Objectives

Three objectives were set for this project:

• Provide the Victorian Government and TAC direct and ongoing access to AV perception to further enhance understanding of how AVs perceive the road environment.
• Provide an intersection black box function to allow detailed assessment of all crashes, near misses & other interactions.
• At a high priority intersection crash hotspot, demonstrate provision to early Connected Automated Vehicles (CAVs) (and C-ITS equipped vehicles) of full situational awareness as if every road user and vehicle were equipped.

The objectives sought to support the four objectives established for the trial program of which this project was part:

• Trial CAV technologies that will inform and support Victoria’s readiness for CAVs vehicles optimise safety benefits.
• Evaluate how CAVs improve road safety trauma outcomes to inform future investment and planning.
• Generate knowledge that will inform the Victorian Government’s and TAC’s future planning for physical and digital infrastructure.
• Spur early deployment of CAVs that will reduce road trauma by providing a testing environment for on-road development of CAVs.

Research questions that the project aims to address

The three research questions adopted for evaluation of the project directly align with the objectives set for the project:

• Has Omni-Aware demonstrated the ability to provide information to inform a detailed assessment of all crashes, near misses and other interactions at the subject intersection?
• Has Omni-Aware demonstrated the ability to provide content for a C-ITS message with the necessary detail, confidence and timeliness to permit C-ITS crash avoidance functions?
• Has the project provided learnings about how AVs perceive the road environment?

A number of relevant sub-questions were explored for each of these research questions, such as:

• What types and quantities of crashes were observed?
• What detail was able to be provided for crashes?
• What types and quantities of near misses were observed?
• What detail was able to be provided for near misses?
• What types and quantities of other interactions were observed?
• What detail was able to be provided for other interactions?
• In a validation exercise, what (if any) types of crashes, near misses and other interactions were not (fully) observed?
• Does performance suffer for any observed road user types?
• Does performance suffer for any observed environmental conditions?
• Does performance suffer for any observed traffic conditions?
• How transferable does performance appear to be to intersection / location conditions different to the test site?

Lessons learned to inform the conduct of trials

• The use of off-road testing permitted the consideration of use cases best examined in a controlled environment rather than on public roads and provided visibility of performance prior to installation at the on-road site.
• The lengthy trial period at the on-road site (six months) allowed for a substantial exploration of both real-world performance and the identification and response to real-world operational challenges.

Lessons learned that inform future technology deployments

• Substantial adaptation was required to make technology developed for Automated Vehicles (AVs) suitable for on-road use, however the available performance once this is undertaken is suggestive of value in doing so.
• During the off-road testing, the technology demonstrated strong potential, resulting from very strong spatial accuracy (within 10cm). the high fidelity of LiDAR point cloud data (clear surfaces of objects such as pedestrians and vehicles) and performance did not appear to degrade significantly in any observed environmental conditions (wind, rain).
• Also during the off-road testing, the Omni-Aware solution demonstrated strong performance in generating and communicating C-ITS warning functions when paired with the Q-Free C-ITS equipment.  Hazard identification by Omni-Aware was rapid (<0.2 seconds).  Display of warning messages within vehicles occurred in 3-5 seconds with opportunities to further reduce this identified.
• The results of the on-road testing both confirmed the strong performance potential of the technology and the myriad of challenges that exist when seeking to achieve achieving high performance results robustly in real-world conditions over extended periods.
• The provision of detailed information of events of interest was demonstrated and this information was available for any event of interest regardless of whether crash, near miss or other event.  The capability was also proven to apply highly configurable (and richly definable) rules-based queries to successfully identify different types of events of interest.
• The use of Surrogate Safety Measures (SSMs) for intersection crash risk assessment was partly demonstrated and further software development would be required for these results to reflect the high levels of performance available in the underlying technology.
• The development of LiDAR sensors is ongoing and sensors suitable for long-term and economical installation in roadside environments should be available in coming years.

Links to other supporting project reports

Victorian Government website for the trial program

Victorian Government press release for the project

Infrastructure Magazine news article on the project