The National Transport Model

The National Transport Model - Introduction

The National Transport Model (Monty) creates a digital representation of people’s daily travel in New Zealand by modelling individual agents and their decisions. In real life, people travel to carry out their daily activities. People consider the trade-offs of travelling at a certain time, using particular travel mode, via a specific route, optimising how they get from place to place based on prior experience and constraints. For example, someone may be a frequent car driver except for their commute to work, where peak congestion makes driving too frustrating. Monty models this decision-making process, representing people as agents, each with their own plausible activity schedules to carry out. The idea behind this approach is that you can:

• generate a representative population of agents for New Zealand (each with their own combination of plausible attributes like age, gender, job and household structure);
• provide this representative population of agents with plausible activity schedules that mimic patterns of activities seen in real life; and
• have all these agents interact with one another in a realistic manner across the transport system as they attempt to fulfil their scheduled activities.

With these components together, we have a means of simulating not only where people are travelling from and going to, but also why they might be making specific decisions and what factors may affect certain groups of people more than others.

Monty divides the modelling process into three main behavioural concepts, providing examples of external influences at each stage. These are described as follows:

• Human Decision-Makers. At the centre of the simulation are individual humans (represented as agents within the simulation), each of which has their own characteristics that influence their desires and needs. For example, agents under the age of 18 will (in most cases) be predominantly concerned about their ability to travel to and from primary and secondary school by some means. Those aged older will demonstrate a much broader variety of concerns depending on whether they are working, completing tertiary study, or performing some combination of archetypical activities.
• Daily Activity Tasks. Just as humans have basic desires and needs, this manifests in the requirement to perform certain tasks over the course of a given day. Those studying must access their place of study, just as those working must commute. Some types of activity are required to start at very specific times whilst others (like shopping) may be more flexible. As such, activity patterns can be influenced through several means including individual socio-demographic attributes, household relationships as well as accessibility to opportunity and available travel options. Additionally, activity patterns can also depend on advances in technology and policy (e.g., autonomous driving or flexible working arrangements).
• Travel Experience. Building on this need to undertake activities, individuals need to travel from one activity to the next whilst considering planned activity starting times. This act of travelling forms a travel experience that is influenced by interactions with other individuals as well as the capacity constraints of the transport system. For example, a road segment might often be congested in the morning peak, leading to severe journey delays for travellers taking this route. This can have a major impact on the individual’s routing decisions and even their chosen activity schedule in certain instances. As a result, some individuals may end up changing their route in response or attempt to travel at a different time. As such, the process is iterative, with individuals learning and adapting their patterns over time.

The aim for Monty is to accurately capture this decision-making hierarchy for agents within a given scenario. Monty achieves this through the orchestration of four distinct components, which map on to the behavioural concepts.

• Network generation. This is responsible for generating the transport network supply for a given scenario. This includes a detailed road network representation as well as public transport timetables for the entire country. This “model” brings together the complex physical world that includes detailed multi-modal networks (incorporating walking, cycling, public transit, different vehicle types), scheduled transit operations, realistic building/activity locations etc.

• Population synthesis. This is responsible for creating agents and their attributes (such as age, gender, employment status etc.) for the entire country of New Zealand. The resulting population of agents is representative of New Zealand’s real population in aggregate, without any individual agent corresponding with a real-life person.

• Activity synthesis. The Activity based model is responsible for assigning agents generated during population synthesis with a synthetic activity schedule. Thus, after both population synthesis and activity synthesis have been used, the result is a population of agents that each possess a plausible set of demographic attributes as well as a schedule of activities to perform during the modelled day.

• Network simulation. This is the Agent based part of Monty, and is responsible for facilitating interactions between all agents and their activity schedules, considering factors such as network capacity constraints. Monty brings all of these modelling elements together under a software framework called Multi Agent Transport Simulation (MATSim)(external link). Through the simulation, agents can modify their activity schedules and decision-making surrounding route choice and mode choice.

More information can be found in our technical development report(external link) and in various presentations(external link) that the team have given.