Telecommute Design Implementation

[i-am-sijia]

In Phase 9B, the Consortium selected a model design for explicitly simulating telecommuters and their behaviors in ActivitySim. The design, simplified in the flowchart below, includes adding two new components and re-estimating some downstream components.

To implement this design with the example SANDAG model, the follow main steps are anticipated:

1. Estimate a Telecommute Arrangement model as designed. This model should incorporate the person’s CDAP pattern as one of the explanatory variables. The estimation process will also examine potential intra-household effects, such as how household members’ daily patterns and/or in-home work patterns may influence each other.
2. Estimate an In-Home Activity Duration model as designed. This model should be applied only to those who have in-home work activities as predicted by the Telecommute Arrangement model. This model may be as simple as drawing from a simple duration distribution or may be estimated with explanatory variables in a multinomial logit form.
3. Re-estimate the Mandatory Tour Frequency and Scheduling Models by incorporating in-home work activity and duration as explanatory variables. This approach will enable the models to better account for the impact of in-home work on the frequency and scheduling of out-of-home mandatory tours, particularly for hybrid workers engaged in both in-home and out-of-home work activities during the simulation period.
4. Re-estimate the Non-Mandatory Tour Frequency and Scheduling models using in-home work activity and duration as explanatory variables to impact the frequency and schedule of out-of-home non-mandatory tours. To limit the time window available for scheduling non-mandatory tours based on in-home work duration, without explicitly scheduling in-home work activities, it may be necessary to define a standard “common work window” for the simulation day (e.g., 7am–7pm). The model can then assess whether a worker has sufficient time during this period for the total in-home work duration when scheduling out-of-home non-mandatory tours.
5. Exploratory analysis of other downstream models related to telecommuting. For example, should non-mandatory tour destination choice model include telecommuting as an explanatory variable?

[joecastiglione]

The telecommuting design memo had referred only to re-estimation of the non-mandatory tour frequency and scheduling models, but it makes sense that the mandatory tour models should be considered as well. All these models (and possibly others?) will also need to be (re)calibrated. In terms of scoping for Phase 11C, what is the full list of tasks (HTS data prep?, estimation + calibration of X models, sensitivity testing, code management testing etc)? Can these tasks be defined and sequenced (either in parallel or serially) such that they can be completed in 3 month tranches?