Saving energy with eco-friendly routing of an electric vehicle fleet

This paper fills the research gap between theoretical vehicle routing algorithms and practical solutions in the field. We use commercially developed prediction algorithms for the energy consumption of vehicles and solve for the energy-efficient routing and charging strategies of an electric vehicle fleet to visit a given set of destinations using meta-heuristics. Then we validate the energy saving performance of the efficient routing solutions with real-world vehicle measurements in a real traffic network. We also conduct a sensitivity analysis via simulation to explore some critical factors in reducing energy consumption. Current literature on vehicle routing problems is limited because they do not validate the energy consumption or the travel time of the solution in the field. This opens questions regarding the routing performance, which faces significant uncertainty from a real traffic network. Also, the travel costs in the network, such as energy consumption and travel time, have been predicted with simple or physics-based models in the theoretical studies, which are limited in capturing the complexity of a real traffic network. Our contributions are as follows. First, we develop a comprehensive framework for routing and charging an electric fleet, integrated with a high-precision prediction algorithm of energy and travel time based on learning a large driving data set. Second, we validate our routing performance in the San Francisco East Bay area with an energy consumption reduction of up to 31% compared to a baseline. Third, we conduct a sensitivity analysis via simulation on some critical constraints of vehicle routing. We observe that relaxing the limitation on operation duration and battery size on the vehicle fleet can expand the solution's feasible space and reduce the optimal energy consumption; however, the benefits diminish as constraints are relaxed to a certain point.