Baltimore Metropolitan Council (BMC) Activity-Based Travel Model Peer Review Report
5.0 Proposed Regional Activity-Based Model Structure
The work on BMC's activity-based model began in June 2013 with a completion date scheduled for June 2016. The model includes three mandatory activities (work, school, and university) and five non-mandatory activities (meal, shop, personal business, social/recreation, and non-school escorting). It is intended that, where possible, parcel-level spatial resolution will be used for the land use data. For scheduling activities, a 30-minute temporal resolution will be adopted. The components of the proposed model structure are discussed next.
Figure 5‑1 shows the overall model structure and model components proposed for the Baltimore region. A brief description of each model component is provided below.
- Auto Ownership: This is a household-level model that predicts the number of autos owned by a household.
- Regular Workplace Location: This is a person-level model that will be applied to each employed individual to predict workplace location zone.
- Regular School Location: This is a person-level model that will be applied to each student to predict school location zone.
- E-ZPass Ownership: This binary choice model is to be used to predict whether or not a household owns an E-ZPass transponder.
- Transit Pass Ownership: This binary choice model will be used to predict whether or not a household owns a transit pass.
- Daily Activity Pattern: For each individual, this model predicts the number of tours (0, 1, or 2+) and the number of stops (0, 1, or 2+) for each activity purpose, including travel to work and school.
- School Escorting: This model will be applied at half-tour level (i.e., home to school and school to home). The model will be used to predict on which half-tours a student is escorted to/from school, which household member escorts the student, and whether escorting is done on a work tour.
- Fully Joint Travel: This model will be applied at household-level to predict the number of fully joint tours with two or more household members and which household members participate in each joint tour.
- Work-Based Sub-Tour Generation: This model predicts the number and purpose of any sub-tours made during a work tour.
- Work Mode Choice: This model predicts the main mode for work tour.
- School Mode and Time-of-Day Choice: This model will be used to predict the main tour mode, (in 30 minute intervals) the time period for arriving at school and the time period for leaving school.
- Work Time-of-Day Choice: This model predicts (in 30 minute intervals) the time period arriving at work, and the time period leaving work.
Figure 5‑1: BMC's Activity-Based Model Component
(Reproduced from "Model Design Plan for BMC Activity-Based Model", Draft Report, September 2013)
- Other Tour Time-of-Day Choice: This model predicts in 30 minute intervals the time period arriving at the primary destination and the time period leaving the primary destination for non-mandatory tours.
- Other Tour Mode and Destination Choice: This model predicts the primary destination zone and main tour mode for non-mandatory tours.
- Intermediate Stop Generation: This model predicts the number and activity purpose of any intermediate stops made on the half-tour. The model prediction is conditional on day pattern.
- Intermediate Stop Location: This model predicts the destination zone of each intermediate stop (conditional on tour origin and destination) and location of any previous stops.
- Trip Mode Choice: This model predicts trip mode conditional on tour mode.
- Trip Departure Time: This model predicts trip departure time, conditional on available time windows.
BMC's goals for the peer review were discussed in section 3.4. One of the key issues presented in detail was how to model road pricing. Specifically, how to improve the treatment of road pricing in the mode choice model. The current proposal to model road pricing includes the following enhancements:
- Simulate value-of-time for each individual, possibly by tour purposes. For this, a truncated log-normal distribution may be used.
- Segmentation of trip tables used in aggregate highway assignment by value-of-time level (This technique is currently being implemented in the Houston H-GAC activity-based model).
The main difference between the proposed approach and more traditional segmented mode choice models where the auto mode is divided into "toll" and "free" alternatives are:
- In the proposed approach, mode choice is applied separately for the travelers in each segment rather than using segmentation to create separate mode alternatives.
- Use of multiple segments of value-of-time instead of usual toll/non-toll segments.
- Though there is no guarantee that a "free" path will be used in the development of travel time skims, the proposed approach is likely to increase the probability of a free path being chosen for the lowest value-of-time segments.
