Automated assessment of infrastructure preparedness for autonomous vehicles

The current paradigm shift in transport and mobility has compelled government agencies to plan for the mass deployment of connected and autonomous vehicles (CAV) technologies in the next decade. Indeed, CAV is considered a truly transformative technology that has great potential for reducing traffic accidents, enhancing quality of life, and improving the efficiency of transportation systems. However, current challenges continue to postpone the public implementation of this technology. Thus, it is imperative to reassess the existing road infrastructure for autonomous driving compatibility. Recent research on this topic focuses on closing the gap between the CAV technologies and the design of the existing roadways. With the input of various stakeholders, there has been considerable effort to assess the CAV readiness of the existing infrastructure.

Two approaches have been recommended to improve the reliability and readiness of CAV systems, namely, identifying locations with substandard conditions and deploying vehicles to infrastructure (V2I) at locations deemed critical. V2I connections provide the CAVs with the necessary information, such as potential hazards, speed limits and obstructions. Autonomous vehicles (AVs) have six levels of advanced driver assistance systems (ADAS), ranging from human driver-operated vehicles supported by ADAS (levels 1 – 2) and those supported by high-level automation features (levels 3 – 5). Therefore, during the transition through the levels, the design of highways must account for both autonomous vehicles and human drivers.

AVs are equipped with an onboard computer system and numerous sensors to map and perceive their surroundings just like human drivers and safely and accurately use different road features. The success of CAVs depends on their ability to achieve effective control and avoid collisions. However, due to the limitations of different components of CAVs, supporting high-level autonomous driving needs remains a big challenge. Moreover, there is limited knowledge on the appropriate strategies for assessing the CAV preparedness of the current road infrastructure on a network level. Such approaches should be automated to avoid the disadvantages of manual surveying commonly used in current road assessment and maintenance. To this end, modern sensing technologies like light detection and ranging (LiDAR) have been deemed effective for digitizing road infrastructure and various road assessment methods.

On this account, researchers at the University of Alberta: Ph.D. candidate Maged Gouda, Mr. Ishaat Chowdhury, Mr. Alexander Epp and Professor Karim El-Basyouny in collaboration with Mr. Jonas Weiß from the Technical University of Munich, developed a raycasting approach for assessing the infrastructure readiness for autonomous vehicle deployment in a virtual environment using LiDAR data. The team tested the highway readiness under different AV driving scenarios. They then conducted a thorough and accurate assessment of the highway design, including sight distance and various road characteristics. Their method was validated using an alternative algorithm. This work is currently published in the journal, Automation in Construction.

The research team showed that some AV scenarios could result in unsafe driving conditions for AVs. Interestingly, the proposed method is suitable for locating speed limits and limited sight distance using highway network data. The study provided an effective approach for the quantitative assessment of different highway geometric design features and effectively determined the overall road infrastructure preparedness for CAVs. The proposed algorithm agreed with the alternative algorithm producing speed limit and sight distance values within 1.94% and 1.8% of each other, respectively. It could thus allow designers to build appropriate roads for AVs and regulators to set legislation guidelines and standards.

In summary, the researchers proposed a novel approach for assessing sight distances of highway roads for different AV driving scenarios based on the designs and specifical of a typical vehicle sensor set. The proposed approach is an ideal tool for policymakers to assess the preparedness and compatibility of roads for various AV scenarios in a technology-independent fashion. The sensor parameters are generalized well to suit the current and future technology and allow regulators to keep up with the rapidly evolving CAV technology. It could enable manufacturers to validate and improve the performance of the sensors. In a statement to Advances in engineering, Dr. El-Basyouny explained their new approach provides a path for developing performance-based design guidelines for engineering suitable highways for autonomous vehicles. He argues that “The driverless car is the vehicle of the future. So, transportation innovations will need to address issues around new energy sources, advanced modes of transport, and smarter physical and technological infrastructure to support our transportation ecosystem.”