STRIDE Student Spotlight: Luan Guilherme Staichak Caravalho, Ph.D. Student, University of Florida

Luan Guilherme Staichak Carvalho is a doctoral student affiliated with the University of Florida Transportation Institute who specializes in machine learning and mathematical optimization. As novel technologies such as connected and autonomous vehicles emerge, Carvalho is applying this knowledge to the optimization of urban intersections.

Carvalho is working with Dr. Lily Elefteriadou at the University of Florida Transportation Institute on projects funded by the Florida Department of Transportation and the National Science Foundation. These projects are multi-disciplinary in nature, which include faculty and graduate students from transportation engineering, computer science, and mechanical engineering. The projects are: FDOT Contract BDV31-977-45, FDOT Contract BDV31-977-10, and NSF Award 1446813.

Work on these projects focus on expanding and refining a software designed by UFTI known as Real-time Intersection Optimizer (RIO). This software, which has been in development for over 4 years, is a traffic signal manager that uses the position of regular vehicles and connected vehicles (common or autonomous) to adjust the best signal plan in order to serve the approaching vehicles with reduced delay and optimizing the usage of green time. The position of vehicles is given either by video feed or through the radio communication between the connected vehicles and RIO system. Signal plans developed by RIO are very dynamic, with the order and length of phases being assigned according to the vehicle demand instead of following a strict pattern.

“RIO offers an optimized trajectory for the connected vehicles, allowing them to avoid unnecessary breaking,” Carvalho said. “Leading vehicles receive instructions so that they reach the stop-bar as soon as the green light time begins, while follower vehicles maintain a gap with the leading vehicle that is both safe and as short as possible.

The combination increases the throughput of the whole intersection, reducing delay and emissions, he added.

Carvalho’s role in this project is to refine the RIO logic from the transportation engineering perspective. For example, allowing the system to operate under increased traffic demand and considering different intersections and other scenarios. He also leads the system deployment for the transportation engineering team to ensure that the RIO simulation models are working properly in field trials. He is also working on incorporating pedestrians within the RIO framework to make sure that pedestrians are correctly recognized by the system.

So far, the research team has been able to deploy an optimization procedure for signal phase and timing that consistently and sharply reduces the delay suffered by vehicles when they utilize an intersection managed by RIO. The only issue observed is a marginal increase in pedestrian delay. This improvement is attainable even when the vehicle demand consists purely of regular (non-connected) vehicles, which becomes gradually better as traffic demand is simulated with higher penetration rates of connected and/or autonomous vehicles.

The team has also been able to identify ways to modify the system to greatly reduce pedestrian delay at the expense of a slightly increased vehicle delay. Carvalho says this alternative is particularly interesting for locations with heavy pedestrian demand such as many university campuses.

As a result of this project, Carvalho and the team of graduate students and faculty have created an optimization system (RIO) that can be deployed and integrated into rea-life urban intersection signalization systems. RIO has the capability of increasing performance at intersections when both conventional and autonomous/connected vehicles are considered.

“Transportation professionals and agencies may benefit from RIO by direct deployment or utilizing it as a basis for different optimization methods, adding more modes of transportation within the optimization framework,” he said.