Predictive Eco-Routing with Mixed Powertrains Under Connected Environment

Eco-routing enables individual vehicles to identify the most energy-efficient routes. Routing strategies, however, might be different between powertrain technologies given their different energy consumption behaviors. Such heterogeneity, coupled with the unpredictability in traffic dynamics due to non-recurring events (e.g. road incidents), poses substantial challenges to effectively implement eco-routing and quantify its energy implications. This paper proposes a novel predictive eco-routing system that leverages the capabilities of connected vehicles (CVs) to deliver optimal routing solutions with predicted traffic information for different powertrain technologies. By integrating power-based energy models with a deep neural network, the system can accurately estimate and predict link-level traffic conditions and energy usage. The proposed system is tested via microscopic simulation and demonstrated to reduce both energy consumption and travel delays for each powertrain technology, with internal combustion engine vehicles (ICEV) and hydrogen fuel cell vehicles (HFCV) achieving more energy reduction than electric vehicles. The results also highlight that predictive routing outperforms real-time routing, especially in networks with non-recurring events like road incidents. The sensitivity analysis verifies the reliability of the proposed system in supporting eco-routing at different congestion levels.