A sequential approach to time-dependent demand calibration: Application, validation and practical implications for large-scale networks

The complexity of transportation systems often dictates detailed representation of time-dependent demand and supply interaction through Dynamic Traffic Assignment (DTA). These complex models involve a large number of global parameters (behavior and congestion features) and main inputs (demand and supply characteristics) that require to be calibrated off-line, while stream of data coming from the field in real time can be used for the local fine-tuning of the simulation in Rolling Horizon (RH). In this paper, we present a sequential approach to calibrate time-dependent demand. This method calibrates several demand time slices in one calibration run, after which it shifts forward analogously to the on-line RH technique. The contribution of this paper is to present in detail the novel methodology, demonstrate its performance on a small-scale network and investigate its scalability to large-scale networks. We also analyze the behaviour of the sequential approach to provide recommendations for application for large-scale networks, as it is of high practical importance. We also suggest the settings that provides best convergence given a good starting point, which is crucial for the extension of this approach to real-time applications