The use of vehicle automation has the potential to make the driving task easier and safer and should lead to opportunities for improved safety and productivity. The question facing road managers is: Are our roads ready for automated vehicles?
The foundation and underpinnings of connected and automated vehicles (CAVs) are here now, with the active safety systems which feature in newer cars, like lane-keeping assist and adaptive cruise control.
The Australian Road Research Board (ARRB) was commissioned by Austroads to undertake research into how ready Australia and New Zealand freeways and highways are for active safety systems and automated driving.
This project undertook an extensive field audit of Australian and New Zealand freeways and highways to inform the assessment of their readiness for active safety systems and automated driving.
The road audit included more than 8 million individual line segments and over 8,000 signs on a 25,000km sample of the road network which, although extensive, still represents less than 2% of the total network.
The focus was on addressing the following key questions:
- What are the locations, incidence rates and characteristics of potentially problematic infrastructure?
- What is the likely impact (and perhaps severity) of each occurrence on real-time Advanced Driver Assistance System (ADAS) and CAV operation, including whether a vehicle can correctly detect the line or sign on each occurrence, and is real-time driving operation affected because of failure to correctly detect or identify an item?
The first question requires an accurate assessment of road attributes across the network. This was achieved through video survey collection and a combination of human assessment and machine vision processing.
The second question assesses the impact of these attributes on CAV operation. For example, a line partially obscured by gravel or worn on a highway where detection fails may have no effect on CAVs or ADAS considering the way in which lines are pre-processed, isolated and averaged over long distances.
To assess this, the project made use of a real-time machine vision system, using hardware processing and representative of the technology fitted to modern vehicles with ADAS functionality.
The audit process comprised several stages:
- Road selection: The roads included sections of the freeway and highway network of Australia and New Zealand..
- Data collection: Data was collected from recent surveys as well as from third-party data providers. New surveys were conducted for the real-time testing component of the audit using a real-time machine vision system incorporating hardware processing.
- Data processing and validation: This involved extracting features to identify signs and lines and their associated attributes (e.g. line width, type and speed limit). The data processing methods used were manual processing, post-processing, real-time processing, and data mining/analysis.
- Test cases: For specific sites video images were selected to investigate the robustness, or otherwise, of the audit results when subjected to variations in road and observation conditions.
The solution we delivered
The audit found that most freeways and highways of Australia and New Zealand can for the most part currently support Advanced Driver Assistance System (ADAS) operation such as lane-keeping assistance, on roads with good quality lines, higher traffic volumes and good cellular availability.
The presence of left and right lane line markings is critical for lane positioning, and there are significant proportions of the road network without edge lines. Increasing the use of edge lines and dividing lines (lane lines and centre lines) will provide a clear immediate benefit for both automated driving and human drivers.
Research suggests that line markings also need to have good contrast with the surrounding pavement for accurate detection. This may be addressed with line maintenance and materials, wider lines (to improve contrast) and consideration for background luminance (of pavement materials), although in many cases contrast and reflectivity are subject to the current lighting conditions.
The audit found that speed limit signs can be reliably detected and read using automated methods and in real-time provided they were in the correct position.
The audit found that cellular infrastructure availability and diversity is generally adequate and not likely to be a limiting factor for CAV operation on major highways where uninterrupted connection is not required.
The project investigated high-definition (HD) map coverage to support future CAV operation, and found that HD maps are not currently available from any of the mapping providers within Australia and New Zealand. Many providers indicated that these may be ready within the next few years and will significantly enhance the suitability of the network for CAV operation.
Fully automated driving readiness will be challenging to achieve, even with advances in CAV technologies and accurate HD maps. Roadworks, temporary lanes, missing lines and lane closures are significant problems for automated vehicle operation. The incidence rate of these conditions is significant.
The real-world performance of ADAS and CAV systems is likely to be further diminished in poor lighting, traffic and weather conditions. The audit was based on best-case conditions for automated driving, in ideal conditions and away from intersections and built-up areas.
More work needs to be undertaken to assess the suitability of road infrastructure, and specialised and advanced real-time equivalent technologies are best suited for this purpose. The audit found that the performance of this equipment was high, outperforming the post-processed system, and its operation is likely to be consistent with current market and near-market CAV vision system technologies.
What our partners think
Our client Austroads is justifiably very happy with the world leading outcomes delivered by this project. Australia is now better placed than ever before to understand its infrastructure requirements for connected and automated vehicles.