Introduction:
Achieving desired levels of accuracy in the outcome of travel demand forecasts produced by micro-simulation of household behavior may require a large sample of households. This may happen when: high levels of spatial or temporal resolution are required of the outcome; sample households do not have a desirable geographical distribution; demand by a small population segments is desired; or a high level of accuracy is desired. In such instances the number of households available in the data set at hand may not be sufficiently large and generation of synthetic households may be required. When the micro-simulation expects daily travel patterns of household members as input data, it calls for generation of synthetic daily travel patterns.
An approach to the problem of synthetic travel pattern generation is proposed in this report. The approach adopted here is sequential. The proposed model system can be decomposed into components to which certain aspects of observed activity-travel behavior correspond. This establishes a link between the mathematical models and observational data. The model components are each relatively simple and can be estimated using commonly used estimation methods and existing data sets.
The problem of synthetic travel pattern generation is first formulated and presented formally in Section 2. The knowledge that has been accumulated on the characteristics of daily travel patterns is briefly reviewed in Section 3. Following this, Section 4 is devoted to the discussion of the relative advantages of sequential and simultaneous modeling approaches. The formulation of the model system and its components is described in Section 5. Significant portions of Section 5 are dedicated to the discussion of behavioral and statistical issues associated with modeling daily activity-travel patterns. Section 6 offers an explorative analysis of history dependence in activity engagement. Section 7 is a conclusion.
tr39.pdf