NSF-Funded Study Creates Strategy to Improve System Performance of Traffic Apps

This study was recently published in IEEE – Transactions on Intelligent Transportation Systems (https://ieeexplore.ieee.org/document/9312471)

A diagram of the CRM-M-IP schematic.

Doctoral candidate Stephen Spana and his adviser, Dr. Lili Du of the UF Department of Civil & Coastal Engineering, worked on a National Science Foundation study aimed at mitigating system traffic congestion by leveraging users’ reaction to real-time traffic data provision such as from Google Maps or Waze or other apps.

For example, traffic conditions can deteriorate when users simultaneously flock to routes which the apps are showing to be uncongested, a phenomenon known as overreaction. Additionally, users exhibiting certain route selection behavior (such as selecting the route with minimum travel time) can lead to a sub-optimal system-level performance. The work that Spana and Dr. Du undertook developed a proactive strategy for information providers (i.e. Google Maps and Waze) to implement under a connected infrastructure environment. The researchers identified this strategy as a coordinated routing mechanism under a mixed-strategy congestion game with information perturbation (CRM-M-IP).

“This routing scheme coordinates the routing decisions of all users to avoid the situation of overreaction,” Spana said. “Additionally, strategically perturbing the traffic information sent to vehicles ensures that the decision-making of users will help improve network-wide conditions.”

While information perturbation will improve system conditions, it will also cause some users to sacrifice individual user optimality. Spana said their work was focused on exploring the trade-off between system benefit and individual user loss resulting from information perturbation.

Stephen Spana will be graduating with his Ph.D. in Summer 2021.

Spana’s role in this project was to develop the CRM-M-IP mathematical model. He was also tasked with demonstrating that a system-level improvement was reached due to the implementation of information perturbation, and that it was more significant than the corresponding individual losses. Additionally, Spana wrote the MATLAB code used for the numerical experiments that were used to verify and test their theoretical results. He also co-authored the journal article that was recently published in IEEE- Transactions on Intelligent Transportation Systems along with Dr. Yafeng Yin of the University of Michigan.

Results from the study showed that information perturbation indeed created a system performance improvement that was more significant than the corresponding user losses

“This result holds at low perturbation levels, meaning small perturbation to traffic information sent to users can result in noticeable system improvement with relatively low individual sacrifice,” Spana said. “These outcomes were confirmed using numerical experiments.

The researchers also used simulation to quantify the system performance improvement (>3%) and average travel time improvements (>3.5%) resulting from following the CRM-M-IP.

This work shows the potential benefits of using this type of strategic information delivery under connected vehicle infrastructure. A successful implementation of the CRM-M-IP could result in a noticeable reduction of system-level travel costs, while requiring negligible user-optimality sacrifices for a small percent of users.