Activity timing choice#

An activity timing choice model is developed to generate the activity start time and duration within POLARIS using data from a GPS-based prompted recall activity-travel survey, UTRACS survey, that collected detailed information of activity planning and scheduling of the households in the Chicago region (Frignani et al., 2010). The model is primarily estimated as a joint discrete-continuous model (Hybrid Random Utility Maximization (RUM)-Random Regret Minimization (RRM) Copula model), where the variables in the discrete component (activity start time) are treated by both utility maximization and regret minimization decision rules and the continuous component (activity duration) is estimated using a hazard duration model. Then, the two activity components of start time and duration are linked through their stochastic error terms using the copula approach. This approach uses the marginal distribution of two random variables and estimates their joint probability distribution without imposing any restrictions.

The current POLARIS model implements two separate processes for routine and non-routine activities. The start and duration of the former are drawn from an observed joint start and duration distribution per activity type, which needs to be provided via the START_DURATION_DISTRIBUTION_FILE parameter (default start_time_duration_data.txt). This data is usually derived from surveys and to account for reporting biases the drawn start times are randomly adjusted by 0 to 10 minutes, and the duration times are randomly adjusted by 0 to 5 minutes. The non-mandatory start and duration model utilizes the RUM portion of the start time model described above to generate the activity start time, and randomly draws the activity duration from a joint distribution of start time-duration based on activity types. This data is provided in the same file as the routine activity start and duration distribution.