volume-01-Issue 1 (2019)
Latest Articles
Modeling framework for supporting taxi policy making
JTTM, volume-01, Issue 1 (2019) , PP 27 - 36
Published: 26 Apr 2019
DOI: 10.5383/JTTM.03.01.004
by Josep Maria Salanova Grau, Miquel Estrada from Centre for Research and Technology Hellas/Hellenic Institute of Transport, Thessaloniki, Greece, 57001 Technical University of Catalonia, Barcelona, Spain, 08034
Abstract: Taxi services account for a significant part of the daily trips in most cities around the world. These services are regulated by a central authority, which usually monitors the performance of the taxi services provision and defines the policies applied to the taxi sector. In order to support policy makers, fleet managers and individual taxi drivers, there is a need for developing models to understand the behavior of these markets. Most of the models developed for analyzing the taxi market are based on econometric measurements and do not account for the spatial distribution of both taxi demand and supply. Only few simulation models are able to better understand the operational characteristics of the taxi market. This paper presents a framework for the development of taxi models both aggregated and simulation-based. It is aimed at assessing policy makers, taxi fleet managers and individual drivers in the definition of the optimum operation mode and the number of vehicles. read more... read less...
Keywords: taxi modeling, agent-based modeling, modeling framework
Evaluating Reinforcement Learning State Representations for Adaptive Traffic Signal Control
JTTM, volume-01, Issue 1 (2019) , PP 19 - 26
Published: 16 Apr 2019
DOI: 10.5383/JTTM.01.01.003
by Wade Gendersa, Saiedeh Razavi from Department of Civil Engineering, McMaster University, Hamilton, Canada, L8S 4L8, McMaster Institute for Transportation and Logistics, Hamilton, Canada, L8S 4L8
Abstract: Reinforcement learning has shown potential for developing effective adaptive traffic signal controllers to reduce traffic congestion and improve mobility. Despite many successful research studies, few of these ideas have been implemented in practice. There remains uncertainty about what the requirements are in terms of data and sensors to actualize reinforcement learning traffic signal control. We seek to understand the data requirements and the performance differences in different state representations for reinforcement learning traffic signal control. We model three state representations, from low to high-resolution, and compare their performance using the asynchronous advantage actor-critic and distributional Qlearning algorithms with neural network function approximation in simulation. Results show that low-resolution state representations (e.g., occupancy and average speed) perform almost identically to high-resolution state representations (e.g., individual vehicle position and speed) using fully connected neural networks, but deep neural networks with highresolution state representation achieve the best performance. These results indicate implementing reinforcement learning traffic signal controllers in practice can be accomplished with a variety of sensors (e.g., loop detectors, cameras, radar). read more... read less...
Keywords: adaptive traffic signal control, deep reinforcement learning, intelligent transportation systems, applied machine learning, transportation simulation, neural networks
Dynamic, multi- and intermodal bike sharing in agent-based modelling
JTTM, volume-01, Issue 1 (2019) , PP 09 - 17
Published: 13 Apr 2019
DOI: 10.5383/JTTM.01.01.002
by Cornelia Hebenstreit, Martin Fellendorf from Institute of Highway Engineering and Transport Planning, Graz, University of Technology, Graz, Austria, 8010
Abstract: Negative effects of traffic, like congestion, air and noise pollution are among the reasons why environmentally friendly solutions are promoted. Bike sharing (bs) is intended to strengthen cycling and public transport. Nevertheless, current transport models rarely consider cycling or even bs, in either detail or holistically. In this paper we present an agent based approach to model cycling and in particular bs within the multimodal simulation environment MATSim. Multimodal trips combining public transport and bs are included as well as within day rescheduling of bs trips as agents may not find a bike or empty return space (parking spot). To minimize such cases, choice probabilities were implemented, so that agents only start their bs trip, if sufficient bikes or parking spots are available. The modules presented in this paper were applied using a MATSim model of the city of Vienna. Agent based bs modelling is an inexpensive option to test the impact of a bike sharing system before implementation. read more... read less...
Keywords: Bike Sharing, Transport Simulation, MATSim, Route Choice, Agent-based Simulation, Intramodality
The matching problem of empty vehicle redistribution in autonomous taxi systems
JTTM, volume-01, Issue 1 (2019) , PP 01 - 08
Published: 10 Apr 2019
DOI: 10.5383/JTTM.01.01.001
by Tatiana Babicheva, Wilco Burghout, Ingmar Andreasson , Nadege Faul from VEDECOM, 23 bis Allée des Marronniers, 78000 Versailles, France, KTH-Royal Institute of Technology, 100 44 Stockholm, Sweden, LogistikCentrum AB, Osbergsgatan 4 A, 42677 V.Frolunda, Sweden, DAVID, Université de Versailles Saint-Quentin-en-Yvelines, 55 Avenue de Paris, 78000 Versailles, France
Abstract: This article discusses empty vehicle redistribution algorithms for PRT and autonomous taxi services from a passenger service perspective. In modern literature reactive methods such as nearest neighbours are commonly used. In this article we first formulate the general matching problem on a bipartite graph of available vehicles and stations. In addition, we propose an index-based proactive redistribution (IBR) 18,19 algorithm based on predicted near-future demand at stations. The results of different redistribution methods implemented on a simple line test case show that none of the proposed methods are optimal in all cases. Test results of six variations of combined proactive and reactive strategies on a test case in Paris Saclay, France with 20 stations and 100 vehicles are given. The combined Nearest Neighbour / IBR provides a promising solution for both peak and off-peak demand, significantly outperforming all other methods considered, in terms of passenger waiting time (both average and maximum) as well as in terms of station queue lengths. read more... read less...
Keywords: Empty Vehicle Redistribution, Fleet-size, Autonomous Taxi, Matching Problem