New Generation Activity-Based Travel Demand Models: From Trips To Lives

For decades, transportation planning has relied on models that simplify an extraordinarily complex question:

How will millions of different people respond when we change the transportation system?

A new generation of travel demand modeling is pushing toward much greater behavioral detail.

Instead of representing populations primarily through aggregate groups and snapshots, researchers are developing complex synthetic populations that can evolve over time, incorporate household and social relationships, and feed into increasingly sophisticated activity-based models.

At the same time, advances in discrete choice modeling and computation are making it possible to model decisions involving enormous numbers of alternatives—bringing travel demand models closer to the complexity of the choices people actually face.

What does this mean for the future of transportation modeling?

At Mobility Forum, I spoke with Michel Bierlaire, Professor at EPFL, about this new generation of disaggregate travel demand models (listen to this episode on Spotify) . This memo is a deep dive of our discussion on three questions:

1. The four big shifts that are changing travel demand modeling: from modeling trips to understanding people’s days, households, and lives
2. How can we model the enormous complexity of people’s daily choices, relationships, and lives? Once we model human behavior in this much detail, the number of possible choices explodes into the billions. How do we make that computable, what data do we need, and how do we know the results are good enough?
3. What does all of this mean for transportation planners? How Swiss Railways put these ideas to work, and what MPOs, DOTs and transportation authorities should consider as they make their next modeling investments.

1. Why a New Generation of Travel Demand Models? Four Shifts in Behavioral Complexity.

The infographic below illustrates four fundamental shifts in how the next generation of activity-based travel demand models represents behavioral complexity—and what these shifts mean for transportation planning.

The value of richer models lies in the scope of questions they can address, not in predictive precision. The added detail captures behavior relevant to the planning and investment decision. The true challenge is in asking the right question.

Complexity Shift 1: From Modeling Single Trips to Modeling An Entire Day

Three example questions a trip-based model cannot answer:

1.  Remote work changes where activities take place, when people travel, and how the rest of the day is organized.

2.  At EPFL and the university next door, classes start at the same time, and the public transportation operator asked whether EPFL could shift its start time to spread the morning peak.

3. In Switzerland the cantons set shop closing times by law, typically between 18:30 and 20:00 on weekdays with Sundays closed, and voters have repeatedly refused to loosen them. What would a later closing hour do to the evening peak?

Answering these questions requires the model to represent the day as one schedule of activities and trips: each choice changes what remains possible for the rest. E.g. someone might shorten lunch to finish work earlier and play tennis in the evening.

Preferences matter, but so do constraints: a day contains 24 hours, some activities are mandatory, and a car may be unavailable. The model represents both what a person wants and what that person can do.

Modeling the trips separately misses the time contraint that connects them. A full-day schedule makes these connections explicit.

The approach has existed since the late 1990s; the frontier is how completely the activities and their connections are represented.

Complexity Shift 2: From an Assumed Decision Hierarchy to Joint Decisions at the Same Level

The 1990s’ activity-based models manage complexity by arranging choices in sequence: primary activities first, then secondary activities, and then destinations, modes and time.

Each step is manageable, but the sequence dictates which decisions come first. In reality, most of the time people do not follow this decision sequence.

In particular, this decision hierarchy was shaken by COVID and working from home. The commuting trip may no longer be the anchor of a day. In Switzerland, leisure, not work, is already the largest trip purpose according to census.

The new approach places the decisions at the same level. Constraints define the feasible schedules, preferences rank them, and observed behavior shows how the choices fit together. Letting the data decide means not prescribing a decision order in advance.

This is not a claim about the psychological process. People do not enumerate every possible day schedule, and each person decides differently. The objective is only to predict behavior under specified conditions while avoiding a hierarchy that may be wrong.

Complexity Shift 3: From the Isolated Individual to Households and Social Networks

A person’s feasible day depends on other people. In a household with one car, one member’s decision to drive removes that option for another. Someone must take the children to school. We want to enjoy dinner together. Shared resources and obligations make individual choices interdependent.

Joint activities require compatible arrival times. Shared cars and obligations are shared constraints. And the outcome is bargained among the members, not chosen by one of them.

These interactions change both constraints and objectives. A household member may accept an inconvenient trip to improve someone else’s schedule, which an individual model misses.

The model must consider compatible combinations of schedules across people.

Beyond the household, colleagues coordinate commutes and friends coordinate leisure. Identifying the relevant relationships is part of understanding demand.

Complexity Shift 4: From a Static Population Snapshot to Synthetic Population with Life Histories

Important life decisions such as moving home, changing jobs, forming a family, and buying a car shape daily travel over years.

A travel model carries those choices in its synthetic population: artificial individuals who reproduce the statistical population without corresponding to any real person.

Separate snapshots lack continuity. A population synthesized for 2025 and another for 2030 may each match its year’s statistics without representing the continuity between the two.

The alternative is to synthesize the dated life histories: birth, education, work, residence, driving license, car ownership. The same person is followed from one survey to the next.

The model links trips to the life events that change them for the same people over time: people decide whether to live near transit when they move; whether to buy a car when a child arrives, and those arrangements persist for years. Remote work changes the trade-off between commute distance and frequency: a more distant home becomes attractive when the commute is less frequent.

The life history of shift 4 determines where someone lives and works, and shifts 1 to 3 decide the day inside those constraints.

2. Managing the Explosion of the Choice Set: Method, Data, Validation

Richer representation is only useful if it can be computed, supported by data, and checked for validity.

All four shifts in behavioral complexity expand the choice set, and they compound to billions of alternatives.

A single trip involves a handful of modes, and a destination choice model may have a few hundred options. A full day schedule adds which activities to perform, in what sequence, where, when, for how long, and by what mode, and the combination quickly explodes.

In the 1990s activity based model, the assumed decision hierarchy reduces the complexity substantially: a model that fixes primary activities before secondary ones never considers most combinations.

However, a model that decides everything at the same level keeps all of them in the choice set: removing the hierarchy removes the pruning.

Households multiply again, because only compatible combinations of members’ schedules are feasible, and feasibility depends on shared cars, shared obligations, and shared arrival times.

Life histories add sequences of events over decades: where someone lived, worked, and studied, and when.

The first three shifts grow the day; the fourth shift grows the population.

The Technical Core: Formulate - Optimize - Simulate - Estimate

Listing billions of combinations is impractical.

Discrete choice modeling supplies a theory of preferences and probabilistic behavior, and methods for learning from observed choices.

Combinatorial optimization supplies ways to describe and search enormous sets of feasible alternatives. It is the same mathematics behind timetabling, vehicle routing, and airline operations.

Step 1: Formulate codifies the rules that define the feasible set. Instead of listing every feasible daily schedule, write the rules that represent the behavioral relationships, including

1. The constraints a feasible day must satisfy: 24 hours, mandatory activities, shared car, dinner together, etc, set by the life history.
2. An objective function representing preferences.

Constraints plus preferences form an optimization problem.

How to write a household’s preferences remains contested: summing the members’ utilities, or maximizing the best-off or the worst-off member, and the framework accepts any of them without settling how families negotiate. Attitudes enter the utility through hybrid choice models with latent variables, whenever suitable data exist.

Step 2: Optimize identifies the best day schedule by searching the space efficiently. Branch-and-bound divides the problem into slices and uses bounds to discard slices that cannot contain the best solution.

The first versions ran on standard solvers, Gurobi and CPLEX. Sometimes they are not enough, so decomposition methods, Benders decomposition and column generation, are now being developed: they split the model into sub-models that are easier to solve, then put the solutions together.

The successes in optimization come from adapting algorithms to the specific structure of the problem.

Step 3: Simulate adds random utility, from one specific schedule to a distribution.

A deterministic optimization yields one preferred schedule.

Random utility adds variation: draw realizations of the random components, solve the optimization for each draw, and repeat, perhaps thousands of times.

The result is a distribution of schedules. The draws are independent and run in parallel, but the computation remains demanding.

Step 4: Estimate learns the preference parameters by comparing the schedule a person actually chose with alternatives they could have chosen.

There are billions of feasible schedules, far too many to list. Most would never be attractive and teach the model nothing; the informative ones are those that compete with the observed schedule.

Metropolis-Hastings, a Markov chain Monte Carlo method, draws alternatives from the combinatorial space so that competitive schedules are drawn more often.

This introduces a bias, but since each alternative's chance of being drawn is known, importance-sampling corrections remove it in estimation.

The Synthetic Population: Prior and Update

The population is built in two layers.

First, demographic models give a starting guess, the prior: when people are born, study, start work, move, get a driving licence, and buy a car.

Then each survey wave corrects that guess (Bayesian updating). In Switzerland the survey is the mobility microcensus, taken every five years. Each survey captures only one year, yet the result is a population in which the same individuals appear in every year: life histories of the population.

One caveat: these life histories are built to match the statistics, not to explain them. They reproduce when people move, change jobs, or buy a car, but not why. This is a deliberate choice: the causes of these long-term decisions are rarely observed and are beyond the scope of the model.

From Algorithms to the Data

Better algorithms let the model represent more behavioral richness. But a model that decides the whole day also needs data on the whole day: what people did, what limited them, and what else they could have done.

The usual source is a travel diary or time-use survey, in which people report their activities and the trips between them. Switzerland's mobility microcensus is one. Today, phones make these diaries easier to fill in.

Relationships are harder to capture. The census tells us who lives in a household, but not who people's friends and colleagues are. That takes a separate survey. The "snowball" method asks each respondent to name their contacts, then surveys those contacts in turn. It finds connections, but the sample is heavily biased. And online friends are not always the friends people actually travel with.

GPS traces, transit smart cards, connected cars, and payment records capture movement on a scale no survey can match: where people went, when, and how often. They show what people did, but not why. They do not record the purpose of a trip, what limited the options, or why someone turned an option down.

Transportation research is data-rich and data-poor at the same time.

Start from the question, not from the dataset.

Privacy adds another limit. A detailed location history reveals where someone lives and works, even without a name. In Europe, the General Data Protection Regulation (GDPR) has made such data harder to use.

When Is a Synthetic Population Good Enough?

When it does what it is supposed to do.

The goal is not to copy the real population. If it were, we could just use the real one; the whole point of a synthetic population is that no real person is in it. So there is no real population to test it against, and the usual out-of-sample test does not apply. The test is different, and it has three steps.

First, check that the basic distributions, such as age and household size, and the key relationships between them match the real population.

Second, test it on the question it was built for. Feed the synthetic people into the model that will use them, such as a traffic simulation, and check that they include the kinds of people the question depends on. A study of a new night bus needs enough shift workers; a study of access for older residents needs enough older people without a car. Then change the scenario, for example raise the fare, and check that the results still make sense.

Third, if the test shows something missing, add it and test again. For the night-bus study, that could mean giving each synthetic person their working hours. A population is good enough for a purpose, not necessarily for every purpose.

The model's predictions of behavior are tested separately, against survey data held back from the model or, where available, data that follow the same people over years. Transportation agencies should be clear about what they want to check: the makeup of the population, today's behavior, or performance over time. Passing one test does not mean passing the others.

3. From Research to Application: Swiss Railways

Swiss Federal Railways (SBB) faces an interesting problem: Peak-hour trains are packed. Yet across the full day, the railway uses only about 28% of its capacity.

So how to bring more passengers onto trains during off-peak hours? This requires SBB to understand the full chain of activities and trips so that the railway can make long-term infrastructure and service decisions. For example, Swiss Railways has considered what would happen if trains operated more like a metro, with service on every line perhaps every 15 minutes.

SBB can’t answer that by looking at individual trips alone. They need to understand people’s entire day and activities behind it - not just commuting. Why are they traveling? What are they doing before and after the train ride? How do they get to the station? What happens if service changes?

SBB had already built its own activity-based model and uses MATSim to understand these broader travel patterns. Michel and his team worked with Swiss Railways to test their newer methodology—using combinatorial optimization to handle the enormous number of possible activity and travel choices within the railway’s existing modeling tool. The project was successful: the model reproduced the observed distribution of activities satisfactorily.

The results successfully reproduced observed activity patterns.

SBB understood the value of activity-based modeling. New modeling methods become useful when they meet a real planning question, an existing modeling foundation, and an agency capable of using the results. The new methodology added capability where it mattered.

What should MPOs, DOTs and transportation agencies do?

Do not do a head-to-head comparison of a four-step and an activity-based model on the same question as the two answer different questions. Traditional models may remain appropriate for the questions they were built for. Start by reviewing what the newer models can do and judge whether that is relevant to your questions. More elaborate models earn their place through the questions they can address and the quality of the resulting analysis. Connect activities, people, and time, decide them together, and the model shows what a policy does to people's lives, not only to trips.

For an agency planning its next data and model investment, I suggest five steps.

The table below applies the five steps framework to one Swiss example, shop closing hours. Each canton sets closing times by law, typically between 18:30 and 20:00 on weekdays, and shops stay closed on Sundays. Voters have repeatedly refused to loosen these rules.

— Jinhua

Listen to our full 46 mins conversation on Mobility Forum Podcast: