Activity plans
If the synthetic population describes who travels, activity plans describe where, when, why and how they travel.
An activity plan represents the daily or weekly sequence of activities and trips undertaken by each synthetic individual.
What a plan contains
An activity plan typically specifies:
| Element | Example |
|---|---|
| Activity purposes | Home, work, education, shopping, leisure |
| Locations | Where each activity takes place |
| Start and end times | When each activity begins and finishes |
| Travel modes | How the person gets between them |
| Connecting trips | The journeys linking successive activities |
The plan is a schedule, not a set of trips. Trips emerge from the need to be somewhere else at a particular time.
Where the behavioural evidence comes from
| Source | Contribution |
|---|---|
| Household travel surveys and travel diaries | Particularly important for deriving behavioural information |
| Mobile-phone, GPS, smart-card and other mobility datasets | Complementary information on observed movements |
Detailed mobility records are rarely openly available
Individual mobility records contain sensitive spatial and temporal information, so they are seldom released as complete open datasets.
This constraint shapes the whole method. Agent-based travel-demand generation therefore commonly combines different data sources:
- Population and census data reproduce the socio-demographic and spatial structure of the population
- Travel surveys and mobility datasets provide activity patterns and travel behaviour
What combining the two produces
Combining synthetic individuals with activity plans produces a synthetic travel demand, in which each agent has:
- A set of socio-demographic attributes
- A temporally and spatially explicit daily travel schedule
This representation provides the interface between transport-demand generation and agent-based transport simulation.
Individual plans can be placed onto a multimodal transport network, where agents:
- Select routes and modes
- Experience congestion and travel times
- Potentially adapt their travel decisions in response to network conditions, or to policy and infrastructure changes
Frameworks such as MATSim use this type of agent-based representation to simulate the interaction between individual travel behaviour and transport supply at large spatial scales. See transport model for that stage.
MITO
The Microscopic Transport Orchestrator (MITO) is an agent-based, trip-based travel-demand model developed at the Technical University of Munich to generate travel demand at the level of individual households and persons (Moeckel et al., 2020).
Rather than applying aggregate trip rates to zonal population groups, MITO uses a synthetic population to represent heterogeneous travel behaviour explicitly.
For each individual, the model determines:
- The number and purpose of trips
- Destinations
- Travel modes
- Timing of travel
It produces person-level travel-demand records that can subsequently be converted into daily travel plans and assigned to a transport network using a simulation model such as MATSim.
MITO's role is to bridge
MITO sits between two kinds of model that do not otherwise speak to each other: synthetic-population or land-use models upstream, and microscopic transport-network simulation downstream.
This is why it appears in the integrated SILO, MITO and MATSim chain.
Required inputs
MITO's primary input is a synthetic population containing household- and person-level socioeconomic characteristics and spatial information. Relevant attributes can include:
- Household size
- Income
- Car ownership
- Age
- Employment status
- Workplace and school locations
- Other traveller characteristics that influence travel decisions
The model additionally requires:
| Input | Purpose |
|---|---|
| Zonal activity opportunities | Destination choice |
| Transport supply information and travel-time matrices for alternative modes | Mode and destination choice |
| Household travel-survey data | Estimating or calibrating trip-generation, destination-choice, mode-choice and time-of-day parameters |
Mandatory destinations may already be fixed
Work and education destinations may be predefined within the synthetic population. Spatial information on other opportunities is required only for modelling discretionary destinations, which are the ones MITO has to choose.
How MITO generates demand: five sequential components
1. Trip generation
Monte Carlo sampling is used to determine the number of trips undertaken by each household for different journey purposes.
2. Travel-time budget
A household-specific travel-time budget is estimated according to:
- Trip frequency
- Trip purpose
- Household size
- Income
- Other socioeconomic characteristics
Why a budget
People do not travel without limit. A time budget constrains the destination choice that follows, preventing the model from assigning a series of individually plausible long trips that together would be impossible to fit into a day.
3. Destination choice
Destinations are assigned to individual trips, by two different routes:
| Trip type | How the destination is chosen |
|---|---|
| Workplace and school | Obtained directly from the synthetic population |
| Discretionary activities | Selected using a logit-based destination-choice model, constrained by the household travel-time budget |
4. Mode choice
A nested logit mode-choice model assigns a travel mode according to:
- Transport-supply conditions
- Trip characteristics
- Traveller attributes
5. Time of day
Travel times within the day are assigned, and the approach differs by trip type:
| Trip type | What is generated |
|---|---|
| Mandatory (work and education) | Preferred arrival times |
| Discretionary | Preferred departure times, using distributions derived from household travel-survey observations |
Arrival versus departure
The asymmetry is behavioural. A person needs to arrive at work by a certain time and will adjust departure to achieve it. A discretionary trip is more naturally described by when someone chooses to set off.