A Model Predictive Control Approach With Limited-Time Termination for Virtually Coupled Train Set

To improve the flexibility and efficiency of railway services, virtual coupling has been investigated widely. It forms multiple train units as a virtually coupled train set (VCTS), deploying advanced communication and control technologies instead of physical couplers. To operate VCTS as desired, all units should drive synchronously, while satisfying a safety constraint to regulate the following distance. Thus, model pre-dictive control approaches are a promising solution to achieve VCTS operation. However, the computation efficiency of MPC is a critical problem to be addressed, considering the limited capabilities of realistic on-board train computers. To solve this problem, this paper proposes a fast-safe MPC approach that can terminate in a limited time while generating a quasi-optimal and feasible controller. By adding barrier functions that are specifically transferred from constraints in VCTS operation, we formulate an augmented cost function for MPC. Then, a fast-safe solver is designed to analytically solve the optimal control problem in MPC, using the gradient of augmented cost function and adaptive updating steps. Experimental results demonstrate that our approach can yield controllers for VCTS in a limited time, while operating VCTS stably and guaranteeing constraint satisfaction.