Adaptive Optimal Control for Designing Automatic Train Regulation for Metro Line

Abstract

Automatic train regulation (ATR) plays an important role in maintaining the service quality of a metro in relation to schedule and headway adherence. However, maintaining service quality in an optimal way with less capacity utilization of infrastructure, particularly in an environment where disturbances occur frequently, is a challenge. Intrinsically, designing ATR is a real time optimal control problem with high nonlinearity, heavy constraints, and stochastic characteristics. Dual heuristic programming (DHP) was successfully employed to design ATR; however, the influence of traffic modeling error on the ATR performance was observed. In this paper, the adaptive optimal control (AOC) method is developed to improve the DHP design in regard to modeling errors as well as optimality, and an ATR designed using AOC method was developed and evaluated. The evaluation shows that the AOC method is able to find a near-optimal solution more rapidly and accurately than the DHP method. Moreover, the ATR designed using the AOC method improves both schedule and headway adherence with less capacity utilization and is more robust against disturbances as well as traffic modeling errors resulting from passenger flow fluctuations.