Case Studies Train Control Optimization

A new era of automatic train control has begun, in which mass transit trains will be commanded with precision beyond the capabilities of past systems. Although transit properties such as San Francisco's Bay Area Rapid Transit (BART) have controlled their trains automatically for decades, the control systems have limited capability. Increases in capacity now require trains to run closer together than these systems can accommodate, so new systems are being developed. These new systems, such as the Advanced Automatic Train Control (AATC) system under development by BART in collaboration with Raytheon Corporation and Harmon Industries, will increase the capacity of the system through more accurate train location and more precise control. Although the need to increase capacity has been the main driver for these new systems, they will also enable smoother service and improved energy management. However, incorporating smoother service and improved energy management into the control system will require optimization of a complex dynamic system. Sandia has collaborated with BART to develop a simulator of the train control and power consumption of the AATC system. The simulator has enabled us to develop enhanced train control algorithms to supplement the safety-critical controller. These algorithms do not attempt to globally optimize the control system with respect to a cost function, but rather they modify the baseline vital control using heuristics to smooth out train operations, and to reduce energy consumption and power infrastructure requirements. Although enhanced train control algorithms provide a valuable rst step toward optimization, they represent only a fraction of the ultimate capabilities of the system. We are now beginning work toward true optimization of the control system. Train control optimization encompasses such classes of optimization as mixed integer nonlinear programming, nonlinear discrete-time optimal control, and multi-objective optimization.