Safely deploying automated vehicles on Australian roads

For the past five years a world-leading piece of autonomous vehicle research and development has been taking place in Queensland, first on a test track, then moving on to public roads in urban and regional environments. The final reports below, and the accompanying summaries, are the outcomes from the Cooperative and Highly Automated Driving Safety Study project.

A prototype, level 4, highly automated vehicle, ZOE2, has been trialled to develop recommendations regarding how to make connected and automated vehicles (CAV) and associated infrastructure more resilient to road safety risks. The vehicle was also used to increase public awareness about CAVs. Final reports for the project are available here.

In Queensland, the Department of Transport and Main Roads’ (TMR) Cooperative and Automated Vehicle Initiative (CAVI) is helping prepare the state’s roads, and road users, for new vehicle technologies, such as CAV, and their safety and mobility benefits. CAVI is delivering several projects (displayed below) under the Cooperative and Highly Automated Driving (CHAD) pilot, for which ZOE1 and ZOE2 research prototype vehicles have been rolling laboratories.

Project articles and final reports in the graphic above include:

• Project 1-007: How automated vehicles will interact with road infrastructure | Final report: Autonomous vehicles and Australian roads: Are they ready for each other?
• Project 1-021: HD maps for automated driving – literature review | Final report: HD mapping Australia’s CAV future
• Project 1-048: Using crowdsourced data to improve road management
• Project 1-059: Safety risk evaluation of the remote operation of HAVs
• Project 1-075: Expanding Operating Design Domain of automated vehicles

The CHAD pilot is being delivered by TMR in partnership with Queensland University of Technology (QUT), and QUT’s Centre for Accident Research & Road Safety (CARRS-Q).

Contents

This particular project, the CHAD Safety Study, was divided into four separate work packages:

• Work Package 1: Driving task transition in automated vehicles
• Work Package 2: Cooperative and Intelligent Transport Systems (C-ITS) and automated vehicle (AV) integration, and benefits evaluation
• Work Package 3: Australian safety challenges for CAVs in the dynamic road environment
• Work Package 4: CAV public awareness and demonstrations

Click the links above to go directly to the summary, final report and fact sheet for Work Packages 1, 2, 3, and 4.

Work Package 1: Driving Task Transition in Automated Vehicles

In this work package, members of the public sat in the driver’s seat of ZOE2 and were taken on a 30-minute drive around a test track, with the purpose of assessing their perception and reactions of the task of the takeover and handover processes in ZOE2 transitioning between automated and manual driving modes.

Data collection

The participants filled out pre- and post-drive questionnaires on their experiences of the driving mode changes. The questions investigated the participants’:

• Acceptance of the automated vehicle (AV)
• Perceived risk of AV manoeuvres, and other experiences related to AV use
• Thoughts on the speed and safety of the transitions between automated and manual driving modes

Additionally, eye-tracking data from the Driver Monitoring System was analysed and compiled, with a view toward recommending good practice guidelines for the human machine interface.

Conclusions and recommendations

The three questions investigated in this work package were:

1. How does experiencing SAE level 3/4 automation affect trustworthiness and acceptability of Automated Vehicles (AV)?
2. How do regular drivers/operators behave in an AV and what are their concerns?
3. How does level 4 automation affect driver’s situational awareness and their reaction time?

Amongst the findings from the public participants were:

• greater acceptance after experiencing the AV, including perceiving it as less risky, with largest changes in pre versus post drive responses were for stressfulness (less stress) and comfort performing a non-driving task (more comfortable).
• the most common concerns raised related to the vehicle behaviour (braking, cornering, and speeds which were described as harsh, jerky, slow and unnatural).

The research team posits that a minimum of 5-6 seconds should be allotted for safe takeover from automated to manual mode to account for the majority of drivers, even at their slowest. However, this only applies:

• if operational requirements prohibit engagement in non-driving tasks, and
• takeover request warnings are only provided through audio and visual means (no haptic alerts via seat and/or seatbelt).

In terms of the vehicle’s performance the first five seconds of takeover showed greater driving performance instability with more deviations from the planned trajectory, but on average, the vehicle remained within 1.75 metres of the trajectory in all conditions, which is within the standard width of an Australian urban lane (3.5m).

WP1 final report and factsheet

More information on the study design, data and analyses, conclusions and recommendations is available in the final report for Work Package 1. Download your copy of the final report and associated fact sheet by clicking the buttons below.

WP1 FINAL REPORT WP 1 FACT SHEET

Work Package 2: C-ITS/AV integration, and benefits evaluation

This work package sought to integrate over-the-air C-ITS messages in AV (ZOE2) and evaluate safety benefits. The study compared simulation results of an AV with C-ITS messages (Connected AV (CAV)) versus an AV without C-ITS messages. Two CITS messages were developed as part of Ipswich Connected Vehicle Pilot, were integrated in ZOE2 and safety benefit of AV was evaluated with and without CITS messages.

Two use cases were assessed:

Data pertaining to AV behaviour was collected at the test track utilising ZOE2, while data pertaining to behaviours of other human driven vehicles were collected on public road for the use case A.

Conclusions and recommendations

It is recommended to consider extending the scope of this study to fully understand the impact of the latest industry standards IEEE 2846,2022. In the meantime, the findings contained in this report should be considered as true at the time of AV behaviour data collection, that is 2021.

Use case A: Occluded pedestrian at a traffic light intersection

• Both AV and CAV manage to stop before colliding with the pedestrian. C-ITS messages do not provide added benefit.
• Rear-end crash likelihood improved by 20 per cent in CAVs; the crash severity also reduced. It is important to note this benefit may likely reduce or diminish if the latest standard, published in 2022, for the AV decision making were to be implemented in ZOE2.

Use case B: Lane closer due to roadworks forcing vehicles to take alternate lane

• CAV has significantly increased likelihood of successful lane change manoeuvre, resulting in less need for human driver takeover.
• No change observed in the likelihood of crashes and near crashes on roadworks zone.

Overall

Over-the-air C-ITS message quality has no impact on reducing the likelihood of crashes and near crashes for both use cases.

WP2 final report and factsheet

More information on the study design, data and analyses, conclusions and recommendations is available in the final report for Work Package 2. Download your copy of the final report and associated fact sheet by clicking the buttons below.

WP2 FINAL REPORT WP 2 FACT SHEET

Work Package 3: Australian safety challenges for CAVs in the dynamic road environment

This work package focused on identifying unique Australian scenarios which a CAV must deal with safely. Following consultation with international experts, three scenarios were identified as unique to Australia. These are;

1. Road trains / long vehicles;
2. Wildlife (kangaroos / wallabies); and
3. Single lane, two-way rural roads

Following identification of three unique Australian scenarios, an extensive data collection exercise was undertaken utilising ZOE2 (about 250 hours of driving). In addition, data gathered during experiments at RACQ Mt Cotton test tracks, and dynamic demonstrations (part of WP4) in locations like Shailer Park, Bundamba, Bundaberg, and Mount Isa were also included.

Key findings

• Machine learning techniques along with sensor fusion techniques can accurately detect and classify road trains.
• Rule based decision-making mechanisms utilising accurate high-definition maps of the road are likely to provide safe interaction between CAVs and road trains.
• Producing a naturalistic driving behaviour is difficult and needs to consider local road user behaviours. These behaviours could be aggressive or extra polite.
• Overhung loads, such as ladder in the back of a utility vehicle, are difficult to accurately detect.
• Current AV trial regime may not be sustainable in three to four years horizon when mass produced AVs are likely to enter Australian market.

Key recommendations

• Regulators should limit the Operational Design Domain (ODD) of the CAV if supporting technologies for safe operation (such as C-ITS messages, accurate high-definition maps) are not available.
• Regulators need to seek safety case from the ADSE (Automated Driving System Entity) demonstrating AV’s safety remain ‘so far as is reasonably practicable’ while interacting with road train/long vehicles on Australian roads.
• Expand the study scope of future projects to prototype and confirm mitigation strategies for the wildlife and single lane two-way rural roads scenarios.
• Ensure AV national law, associated regulatory framework and national AV regulator are in place by 2026.
• Continue supporting National Transport Commission in developing a nationally consistent approach to on-road enforcement for AVs.

WP3 final report

More information on the study design, data and analyses, conclusions and recommendations is available in the final report for Work Package 3 providing guidance for CAV on-road deployment. Download your copy of the final report by clicking the button below.

For this work, the researchers:

• Conducted a qualitative, in-depth investigation surrounding public perceptions of CAVs with participants residing in urban (Brisbane) and regional (Toowoomba) areas of Queensland. There were 11 focus groups held, comprising 43 individuals (18 males, 25 females) who were licensed drivers.
• These groups were asked about their beliefs about CAVs, and their advantages, disadvantages as Level 4 CAV private vehicles, and Level 5 CAV vehicles, in both private and shared use.

A further focus group was held, in which public awareness messaging derived from the earlier focus groups was tested. In the final component of this pre-CAV experience, the messaging further refined by the findings from the focus groups was tested via a questionnaire.

Public demonstrations

ZOE2 was displayed to the public at a Gold Coast show, in stationary not driving mode, with again a questionnaire completed.

Then the research switched to the car in motion, with some members of the Bundaberg general public taken on an approximately 20-minute course on public roads. Not only were there thoughts on the drive collected, but also on the in-car messaging of the vehicle’s automated/manual modes and the interchange between them.

Amongst the results from the public demonstration were:

• the messaging and the demonstrations (both static and dynamic) were well-received by the public
• both demonstrations reported improved knowledge about automated vehicles (AV), as well as positive increases in acceptance measures comprising attitudes towards, intentions and willingness to use, as well as trust in, AVs
• 96.4% and 98.3% of intervention participants from the static and dynamic demonstrations, respectively, reported they would recommend others to attend the demonstration they had attended to increase their knowledge about automated vehicles.

WP4 final report and factsheet

More information on the aims, method, and key findings of Work Package 4 are available in its final report. Download your copy of that final report and associated fact sheet by clicking the buttons below.

WP4 FINAL REPORT WP4 FACT SHEET