DiDi Chuxing

Submission on

NTC Discussion Paper 'Changing driving laws to support automated vehicles’

24 November 2017

Introduction

This submission is provided by DiDi Chuxing (DiDi) to the National Commission (NTC) in response to its discussion paper 'Changing driving laws to support automated vehicles' (Discussion Paper). The Discussion Paper sets out a series of potential reforms and strategies to set a long term strategic direction for the introduction of automated vehicles in Australia.

DiDi congratulates the NTC on seeking feedback from industry to address the legal and regulatory barriers to the introduction of automated vehicles and the preparation it is taking for the transformed transport system.

DiDi shares the NTC's vision that driverless cars will have significant benefits for society. As an international participant in the global transportation industry, DiDi welcomes the opportunity to provide its comments on the regulation of automated vehicles in Australia.

DiDi Chuxing

Who is DiDi Chuxing?

DiDi is a world leader in mobile transportation platforms, providing consumers with comprehensive transportation services, such as :

(a) DiDi taxi;

(b) DiDi premier;

(c) DiDi express;

(d) DiDi luxury car;

(e) DiDi bus;

(f) DiDi minibus;

(g) DiDi designated driving;

(h) DiDi car rental;

(i) DiDi enterprise solutions; and

(j) bicycle sharing through strategic investment with .

DiDi operates in over 400 cities in China, catering for more than 440 million users. The daily orders received by DiDi have exceeded 25 million. Additionally, DiDi uses artificial intelligence technology to assist cities to establish smart traffic solutions. In August 2016, DiDi acquired China.

DiDi strives to cooperate with a broad range of communities and partners to solve transportation, environmental and employment challenges through the use of data-driven deep learning technologies as well as improve user experiences, promote social values and build an open, efficient and sustainable new mobile transportation ecosystem. In 2016, DiDi was included on the Fortune "Change the World" top 50 list. In the same year, DiDi was named as one of the 50 Smartest Companies by the MIT Technology Review. In 2015, DiDi was named as a Global Growth Company in Davos.

Page 2

While DiDi does not have an established presence in Australia, it has a global concern about the appropriate and consistent regulation of autonomous vehicles.

DiDi believes that transportation is the very artery of urban vitality and a transportation system that is plugged in to new data technologies is essential to initiatives around the world.

DiDi's international vision now extends to building the best-in-class international research network, advancing the global transportation revolution by leveraging innovative resources. In March 2017, DiDi officially launched DiDi Labs in Mountain View, California.

DiDi Labs primary focus is on AI-based security and intelligent driving technologies. However, it also works in tandem with the broader DiDi research network to advance DiDi's global strategy, apply research findings to products and services, and help cities develop smart transportation infrastructure.

For further information regarding DiDi please visit http://www.DiDichuxing.com/en/.

DiDi and automated vehicles

Driverless technology promises to disrupt the automotive industries, and fundamentally change the approach to car ownership and use.

DiDi is aware that Australia is a 'driving country'. Research indicates that private car ownership has steadily increased by 43% since 2000, with over 90% of households having access to a car.1 This prolific trend towards privately owned vehicles is a significant contributor to the pollution and congestion experienced in Australian cities.

DiDi recognizes the Australian government's substantial efforts to promote legislation to allow for the development of autonomous vehicles. DiDi welcomes the opportunity to provide submissions on its perspective on the proposed Australian legislation to the NTC. In order to derive the benefits of automated vehicles private ownership of vehicles must change. If driverless vehicles continue to be privately owned there is, in fact, the risk of further congestion and pollution due to more cars on the roads. Particularly if privately owned autonomous vehicles are available 'on demand'.

Ride share services bridge the gap between public transport and private car use and ownership. They have grown significantly around the world over the past decade and should continue to grow exponentially. Given the significant role ride share services will play in the future of the global transport industry DiDi offers its views on the regulation of the industry for the NTC's consideration.

DiDi's submissions on each of the questions posed by the NTC are set out below.

As an overarching concern, DiDi believes that Australia's continued co-operation with the international automotive industry is essential. DiDi wishes to ensure that automated vehicles and ride share services are developed, used and regulated in a globally harmonious manner. While regard needs to be had to unique local conditions, any regulation in Australia should wherever possible be compatible with current and future international regulations and standards.

Similarly, DiDi supports a nationally consistent approach being taken in Australia to automated vehicle regulation, vehicle-to-vehicle technology and vehicle-to-infrastructure in order to achieve the desired effect of enhanced road safety and allow for passage across State boundaries. DiDi holds significant concerns that a failure to do so (in Australia or elsewhere) will increase red tape and reduce innovation.

1 http://www.theaustralian.com.au/news/inquirer/if-car-ownership-keeps-growing-australians-will-be-outnumbered/news- story/92f65fe440421d9ede7155e895db723c

Page 3

From the perspective of technical development of unmanned driving, the industry generally believes that the time when Level 4 and Level 5 ADS are maturely applied will be between 2020 and 2025. DiDi takes the view that, at present, it is still too early to establish a complete set of laws in relation to unmanned driving.

DiDi acknowledges the efforts of the NTC in engaging in consultation with industry about automated vehicle trials and in developing the National Guidelines for automated vehicle trials released in May 2017.

DiDi supports these efforts and recommends that government should first provide support for trials of unmanned driving technologies to assist with the further development of technologies in this space.

To facilitate this process, DiDi suggests that the Australian government should divide the legislation on autonomous (unmanned) vehciles into two stages, a testing stage and a deployment stage, giving priority to laws that allow autonomous vehicles to carry out trials on public roads in Australia.

DiDi makes the following recommendations regarding the public road tests for autonomous (unmanned) vehciles:

(a) autonomous vehicles should be allowed to carry out trials on public roads throughout Australia;

(b) eligible companies should be entitled to apply for a 'testing license', including autonomous vehicle R&D companies in Australia and internationally;

(c) companies applying for testing licenses should be required to provide a certain amount of security funds or to establish a self-insurance policy;

(d) companies applying for testing licenses should be required to provide a security management plan relating to proposed autonomous vehicle trials;

(e) in the instance of an accident involving an autonomous vehicle colliding with an ordinary vehicle or pedestrian in circumstances where the accident is caused by the failure of the ordinary vehicle or pedestrian to comply with traffic laws, DiDi suggests that there ought to be circumstances in which the ADSE should only be held proportionately liable to the extent that the accident is caused by the failure of the autonomous vehicle. In the example of a manufacturer of the ADS, examples of limits on liability for the NTC to consider are outlined at paragraph 2.8 below;

(f) as technology advances autonomous vehicle research and development companies should be allowed to decide whether to have fall back users in their autonomous vehicles during trials provided that the required safety assurances are given; and

(g) autonomous driving companies should be allowed to conduct high-definition map data collection for the purpose of testing. Submissions on questions

2. Do you agree that reform to existing driving laws is required to:

(i) allow an ADS to perform the dynamic driving task when it is engaged?

(ii) ensure a legal entity (ADSE) is responsible for the actions of the vehicle when the ADS is engaged?

Question 1(i):

Page 4

2.1 Yes. DiDi applauds the Australian government on its efforts to reform Australian driving laws to adapt to the rapid growth of ADS.

2.2 DiDi agrees that the law should allow an ADS to perform the dynamic driving task when it is engaged.

Question 1(ii):

2.3 Yes. DiDi agrees that the law should ensure a legal entity (the ADSE) is responsible for the actions of the vehicle when the ADS is engaged

2.4 DiDi supports the amendment of legislation to clarify the potential liability for vehicles in autonomous mode, but notes that it is important that any legislation operates in a flexible manner to reflect the ever evolving autonomous vehicle developments and the benefits of autonomous vehicle technology.

2.5 A key issue in relation to the above questions is the definition of the ADSE. DiDi notes that driverless technology is a comprehensive technology. This means that an ADS includes, but is not limited to:

(a) hardware (such as vehicles, sensors and computing devices);

(b) software (such as perception, decision making and control algorithms);

(c) high definition maps;

(d) human machine interface(HMI) systems; and

(e) V2X systems.

2.6 It is therefore important to draw attention to the fact that the providers involved in the production of an ADS includes various entities including but not limited to:

(a) OEMs;

(b) tier one suppliers;

(c) sensor suppliers;

(d) algorithmic companies;

(e) high-definition mapping companies; and

(f) V2X companies.

2.7 DiDi has formed the view that the ADSE should refer to all of the above suppliers involved in the production of the ADS instead of simply the prover of the vehicle or service. Where an autonomous vehicle, with its ADS engaged, causes an accident due to an ADS error, DiDi suggests that liability should be determined on the basis of the cause of the accident. For example, a failure in sensors should be attributed to the entity responsible for the sensors in the vehicle.

2.8 Depending on the fault or error, it may be necessary to limit or apportion liability. By way of further example, in respect of the liability of the manufacturer of the ADS, DiDi recommends that this should be subject to certain limitations of liability, such as where:

(a) the autonomous vehicle satisfies the minimum safety standards for each level of automation. Different regulations ought to apply for different levels of automated vehicles;

(b) third party modifications to autonomous vehicles or failure by vehicle owners to maintain vehicles to manufacturer's specifications impact the vehicle's ability to perform the dynamic driving task in compliance with applicable laws;

Page 5

(c) there is a failure of communications between an autonomous vehicle and fixed infrastructure. In this instance, liability may be apportioned between the owner and operator of that fixed infrastructure (which may be government or a private sector entity) having regard to the cause of the failure;

(d) human intervention is or is not present when it ought to be. This may be either by negligent intervention or a failure for a 'fallback ready user' to intervene when required. This limitation would only apply where vehicles are not highly or fully autonomous; and

(e) the ADS is hacked or disrupted by a third party.

2.9 DiDi recommends that a national or unified approach is essential. A patchwork regulatory approach could potentially lead to greater safety risks because, for example, of the potential risk that vehicles will fail to properly interact. Further, inconsistent regulation between different territories would present a barrier to the introduction of autonomous vehicles, restrict innovation and de-incentivise entry for international manufactures and operators.

2.10 In the same vein, the success of the autonomous vehicle legislative reform is interconnected with the need for a unified approach to rideshare regulation across Australia. Currently, there are differences in the ride share requirements in each Australian jurisdiction. Cross boarder trips will become increasingly more difficult as the move towards automated vehicles continues unless a cohesive approach to ride share services (and automated vehicles) is taken.

2.11 DiDi recommends that, to the extent possible, legislation should recognise incentives for continuing innovation and deployment of autonomous vehicles (while recognising the need to incentivise safety).

3. Do you agree that if the ADS is engaged, legislation should provide that the ADS is in control of the vehicle at conditional, high and full levels of automation? If not, do you think a human in the vehicle should be considered in control of the vehicle, and at what levels?

3.1 DiDi agrees with Option 2 identified by the NTC that for highly automated (level 4) and fully automated (level 5) vehicles it is appropriate to recognise that the ADS is in control of the vehicle when engaged, subject to the concerns and matters for consideration expressed by DiDi at paragraph 2.8 above.

3.2 DiDi is of the view that when a level 3 ADS is engaged that a human in vehicle should be considered in control of the vehicle.

4. Do you agree that the proper control offence should not apply to the ADS, provided there are appropriate ways to hold the ADSE to account for the proper operation of its ADS?

4.1 At present, DiDi has not formed a view on this issue.

5. Do you agree that if a safety assurance system is approved that requires an ADSE to identify itself, the identified ADSE should be responsible for the actions of the vehicle while the ADS is engaged? If the ADSE is not identified through the safety assurance system, how should the responsible entity be identified in legislation?

5.1 DiDi agrees that the ADSE should be identified under any proposed safety assurance system.

6. Do you agree that when the ADS is engaged:

Page 6

(i) an ADSE should be responsible for compliance with dynamic driving task obligations?

(ii) obligations that are part of the dynamic driving task that the ADS cannot perform should be modified where appropriate, or the ADS exempted from the obligation?

(iii) an ADSE should not be responsible for existing driver duties and obligations that are not part of the dynamic driving task?

6.1 DiDi takes the view that an ADSE should be responsible for driving compliance when an ADS is engaged in the dynamic driving task.

6.2 When an ADS is unable to perform certain components of the dynamic driving task, it should refuse to perform or suspend its dynamic driving task and hand over the driving control to a human driver.

6.3 The ADSE should not be liable for accidents which are caused by the failure of the human driver to take over the dynamic driving task promptly. The ADSE should not be liable for the existing driver duties and obligations.

6.4 Industry will need certainty as to the scope of the 'dynamic driving task' and how it differs at each level of automation.

6.5 Further, the legislation will need to recognise that different types of ADS' will have different capabilities in terms of the dynamic driving task. That is, automated vehicles may be capable of the driving task on highways and in specified urban environments, but some may not have the capability to operate in other areas.

7. How should legislation recognise an ADS and an ADSE? In assessing the options in section 5.6, please consider the following factors:

(i) legislative efficiency

(ii) timeliness

(iii) impact on compliance and enforcement

(iv) impacts on other schemes such as compulsory third-party insurance

Are there other options that you prefer? Please provide details of how it would work.

7.1 As previously discussed, regarding the definitions of ADS and ADSE, ADS is a comprehensive technology. An autonomous vehicle consists of a variety of hardware and software as discussed above at paragraph 2.3.

7.2 DiDi recommends that an ADSE should be defined to include all entities contributing the an ADS as outlined at paragraph 2.6.

8. Do you agree that driver obligations need to be assessed to ensure there are no obligations that cannot be fulfilled if an ADS is in control? If gaps are identified, should other appropriate entities—such as fallback-ready users, other vehicle occupants, registered operators and operators—be made responsible for the obligation?

8.1 DiDi is of the view that it is presently not possible to evaluate whether an ADS can accomplish all of the dynamic driving tasks through a set of assessments. For now, DiDi is not aware of any ADS that that would fully satisfy level 5 full autonomy.

8.2 DiDi recommends that a set of evaluation criteria should be developed to test an ADS. However, this evaluation criteria should refer to the criteria and capabilities currently

Page 7

used to assess human drivers during license examinations and should not pursue the absolute safety of the driverless system in any external environment.

8.3 As discussed further in relation to question 9, DiDi takes the view that the fall back user in the vehicle should assume some responsibility before fully-fledged driverless technology is implemented.

9. Do you agree that obligations on a fallback-ready user of a vehicle with conditional automation, who will be required to take over driving if requested by the ADS should include:

(i) sufficient vigilance to acknowledge warnings and regain control of the vehicle without undue delay, when required?

(ii) holding the appropriate licence for the vehicle type?

(iii) complying with drug, alcohol and fatigue driver obligations?

Do you agree that the fallback-ready user should be allowed to perform secondary activities?

Fallback user

9.1 DiDi holds the view that the fall back user in an autonomous vehicle should be responsible for control of the vehicle when an ADS up to and including level 3 automation is engaged, as articulated above at 3.2.

9.2 This would include:

(a) being sufficiently vigilant to acknowledge warnings and regain control of the vehicle without undue delay when necessary;

(b) holding the appropriate driving license for the vehicle type; and

(c) complying with drug, alcohol and fatigue driver obligations.

Licenced user

9.3 DiDi supports the fallback user requiring an appropriate licence to operate a lower level automated vehicle (up to and including level 3 automation). This is because it anticipates times (although limited) that a human could be required to have control of, and drive, the vehicle.

9.4 For high and full automation (level 4 and 5) the requirement for a human 'driver' (or fallback user) should fall away. Accordingly, it will be unnecessary to restrict use of the automated vehicle to require a licenced passenger. However, DiDi notes that there may be circumstances in which a passenger in a vehicle capable of level 5 automation chooses to take over and drive. In this circumstance, DiDi recommends that they must be appropriately licensed.

9.5 It is anticipated that by the introduction of high and full vehicle automation the role of driver licensing be severely diminished or non-existent. It will be important for an alternative safety assurance system to apply.

Drug and alcohol offences

9.6 In relation to potential for a person under the influence of drugs and / or alcohol DiDi notes that a similar risk exists in taxis (i.e. a drunk passenger can the steering wheel or handbrake).

9.7 For up to level 3 automated vehicles, DiDi agrees that it is appropriate for the fallback user not to be intoxicated given they may need to drive the vehicle.

Page 8

9.8 DiDi suggests that for level 4 and above intoxicated passengers should be entitled to travel alone in the automated vehicle. Technically, under current rules this passenger would be considered 'in charge of' the vehicle. Changes will therefore be required to the legislation to reflect, for example, that the ADS is in control of the vehicle and not the human passenger. An intoxicated passenger that wrongfully choses to take control should also commit the offence of drink driving.

9.9 DiDi suggests that, in any event, it is perhaps premature to be expressing a view on this in terms of high and full automated vehicles given there will be significant further technological developments before this stage is reached. For example, technological advances may lead to the installation of systems in automated vehicle that will prevent an intoxicated person from taking control. For example, where a vehicle is capable of operating highly or fully autonomously in order for a driver to takeover they may need to pass a breathalyser test.

Fatigue driver obligations

9.10 In automated vehicles classified at lower levels of automation (up to level 3), DiDi supports the fallback user being required to remain awake should they be required to take control.

9.11 In conjunction, DiDi supports the use of, for example, in-car cameras to monitor the fallback user's drowsiness and for the system to sound an alert if the person is falling below certain levels. For instance, the ratio of blinks could be indicative of drowsiness.

9.12 For high and full levels of automated vehicles the obligation of a fall back user should not exist. Therefore, driver fatigue rules should not apply unless, for example, a 'passenger' in a fully autonomous vehicle elects to take over and drive.

10. Do you think it is necessary to impose readiness-to-drive obligations on humans who will take over driving when a vehicle with high automation that includes manual controls reaches the limit of its operational design domain?

10.1 DiDi agrees when a Level 4 ADS is engaged, it is necessary to impose readiness-to- drive obligations on humans who will take over the dynamic driving task when a vehicle with high automation that including manual controls reaches the limit of its operational design to carry out the dynamic driving task.

11. Do you agree that no readiness-to-drive obligations should be placed on passengers in dedicated automated vehicles (designed to be ‘driverless’)?

11.1 Yes, DiDi is wholly supportive of no readiness-to-drive obligations being imposed in dedicated automated vehicles.

11.2 Imposing an obligation on passengers in driverless cars to hold a licence or be able to satisfy certain 'readiness' obligations would defeat the reach and purpose of driverless cars.

11.3 A significant advantage of driverless cars is the possibility to provide a safe means of transport for those who could not otherwise operate a vehicle on their own, such as the elderly, minors, injured or persons under the influence.

11.4 Removing this benefit would:

(a) considerably diminish the societal benefits of driverless cars and limit their use;

(b) be inconsistent with the anticipated trend that licensing regimes will change and most persons will not be licenced to drive; and

Page 9

(c) potentially require, for example, ride share providers to engage a now redundant 'driver' who is ready to drive when transporting passengers who could not satisfy the ready to drive requirements.

12. Should exemptions from the drink- and drug-driving offences concerning starting a vehicle and being in charge of a vehicle be provided to a person who is starting, or who is a passenger in, a dedicated automated vehicle?

12.1 Yes. For the reasons outlined at DiDi's answer to question 10 above, an intended benefit of driverless cars is to safely provide transport to persons under the influence or who could not otherwise drive (under current Australian transport rules and regulations).

13. Should exemptions from the drink- and drug- driving offences concerning starting a vehicle and being in charge of a vehicle be provided to a person who is starting a vehicle with high or full automation that includes manual controls?

13.1 Yes. Where full automation is available DiDi considers that it should be available to be used to the full extent and passengers should be entitled to ride in the vehicle irrespective of whether or not they are under the influence of drugs or alcohol.

13.2 However, a passenger should be held responsible if they affect the operation of the autonomous vehicle while under the influence of alcohol or drugs.

13.3 As noted at paragraph 9.9 DiDi suggests that as technology advances methods may develop to ensure that a person under the influence of drugs or alcohol is prevented from taking over the autonomous vehicle. However, a person who seeks to do this while under the influence of drugs or alcohol should commit an offence.

14. How do you think road traffic penalties should apply to ADSEs?

14.1 DiDi is of the view that an ADSE should be held liable for any traffic accident caused due to a technical failure in the ADS. As for how the ADSE should be punished for a traffic accident, DiDi is yet to form an opinion.

15. Do you think obligations and penalties on ADSEs in the safety assurance system should complement, or be an alternative to, road traffic offences?

15.1 DiDi has not formed an opinion on this issue at this time.

Page 10