.png){kind=icon}
.png){kind=icon}
Return
Real-time motion planning for virtual coupling train operations: A learning-based combinatorial optimization method
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2026.115529.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In response to the increasing demand of flexible train formations, the concept of virtual coupling is recognized as a promising technology. The virtual coupling motion planning problem involves both trajectory optimization in a specific operational state and multi-state transition decision-making. To describe the entire virtual coupling process consisting of multiple operational states, we propose a mixed integer nonlinear programming (MINLP) model that achieves state transitions between different operational states. To speed up the optimization process for real-time generation of virtual coupling train operation trajectories, our work designs a learning-based combinatorial optimization method comprising both offline deep learning and online optimization procedures. We innovatively combine operational mechanism with neural network prediction to determine the values of 0-1 variables, significantly enhancing the reliability of their values. Then we solve a nonlinear programming (NLP) problem with the fixed 0-1 variables by the sequential quadratic programming (SQP) algorithm. Under the model predictive control (MPC) framework, our method dynamically generates optimal trajectories through a rolling horizon optimization strategy. Furthermore, a new optimization problem is formulated to address the infeasibility problems arising from mispredicted 0-1 variables, ensuring that virtual coupling trains achieve efficient coupling and decoupling while maintaining safe operation. We evaluate the effectiveness and robustness of the proposed motion planning method through numerical experiments. The experimental results indicate that the proposed method achieves rapid and smooth coupling and decoupling processes while ensuring operational safety.
Journal
IF:
8
Papers:
5.3K
Citations:
3.5W
Organization
No organization information available