Multi-agent Service Area Adaptation for Ride-Sharing Using Deep Reinforcement Learning

Naoki Yoshida, Itsuki Noda, Toshiharu Sugawara

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper proposes a method for adaptively assigning service areas to self-driving taxi agents in ride-share services by using a centralized deep Q-network (DQN) and demand prediction data. A number of (taxi) companies have participated in ride-share services with the increase of passengers due to the mutual benefits for taxi companies and customers. However, an excessive number of participants has often resulted in many empty taxis in a city, leading to traffic jams and energy waste problems. Therefore, an effective strategy to appropriately decide the service areas where agents, which are self-driving programs, have to wait for passengers is crucial for easing such problems and achieving the quality service. Thus, we propose a service area adaptation method for ride share (SAAMS) to allocate service areas to agents for this purpose. We experimentally show that the SAAMS manager can effectively control the agents by allocating their service areas to cover passengers using demand prediction data with some errors. We also evaluated the SAAMS by comparing its performance with those of the conventional methods.

Original languageEnglish
Title of host publicationAdvances in Practical Applications of Agents, Multi-Agent Systems, and Trustworthiness. The PAAMS Collection - 18th International Conference, PAAMS 2020, Proceedings
EditorsYves Demazeau, Tom Holvoet, Juan M. Corchado, Stefania Costantini
PublisherSpringer
Pages363-375
Number of pages13
ISBN (Print)9783030497774
DOIs
Publication statusPublished - 2020
Event18th International Conference on Practical Applications of Agents and Multi-Agent Systems, PAAMS 2020 - L’Aquila, Italy
Duration: 2020 Oct 72020 Oct 9

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12092 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference18th International Conference on Practical Applications of Agents and Multi-Agent Systems, PAAMS 2020
CountryItaly
CityL’Aquila
Period20/10/720/10/9

Keywords

  • Deep Q-networks
  • Multi-agent learning
  • Ride-share service

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Multi-agent Service Area Adaptation for Ride-Sharing Using Deep Reinforcement Learning'. Together they form a unique fingerprint.

Cite this