Eco-Approach with Traffic Prediction and Experimental Validation for Connected and Autonomous Vehicles

This work proposes a fuel efficient vehicle speed control strategy for a connected and autonomous vehicle to pass the intersection (the eco-approach application). A control framework is developed to integrate traffic prediction (enabled by connectivity) and speed optimization. The traffic prediction is based on traffic flow model and can be applied to mixed-traffic scenarios where both connected and non-connected vehicles are on the road. Real-time information from connected vehicles and signal lights provides 'partial' measurement of the traffic states. The unknown traffic states are estimated using an observer (Unscented Kalman Filter). Uncertainties in the traffic prediction are systematically considered during the vehicle speed optimization to improve the robustness of the optimal control. Effects of powertrain states (the transmission gear ratio) on fuel consumption are considered as well. The vehicle speed optimization problem is real-time implementable using an efficient nonlinear programming solver. The optimal speed control strategy is evaluated in both a simulated traffic scenario and a real-world traffic scenario. In the simulated traffic scenario, fuel benefits vary from 5.3% to 9.4% as the penetration rate of connectivity increases. This is satisfactory compared to 11.7% fuel benefits with perfect traffic prediction. In the real-world traffic scenario, experimental results show that 6.9% fuel benefits can be achieved with two connected preceding vehicles and 11.2% fuel benefits can be obtained with perfect prediction. The results demonstrate the effectiveness of the traffic prediction and the optimal vehicle speed control strategy.