As the site of frequent vehicle collisions, roadway intersections can be dangerous places.
Navigating automated vehicles (AVs) safely through intersections can be especially challenging, given that cars crossing the intersection have a number of turning options, human drivers behave unpredictably, and right-of-way rules can be unclear.
University of Minnesota (UMN) research has found that a safety feature designed for AVs shows promise for helping these vehicles navigate safely and avoid collisions at intersections.
Led by Rajesh Rajamani, a professor of mechanical engineering and CTS scholar, the research project investigated the use of a control barrier function (CBF)—an algorithm used to adjust the speed of an AV so it avoids meeting any cross traffic at an intersection.
CBFs have been applied in AV research for collision avoidance, adaptive cruise control, and lane keeping. This project focused on tailoring and validating CBF systems in simulations, with the goal of making them suitable for real-time use in AVs.
“With a control barrier function, the AV either slows down or speeds up so that it goes through the intersection at the correct time so there is no conflict with the cross traffic,” Rajamani says.
For the first part of the project, researchers analyzed theoretical models for CBFs as well as developed simulations to evaluate how well CBFs can help an AV navigate safely through an intersection. Through extensive testing in the CARLA autonomous driving simulator, the researchers demonstrated that CBFs are effective in guaranteeing vehicle safety even in complex and dynamic intersections that involve cross traffic.
For the second part of the project, the team used the MnCAV Ecosystem’s automated vehicle to conduct a real-world proof-of-concept experiment. The 2021 Chrysler Pacifica minivan serves as a customizable, experimental testbed for AV research, education, and outreach at the UMN.
“We set up hardware consisting of a radar and a wireless communication system at an intersection. This hardware finds out if there is cross traffic and transmits information about the location and velocity of cross traffic to the autonomous vehicle,” Rajamani explains.
Researchers tested two scenarios at a two-way stop-sign intersection. In the first scenario, the MnCAV vehicle approached the stop sign and came to a complete stop. The radar confirmed no cross traffic, so the MnCAV vehicle proceeded through the intersection safely and smoothly. In the second scenario, when a cross-traffic vehicle was detected by radar, the MnCAV vehicle held at the stop sign and waited until the intersection was clear before advancing.
While the CBF control system was not used onboard the MnCAV vehicle during the field tests, the experiments confirmed that the vehicle handled simple intersection conditions as expected.
The layered approach to this research, combining theoretical analysis, simulations, and real-world experiments, offers a promising roadmap for advancing AV safety technologies at intersections, Rajamani says.
One possible next step in research includes using a CBF control system on the MnCAV vehicle to evaluate real-time performance in the field.
This project was supported by CTS seed funding. Awarded biennially, this funding aims to help CTS scholars develop expertise in emerging areas and foster strategic relationships that position them for future funding opportunities.
—Peter Raeker, contributing writer
