An aggregate approach for the calibration of time-dependent demand in Dynamic Traffic Assignment models using SPSA algorithm

Demand calibration in Dynamic Traffic Assignment is often regarded as an optimization problem. The most widely algorithm used in the literature is SPSA, using single O-D pairs as calibration variables. This approach is very difficult to apply to large-scale networks due to the large number of variables. In this paper we present a novel approach that can lead to superior demand calibration results. It consists of the aggregation of single demand components based on a correlation object. An analytical model was created to represent real-world dynamics and is used for tests. We tested different scenarios regarding different levels of measurement errors. The results show that the calibration using SPSA algorithm with total origin demand and total destination demand can converge to the solution, whereas using standard approach, i.e. O-D pair calibration, breaks. This approach has a number of advantages which are analysed and discussed.