Long Short-Term Memory-Based Model Predictive Control for Virtual Coupling in Railways

Abstract

The increasing need for capacity has led the railway industry to explore new train control systems based on a concept called virtual coupling. Inspired by the platooning of autonomous vehicles, the safe operation of virtual coupling is guaranteed by a relative brake distance-based train separation method. This paper proposes a novel long short-term memory (LSTM)-based model predictive control (MPC) method for train operations. An MPC-based control design for virtual coupled train operations is presented. The LSTM is introduced to model the dynamics of the preceding train to approximate the actual train operations. With the train dynamics models, the operation trajectories of the preceding train are predicted based on planned control inputs. A study of a metro line in Chengdu was chosen to analyze the proposed control approach. The simulation results of different scenarios show that compared with the conventional MPC methods, the proposed LSTM-based MPC can reduce the speed differences and position differences of tracking trains by up to 35 % and 25 % , respectively.

Railway Virtual Coupling: A Survey of Emerging Control Techniques

“…Longer prediction horizons commonly mean longer computing time. Researchers in [53,54,63] reported increases in both controller performance (less tracking errors) and computing time (see Fig. 8) with the increases of prediction steps.…”

“…Su et al [53] removed the mass term and replaced the train speeds by train positions to conduct VC simulations. In [54], three cost functions were used for acceleration, speed, and position. They all had the same format as the one used in [53].…”

“…Three cost functions were not combined but used individually according to different train operational conditions. Both [53] and [54] used a Neural Network model to predict the dynamics of a preceding train. When multiple parameters are considered, the cost function can be written as:…”

A Survey on Control Methods for Virtual Coupling in Railway Operation

“…Stability: [43], [44] combined deep learning with a MPC method considering the effect of communication delay of unit trains in the platoon. The Long Short-Term Memory (LSTM) neural network is used to predict the platoon's operating state of the preceding unit train.…”

“…Able to improve the stability [43], [44] Suitable for finding the optimal control strategy [45] simulated and verified by numerical calculation and simulation analysis showing that the improved station tracking model has the same capability as the relative moving block, and the VC model has the largest capability. The VC model has the strongest delay recovery ability after the system is subjected to initial delay.…”