Combining immediate customer responses and car–passenger reassignments in on-demand mobility services

Marvin Erdmann, Florian Dandl, Klaus Bogenberger

Research output: Contribution to journalArticlepeer-review

8 Scopus citations


This paper presents a two-step information system for operating an on-demand mobility (ODM) service using the advantages of both immediate responses based on heuristics and global (re)optimization of car–passenger assignments. Service providers using such a model can offer a better user experience due to shorter response times, while they also benefit from the increased profits made possible by the global optimization of assignments. This study compares two immediate response strategies (IRSs) in terms of their system performance and individual ability to correctly predict customers’ pickup time windows. It also compares the key performance indicators (KPIs) of global reassignment optimization without immediate responses and several constrained cases in which customer acceptances and rejections are handled by IRSs. Ten combinations of IRSs and service model variations are tested in simulations using the ODM demand data from New York City taxis obtained over one week for varying fleet sizes of between 1,000 and 6,000 vehicles. The results show that in general, the list-based assignments (LBA) approach outperforms the nearest neighbor policy (NNP) as an IRS in most of the scenarios evaluated with respect to KPIs, such as requests served and profit generated for the service provider, while it also produces more empty vehicle mileage and longer customer waiting times. The pickup time window predictions of both LBA and NNP were correct in 68% to 72% of cases in scenarios in which no constraints are induced by the IRS. It was also found that global (re)optimization of assignments helps to improve the profit generated for the service provider, especially if the decision as to which requests are accepted is made during global optimization rather by the IRS. However, such a service model would imply an average customer response time of half the optimization period, which was set to 30s in this study compared to the immediate responses given when using an IRS.

Original languageEnglish
Article number103104
JournalTransportation Research Part C: Emerging Technologies
StatePublished - May 2021


  • Car–passenger assignments
  • Dial-a-ride problem
  • Immediate response
  • List–based assignments
  • Nearest neighbor policy
  • On–demand mobility


Dive into the research topics of 'Combining immediate customer responses and car–passenger reassignments in on-demand mobility services'. Together they form a unique fingerprint.

Cite this