If self-driving cars are to be widely adopted, we need to know that they can navigate difficult traffic scenarios, such as joining heavy traffic when lanes merge on a highway. To tackle this, North Carolina State University researchers devised a method that enables autonomous car software to perform the necessary calculations more rapidly, boosting both traffic and security in simulated autonomous vehicle systems.
Image Credit: Nick Fewings
Right now, the programs designed to help autonomous vehicles navigate lane changes rely on making problems computationally simple enough to resolve quickly, so the vehicle can operate in real-time. However, simplifying the problem too much can create a new set of problems, since real-world scenarios are rarely simple.
Ali Hajbabaie, Study Corresponding Author and Assistant Professor, Civil, Construction and Environmental Engineering, NC State University
“Our approach allows us to embrace the complexity of real-world problems. Rather than focusing on simplifying the problem, we developed a cooperative distributed algorithm. This approach essentially breaks a complex problem down into smaller sub-problems and sends those to different processors to solve separately. This process, called parallelization, improves efficiency significantly,” Hajbabaie said.
At this moment, the scientists have only evaluated their approach in simulations in which the sub-problems are distributed across multiple cores in the same computing machine. Moreover, if autonomous vehicles apply the strategy on the road, they will network with one another and share the computational sub-problems.
The scientists examined two things in solid evidence testing: if their technique enabled autonomous car software to address merging problems in real-time and how the novel “cooperative” approach influenced traffic and safety contrasted to an actual model for guiding autonomous vehicles.
In terms of calculation time, the scientists determined that their approach enabled autonomous vehicles to traverse complex freeway lanes combining scenarios in live time in moderate and heavy traffic, with spottier efficiency when traffic volumes became high.
However, when it came to enhancing traffic and safety, the new strategy performed admirably. In other instances, especially when traffic volume was low, the two systems performed similarly.
However, in most cases, the new strategy outperformed the old version by a wide margin. Furthermore, there were no situations where vehicles had to come to a halt or where there were “near crash conditions” with the new technique. The findings of the other model contained numerous scenarios with thousands of stoppages and near-crash circumstances.
For a proof-of-concept test, we’re very pleased with how this technique has performed. There is room for improvement, but we’re off to a great start.
Ali Hajbabaie, Study Corresponding Author and Assistant Professor, Civil, Construction and Environmental Engineering, NC State University
“The good news is that we’re developing these tools and tackling these problems now so that we’re in a good position to ensure safe autonomous systems as they are adopted more widely.”
Journal Reference
Tajalli, M., et al. (2022) Distributed cooperative trajectory and lane changing optimization of connected automated vehicles: Freeway segments with lane drop. Transportation Research Part C: Emerging Technologies. doi.org/10.1016/j.trc.2022.103761.