foresee a day where instead of having one autonomous system for ten or a hundred people in the ADF will have a ratio, that’s the opposite. We might have a hundred or a thousand for every person in the ADF”.

He asks, “how do we team robotics, autonomous systems and artificial intelligence (RAS-AI) with people in a way that makes us more likely to be successful in missions, from warfighting through to humanitarian assistance, disaster relief; and do it in a way that accords with the values of Australia and our institutional values?”

Wing Commander Michael Gan says, “robotics and autonomous systems have a great deal of utility: They can reduce casualties, reduce risk, they can be operated in areas that may be radioactive or unsafe for personnel. They can also use their capabilities to go through large amounts of data and be effective or respond very quickly to rapidly emerging threats”.

He goes on to say “however, because a lot of this is using some sort of autonomous reasoning to make decisions, we have to make sure that we have a connection with the decisions that are being made, whether it is in the building phase, whether it is in the training phase, whether it is in the data, which underpins the artificial intelligence, robotic autonomous systems”.

Trusted Autonomous Systems CEO, Professor Jason Scholz points out that “Defence has a set of behaviours about acting with purpose for defence and the nation; being adaptable, innovative, and agile; be collaborative and team-focused; and to be accountable and trustworthy to reflect, learn and improve; and to be inclusive and value others. All of these values and behaviours are included whether we are a ‘robotic and autonomous systems’ augmented force, or not”.

Managing Director of Athena Artificial Intelligence Mr Stephen Bornstein says, “When it comes to RAS-AI in Defence and ethics associated with them…. it’s very important to consider how a given company or a given AI supplier is establishing trust in that RAS-AI product”. He says that “ultimately, that assurance should be the most important thing before we start giving technologies to soldiers, seamen, or aircrew”.

Personnel engaging with the content should gain a clearer idea how to reflect on ethical issues that affect human and RAS-AI decision making in defence contexts of use including the limits and affordances of human and technologies to enhance ethical decision-making, as well as frameworks to help with RAS-AI development, evaluation, acquisition, deployment and review in Defence.

The videos draw on a framework from The Defence Science & Technology report ‘A Method for Ethical AI in Defence’ to help Defence operators, commanders, testers or designers ask five key questions about the technologies they’re working with.

1. Responsibility – who is responsible for AI?
2. Governance – how is AI controlled?
3. Trust – how can AI be trusted?
4. Law – how can AI be used lawfully?
5. Traceability – how are the actions of AI recorded?

The videos consider four tools that may assist in identifying, managing and mitigating ethical risks in Defence AI systems.

https://theodi.org/article/data-ethics-canvas/

The ‘Data Ethics Canvas’ by the Open Data Institute encourages you to ask important questions about projects that use data and reflect on the responses. Such as the security and privacy of data collected and used, who could be negatively affected and how to minimise negative impacts.

The AI Ethics Checklist ensures AI developers know: the military context the AI is for, the sorts of decisions being made, how to create the right scenarios, and how to employ the appropriate subject-matter experts, to evaluate, verify and validate the AI.

The Ethical AI Risk Matrix is a project risk management tool to identify and describe identified risks and proposed treatment. The matrix assigns individuals and groups to be responsible for reducing ethical risk through concrete actions on an agreed timeline and review schedule.

By Rachel Horne – Assurance of Autonomy Lead/Director of Autonomy Accreditation – Maritime

Autonomous systems technology offers the ability to increase safety and efficiency, while lowering economic and environmental cost. While some level of autonomy has been seen in commercial products for a number of years, for example the basic thermostat or the Roomba, in the last five years there has been a rapid acceleration in the capacity and availability of unmanned aerial vehicles known as drones, and in uncrewed surface and sub-surface vessels, also called autonomous vessels.

For this rapid acceleration to continue, and to ensure this technology can integrate into commercial and defence operations, autonomous systems need to be trusted by the government, regulators, operators, and the broader community. An integral part of gaining trust is having a clear, well-tailored regulatory framework, consistent assurance requirements and agreed assurance methodology, and support from the regulator. These same factors also facilitate innovation and promote growth in industry by providing certainty.

Coral AUV. Image by AIMS

New project: Development of an Australian Code of Practice

The NASF-P (National Accreditation Support Facility Pathfinder) team have commenced a number of new projects to address the challenges outlined above. One of these projects is aimed at addressing the lack of tailored standards for autonomous and remotely operated vessels by developing an Australian Code of Practice for the Design, Construction, Survey and Operation of Autonomous and Remotely Operated Vessels. This Code will represent best practice, and is intended to provide certainty for industry by providing a set of regulator-acknowledged standards that they can use to design, construct, survey and operate autonomous and remotely operated vessels. The Code of Practice will be voluntary, and will be updated periodically.

This project, led by Maaike Vanderkooi on behalf of TAS, will begin with a review of available Codes of Practice and Standards, for example the UK Maritime Autonomous Surface Ships (MASS) UK Industry Conduct Principles and Code of Practice, and Lloyd’s Register Unmanned Marine Systems Code. The project will then develop a draft Australian Code of Practice, using input from key stakeholders, which will then be released for broader public consultation. The intent is to release a draft Code of Practice by October 2021, which will be available for use by industry and the regulator.

Maaike Vanderkooi has been chosen to lead the project as a result of her extensive experience in developing regulatory frameworks in the maritime, heavy vehicle and ports arenas, and her experience in developing, reviewing and impact assessing commercial vessel standards.

Maaike Vanderkooi

TAS will engage closely with key stakeholders, including the Australian Maritime Safety Authority (AMSA), the Australian Association for Unmanned Systems (AAUS) Maritime Working Group, the Marine Surveyors Association Inc, and the Australasian Institute of Marine Surveyors, throughout this project to ensure the Code of Practice is practical and appropriate for use by Australian industry and the regulator. There will also be opportunities for input by interested parties throughout the project.

Engagement opportunities

• We are looking for people with direct experience applying current Codes of Practice or Standards to autonomous and remotely operated vessels, to discuss their experience and provide feedback to us in May 2021;
• We will hold a series of workshops with key stakeholders between May and August 2021; and
• We will release the draft Code of Practice for public consultation in August 2021, and welcome all thoughts and feedback.

If you would like to contact us in relation to this project, to offer feedback, suggestions, or your assistance, please email us at NASFP@tasdcrc.com.au.

QUT WAM-V in operation at AIMS. Image by AIMS

Other NASF-P projects underway

The NASF-P team have a number of projects underway, including:

• Preparation of a Body of Knowledge on the assurance and accreditation of autonomous systems;
• Air domain: development of an end-to-end acceptable process for the design, build, test and evaluation of autonomous detect and avoid (DAA) systems for certain types of airspace;
• Maritime domain: development of a repeatable, regulator-accepted methodology to demonstrate compliance with COLREGS for autonomous and remotely operated vessels; and
• Preparation of a business case for a new, independent, National Accreditation Support Facility, based in Queensland, that will better connect operators and regulators to facilitate more efficient assurance and accreditation.

The NASF-P team recently worked with Queensland AI Hub, Australian Institute of Marine Science, and AMC Search, supported by Advance Queensland, to deliver a world-first pilot course ‘Autonomous Marine Systems Fundamentals for Marine Surveyors’. This course, which was created to address the gap in experience with autonomous marine systems amongst the accredited marine surveyor community, had nine participants from around Queensland.

Participants of the pilot course at AIMS, March 2021. Image by TAS

If you would like to find out more about our work, or provide feedback on where you see the key risks and opportunities for the autonomous systems industry in Australia, please contact us as NASFP@tasdcrc.com.au