Optimal Eco-Approach Control With Traffic Prediction for Connected Vehicles

This work proposes a unified framework for the eco-approach application that integrates traffic prediction, vehicle optimization, and implementation. The eco-approach application is formulated as either a car-following optimization problem or a single vehicle optimization problem, depending on whether a preceding vehicle exists. The traffic prediction scheme anticipates future traffic conditions and describes the traffic dynamics on the road segment of interest using state variables: traffic flow, density, and speed. With the information enabled by connectivity, the traffic state estimation is updated using an observer. Uncertainties in the traffic prediction are considered using a robust optimization approach. The robust optimization problem is discretized and solved by an efficient nonlinear programming solver. The proposed eco-approach framework is implemented to a single lane single intersection scenario for 12, 8, 4, and 1 connected vehicle scenarios. The fuel benefits vary from 11.0% to 6.7% as the penetration rates of connectivity decrease. The performance is satisfactory compared to the 12.0% fuel benefits with perfection traffic prediction.