Real-time marginal cost estimation for eco-routing navigation guidance

Principal Investigator(s):

Michael Levin, Assistant Professor, Civil, Environmental and Geo-Engineering

Project summary:

Transportation caused 27% of US greenhouse gas emissions in 2022 (EPA), so reducing emissions from transportation is important to resolving the climate crisis. Towards this goal, Google Maps recently introduced a new default navigation mode: they suggest routes to drivers to minimize their own CO2 emissions instead of minimizing their travel time. Since Google Maps is widely used, this switch could result in correspondingly large changes in how drivers choose routes. However, the goal is to reduce total systems emissions (system-optimal (SO) eco-routing), but Google Maps routes drivers to minimize their individual emissions, which is user-optimal (UO) eco-routing. With a large proportion of drivers trying to minimize their individual emissions, that significantly changes congestion patterns, and traffic congestion increases emissions per mile compared to less-congested traffic. In fact, my preliminary work in IEEE Transactions on Intelligent Transportation Systems demonstrates that UO eco-routing could have higher emissions than UO travel time routing. In other words, Google Mapsa?? well-intentioned change to routing behavior could in fact be causing transportation emissions to become worse! The purpose of this project is to develop a real-time navigation guidance that gets closer to SO eco-routing, i.e. an improved algorithm that Google Maps or similar systems could use instead. The main challenge is predicting how the route choices of drivers affect congestion for other drivers. Since the purpose is for actual use, realistic models of traffic flow behavior need to be used, which also make computation more difficult, and the system must operate in real-time to be useful. Therefore, my theoretical approach to this problem is based on marginal-cost routing, which is guaranteed to find SO eco-routes if marginal costs are estimated correctly. Due to the difficulty of calculating marginal costs, I propose a heuristic to estimate them, and simulations to validate the effectiveness of the heuristic.

Project details: