Effects of autonomous first- and last mile transport in the transport chain

Mikkel Thorhauge*, Anders Fjendbo Jensen, Jeppe Rich

*Corresponding author for this work

Research output: Contribution to journalJournal articleResearchpeer-review

58 Downloads (Pure)


In this paper we explore user preferences with respect to autonomous shuttle buses when these operate as a first/last mile feeder mode to the main public transport system. We analyse the extent to which improvements of the first- and last mile public transport, through the use of autonomous door-to-door shuttle services, can be expected to affect the choice of mode for the entire transport chain in comparison with choice of mode for the first- and last mile trip. The campus of the Technical University of Denmark (DTU) was used as a case study site, and pivoted stated choice experiments were developed as part of a real-life experiment where several autonomous buses were operated at the DTU campus. In the experiment, choice tasks were directed to users that specifically used the shuttle service. Our findings indicate that while AV shuttles did indeed attract some existing public transport users, almost no effect on the overall mode shares was observed. This even holds for scenarios where speed and frequency of the buses were improved considerably compared to those of the experimental setup. The paper thereby underlines that while new autonomous driving technologies have the potential to improve first and last mile services for public transport, they will have a limited effect on the overall choice of mode for the entire trip chain in their current form.

Original languageEnglish
Article number100623
JournalTransportation Research Interdisciplinary Perspectives
Publication statusPublished - 2022


  • Autonomous shuttle buses
  • Demand modelling
  • First/last mile solutions
  • Mode choice
  • Public transport systems


Dive into the research topics of 'Effects of autonomous first- and last mile transport in the transport chain'. Together they form a unique fingerprint.

Cite this