A transit route simulator for the evaluation of control strategies using automatically collected data

Abstract

This thesis develops and tests an extensible simulation model that uses automatically collected transit data to simulate transit route operations, demand, and control mechanisms. This model is motivated by the increasing availability of automatically collected transit data, which enables more detailed simulation and validation and also allows for advanced control strategies that can be evaluated using simulation. A framework is presented for using simulation to evaluate the improvement in service quality enabled by data. Most previous transit route simulation models included an explicit representation of traffic flow, which requires extra input data and introduces extra complexity. A detailed simulator design is presented that uses only transit-derived data to simulate vehicle and passenger movements and outputs a detailed log for flexibility of performance measurement. A case study of operations on the CTA's Route 9 Ashland was used to demonstrate and test the simulator. The simulator could be used to test alternative operator and supervisor behavior strategies and supervisor deployment schemes, as well as potential technological advances involving real-time data. Schedule, vehicle movement, terminal departure punctuality, passenger demand, and dwell time inputs for the simulator were derived from Route 9's schedule, AVL data, and APC data. The case study simulation was subjected to validation tests that compare simulated and real headway regularity, trip travel time, and maximum load statistics. Significant differences were found in all three tests. Adjustments were employed in attempt to make the simulation match reality.

(cont.) The results of adjustments to input parameters show that dwell times are an important source of headway variability. The results of adjustments to operator and passenger behavior and of controls indicate that effects that apply only to bunched vehicles have limited impact on service. After each of these adjustments, the simulation still did not pass validation tests. A prime cause for this result may be the intelligent behavior of transit agency personnel, particularly operators, a potentially fruitful area for future research.