Data-Driven Energy-Optimized Speed Trajectory for Urban Driving Electric Vehicles Utilizing Traffic Flow Estimation

This study proposes a practical-oriented speed trajectory optimization strategy that minimizes the expected energy consumption of electric vehicles (EVs) passing through multiple signalized intersections in mixed-traffic urban environments. To this end, traffic flow is estimated by averaging and clustering speed trajectories from low-frequency probe vehicle data. The Gaussian mixture model (GMM) is used to obtain the vehicle's average speed probability distribution between signalized intersections. Using the estimated traffic conditions and probability distributions, a two-stage optimization algorithm is conducted. The offline stage estimates energy consumption between multiple consecutive intersections. Then, the online stage derives the optimized speed trajectory from the estimated energy consumption tables by using dynamic programming (DP) under speed limitations in accordance with traffic flow. The proposed strategy does not require additional vehicle-to-infrastructure (V2I) communication, and its algorithm can be performed recursively, thereby alleviating both implementation cost and computational burden. Numerical simulation demonstrates the proposed strategy's merit compared to a standard driver model and an optimization strategy that utilizes V2I communication. The proposed strategy has been successfully evaluated using a system developed by our group. Experimental results show that the proposed strategy can effectively estimate traffic flow and reduce energy consumption by 7.6% compared to a preceding vehicle.