Eco-cruising control under cut-in maneuvers using spatiotemporal attention-based trajectory prediction

With the rapid advancement of connected and automated vehicles technologies, eco-cruising control strategies have attracted increasing attention due to the significant potential for energy conservation. However, stochastic disturbances induced by human-driven vehicles (HDVs) can markedly compromise both energy efficiency and driving comfort. To tackle this challenge, a spatiotemporal prediction-based eco-cruising strategy is proposed to optimize velocity profiles by considering the spatiotemporal interaction among surrounding vehicles in a hierarchical framework. In the upper layer, a graph-based attention model is developed to capture spatiotemporal features from historical HDV trajectories, thereby accurately predicting their motion trajectories in mixed traffic environments. In the lower layer, a model predictive control integrated with a quadratic programming algorithm is designed to achieve ecological speed optimization accounting for random cut-in behaviors. Real world motion data are leveraged to validate the proposed approach, demonstrating that the proposed prediction model can accurately predict vehicle lane-change trajectories, while the overall performance is significantly improved under random cut-in maneuvers. Specifically, the proposed method achieves an average improvement of 16.23% in energy efficiency while ensuring preferable travel efficiency and driving comfort.