The Commission has recently released a discussion paper on Regulatory options to assure automated vehicle safety in Australia, outlining four regulatory options to govern the safety of driverless cars and other autonomous vehicles in Australia:
- Continuing the current approach
- Self-certification
- Pre-market approval
- Accreditation.
Automated vehicles that do not require human driver input into the driving task for at least part of the journey are expected to arrive on our roads from around 2020. Currently there is no explicit regulation covering these automated driving functions. Manufacturers are aiming to ensure automated driving functionality improves road safety, but this technology may also create safety risks for road users. The purpose of this paper is to seek feedback on:
- whether there is a need for explicit regulation of automated driving functions, above existing transport and consumer law
- if there is a need for regulation, what form this should take.
We are seeking feedback from governments, road safety experts, automated vehicle manufacturers, technology providers, insurers and other stakeholders on these questions. This paper examines:
- how safety of automated vehicle functions should be assessed
- the options for a safety assurance system
- the criteria that should be used to decide among those options
- institutional arrangements, road access and compliance.
Based on the feedback we receive, we will make recommendations to transport ministers in November 2017 on the preferred approach and the next steps to implement any required changes to legislation.
Australian governments have started work to remove legislative barriers to increasingly automated road vehicles. These barriers relate primarily to road traffic laws that implicitly require a human driver. Without further action, once these barriers have been removed, governments would have no regulatory mechanism to proactively ensure automated driving technologies are safe.
Automated driving technologies are progressively undertaking more of the driving task, and it is likely this technology will improve road safety, mobility, productivity and environmental outcomes. However, the technology is highly innovative and diverse and requires further testing and evaluation. From a regulatory perspective, there are four key issues:
- Should governments have a role assuring the safety of automated vehicles?
- What are our measures of safety, and what is the level of safety required?
- How does a safety assurance system balance safety outcomes with innovation, certainty and regulatory efficiency?
- Where does a safety assurance system fit within the existing regulatory framework for road transport, and how does it interact with existing laws?
In November 2016 the Transport and Infrastructure Council directed the National Transport Commission (NTC) to develop a national performance-based assurance regime designed to ensure the safe operation of automated vehicles. This will form a key component of an end-to-end regulatory framework to support the safe commercial operation of automated vehicles. Based on the feedback to this discussion paper, the NTC will recommend a preferred approach to ministers in November 2017, along with the next steps on regulatory reforms to support this approach.
In the absence of agreed Australian or international standards specific to automated vehicle technologies, governments need to consider the uncertain safety outcomes associated with different applications of automated driving, and whether the safety risk justifies additional government oversight and regulatory intervention. In Australia this type of oversight would be in addition to existing general consumer and product liability laws as well as extensive regulation covering vehicle standards and vehicle operation.
As the performance of the vehicle technology becomes increasingly safety-critical, new regulatory approaches may be needed to ensure initial and ongoing safety. Such approaches will need to cover all potential technology providers, from traditional automotive manufacturers to companies and individuals developing after-market devices to modify existing vehicles.
There is a risk that, without a national and coordinated response to automated vehicle reform, Australia’s complex regulatory framework will result in inconsistent regulation or over-regulation of automated vehicles across states and territories.
Regulatory options for safety assurance of automated vehicle functions
The NTC has developed four regulatory options for consultation for the safety assurance of automated vehicle functions. These are based on our assessment of the current regulatory framework and a review of safety literature and international developments. The four options are:
1. Continue current approach – no additional regulatory oversight, with an emphasis on existing safeguards in Australian Consumer Law and road transport laws.
2. Self-certification – manufacturers make a statement of compliance against high-level safety criteria developed by government. This could be supported by a primary safety duty to provide safe automated vehicles.
3. Pre-market approval – automated driving systems are certified by a government agency as meeting minimum prescribed technical standards prior to market entry.
4. Accreditation – accreditation agency accredits an automated driving system entity. The accredited party demonstrates it has identified and managed safety risks to a legal standard of care.
We are seeking feedback on these regulatory options, recognising that the regulatory solution may draw upon elements across these options. Stakeholders are also welcome to propose new regulatory options. ...
In many ways, the regulatory options reflect the risk appetite of the community and how the optimum role of government is perceived and understood by the community. In broad terms, the greater the risk appetite, the less we need explicit regulation, or a proactive role for governments to ensure automated vehicle safety.
In line with developments in other countries, the NTC proposes that the safety risks are sufficiently high or unknown to warrant some level of regulatory oversight and government involvement in the safety assurance system. ...
We have proposed eight assessment criteria against which the regulatory options for the safety assurance system have been evaluated. ...
Proposed assessment criteria for the design of the safety assurance system
1. Safety
- The model should support automated vehicle safety, including the ongoing safety over the full lifespan of the vehicle.
- The model should provide certainty about who is responsible for testing, validating and managing safety risks.
2. Innovation, flexibility and responsiveness
- The model should be technology-neutral and allow innovative solutions
- The model should allow government to respond and adapt to the changing market and evolving technology.
3. Accountability and probity
- The model should ensure the decision-making process is transparent, accountable and, where appropriate, appealable.
- There should always be an entity (whether an individual or a corporation) that is legally accountable for the automated driving system.
4. Regulatory efficiency
- The assurance process should be as efficient as possible and result in the least cost for industry and government, proportionate to the risk.
- The process of assurance should minimise structural, organisational and regulatory change necessary to implement the model.
5. International and domestic consistency
- The model should support a single national approach, or state-based approaches that are nationally consistent.
- The model should be adaptable if and when there is international consistency. International approval processes and standards should be recognisable.
6. Safe operational design domain
- The model should be able to take into consideration the operational design domain of an automated driving system.
7. Other policy objectives
- The model should be able to support non-safety policy objectives including cybersecurity, traffic management, environmental protection and the provision of data for enforcement or insurance purposes.
8. Timeliness
- The model should be able to be implemented and operational when the technology is ready.
Our initial assessment of the regulatory options suggests there are significant disadvantages associated with not developing a safety assurance system and continuing with the current approach (see Table 2). This is primarily because the ADRs do not have regard to automated driving technologies. Furthermore, existing safeguards, including vehicle recall powers, are focused on the technical integrity of the vehicle and do not consider environmental or human performance safety factors. This may lead to road safety risks, particularly in relation to vehicle modifications and after-market fitment.
Self-certification is a light-touch approach that, like the ‘continue current approach’ option, relies on existing safeguards but could introduce voluntary or mandatory compliance with automated vehicle safety principles and criteria. Showing compliance with these criteria would allow automated driving system entities to demonstrate to government that their vehicles are safe and therefore suitable to be registered under state and territory laws. Self-certification could be supported by a legislated primary safety duty for manufacturers, suppliers and automated driving system entities to provide safe automated vehicles.
Pre-market approval possibly provides the highest certainty for government and consumers that automated vehicles will be safe. However, this option is also regulation- and resource intensive and could stifle safety-related innovation if testing standards and procedures do not keep pace with technology changes.
Accreditation provides a comprehensive, risk-based and proven framework within which safety can be regulated. It focuses on outcomes, risk management and continuing improvements to safety. The accreditation model has demonstrated safety benefits in other high-risk industries including mining, rail and aviation. However, accreditation would involve a major reform of road safety, includes substantial set up costs and is not an approach that other countries are known to be exploring. ...The paper comments
Implementation issues
A number of key issues need to be considered in implementing an approach to the safety assurance of automated vehicle functions, in particular:
- how to evaluate and validate safety
- institutional arrangements to support the approach
- how to manage access to the road network
- how to ensure compliance with the requirements of the approach selected.
How to evaluate and validate safety
Evaluation and validation of automated vehicle safety is a foundation issue for the development of the safety assurance system. The proportionate and appropriate role of a government agency to test the safety claims made by a manufacturer or technology provider will largely depend on the regulatory model adopted in Australia.
We are seeking feedback on whether safety should be defined and measured according to the rate of technical failure and incidents that result in harm to people, or be based on an agreed metric of safety such as crash rates.
The NTC is proposing that the onus be placed on the automated driving system entity to demonstrate the methods they have adopted to identify and manage safety risks.
Institutional arrangements to support the approach
If there is a role for government in safety assurance for automated vehicle functions? Which government body will have that role? Responsibility for motor vehicle safety regulation is currently shared between the Commonwealth and the states and territories. The current mix of regulatory responsibilities adds complexity to the development of a safety assurance system and the potential institutional arrangements to oversee the safety assurance system. We are seeking feedback on institutional arrangements, including the types of government entities that could support a safety assurance system.
Institutional arrangements are heavily dependent on the safety assurance option chosen, therefore the NTC is proposing that institutional models are further developed after a regulatory option has been agreed.
How to manage access to the road network
For the foreseeable future, automated vehicle functionality will be limited to parts of the road network (for example, only sealed roads). This raises the question of the role of registration authorities and road managers (including local governments) in managing access to the road as part of the safety assurance system. We are seeking feedback on the role of road managers and whether registration authorities and road managers should authorise automated vehicle access to their road network in addition to safety assurance processes.
The NTC is proposing that a national approach should be adopted that incorporates automated vehicle registration and network access into the safety assurance process. However, access issues should be further explored once a regulatory model has been agreed.
How to ensure compliance
How do governments ensure compliance with any safety assurance system? We are seeking feedback on how to ensure compliance – including what regulation (if any) is needed to ensure automated driving system entities and other parties comply with safety obligations.
We suggest that compliance could be ensured through a primary safety duty for parties to provide safe automated vehicles with associated penalties and/or specific sanctions and penalties for the automated driving system entity.
The best way to ensure compliance will depend significantly on the regulatory model agreed. Sanctions and penalties in road traffic laws could also cover automated driving system entities through the NTC reforms to driver legislation.
Consultation questions
1. Should government have a role in assessing the safety of automated vehicles or can industry and the existing regulatory framework manage this? What do you think the role of government should be in the safety assurance of automated vehicles?
2. Should governments be aiming for a safety outcome that is as safe as, or significantly safer than, conventional vehicles and drivers? If so, what metrics or approach should be used?
3. Should the onus be placed on the automated driving system entity to demonstrate the methods they have adopted to identify and mitigate safety risks?
4. Are the proposed assessment criteria sufficient to decide on the best safety assurance option? If not, what other assessment criteria should be used for the design of the safety assurance system?
5. Should governments adopt a transitional approach to the development of a safety assurance system? If so, how would this work?
6. Is continuing the current approach to regulating vehicle safety the best option for the safety assurance of automated vehicle functions? If so, why?
7. Is self-certification the best approach to regulating automated vehicle safety? If so, should this approach be voluntary or mandatory? Should self-certification be supported by a primary safety duty to ensure automated vehicle safety?
8. Is pre-market approval the best approach to regulating automated vehicle safety? If so, what regulatory option would be the most effective to support pre-market approval?
9. Is accreditation the best approach to regulating automated vehicle safety? If so, why?
10. Based on the option for safety assurance of automated vehicle functions, what institutional arrangements should support this option? Why?
11. How should governments manage access to the road network by automated vehicles? Do you agree with a national approach that does not require additional approval by a registration authority or road manager?
12. How should governments ensure compliance with the safety assurance system?
