Conference Papers 2020

Seminar
Day 2 (10 Sep 2020), Session 5, Optimisation in Modelling, 13:00 - 15:00

Status
Accepted, documents submitted

Submitted by / Abstract owner
Luuk Brederode

Authors
Luuk Brederode
Tanja Hardt
Bernike Rijksen

Short abstract
We describe the development and application of a microscopic tour based demand model which applies a variance reduction technique that guarantees to eliminate statistical noise at any level of spatial aggregation, leaving only discretization errors.

Abstract
Fueled by the emergence of the internet and mobile devices, in the past decade, digital platforms for the usage and/or sharing of commodities and services have been developing rapidly (e.g.: Spotify, AirBnB, NetFlix, Amazon, YouTube, Software-as-a-Service products, cloud-services, etcetera). This has triggered a change in attitude towards what’s important in life. Whereas previously, owning commodities was the dominating satisfier, millennials and younger generations lean far more towards using (or ‘experiencing’) services to satisfy their needs.

This is no different in the domain of mobility where market shares of services such as Uber, Lime, MoBike and Lyft (together with traditional public transport services) are steadily increasing. It is expected that the introduction of Mobility-as-a-Service and autonomous (taxi) vehicles will further accelerate this process. An important difference with user owned modes (such as cars or bikes) is that (shared) services may be accessed/egressed (e.g. Uber) or picked up/dropped (e.g. Lime) anywhere, whereas the user owned modes always need to be picked up from- and returned to home.

Traditional strategic transport demand models are not suitable to model this increased flexibility of the (shared) services, as they generally assume that the mode choices of travelers are made before the start of a tour (i.e.: at home) which is accurate for user owned modes. However, for the (shared) services mode choice may change anywhere within the tour, because the availability and the attractiveness of the services vary per location. At the same time, these ‘within-tour’ mode choices need to meet preconditions induced by choices made earlier in the tour (e.g. a user owned mode must still be returned home) and properties of the individual traveler (e.g. whether the traveler has installed a smartphone app for and/or is subscribed to a specific service).

To be able to include the ‘within-tour’ choices, a shift from macroscopic trip based towards microscopic tour based demand models is required. In the latter approach, the discrete choices of individual travelers are modelled which allows to consider all dependencies due to previous choices and properties of the traveler.

In the typical strategic application context, model outputs of a scenario (e.g. a set of altered policy measures) are compared to model outputs of a reference case to assess the effects of the scenario. For this reason, we aim to single-out differences only caused by or related to the differences between model input of the scenario and the reference case. However, in microscopic demand models, the discretization of choices introduces statistical noise and discretization errors in the model outcomes, thereby violating this requirement. To still be able to use microscopic demand models in the strategic context, differences caused by statistical noise and discretization errors should be negligible or non-existent.

This paper describes the development of a new microscopic tour based demand model for application in the Dutch context. The model applies a newly developed variance reduction technique that guarantees to eliminate all statistical noise at any level of spatial aggregation, leaving only discretization errors (which are negligible by definition). To the best of the authors’ knowledge, no other existing microscopic demand models possess these properties.
The methodology as applied on the different discrete choice models within the tour generator and the mode and destination choice models is presented as well as its first application as the demand model within the strategic transport model system of the city of Almere.

Programme committee
Transport Models

Topic
The future of transport