Extending KNIME with Floating Car Data Analytics Capabilities (dataset and materials)

Abstract
The ever growing availability of high resolution mobility data has triggered the development of a number of data-driven solutions, leading to a significant improvement in the body of knowledge on Intelligent Transportation Systems (ITS). Nevertheless, to date, an ITS practitioner willing to perform data analytics studies has to face a number of technological challenges, due to a plethora of different data formats. Indeed, while in other data-driven domains a number of well-established tools, such as KNIME or RapidMiner, is available to support the definition of Knowledge Discovery from Data (KDD) pipelines, when dealing with spatio-temporal data, a lot of steps have to be manually implemented, significantly hindering productivity.

To address this issue, we propose a solution we developed to support ITS practitioners in the definition of KDD processes on mobility data. Indeed, by exploiting the modular capabilities of the KNIME Analytics Platform, we developed a collection of new components specifically designed to automatize some standard KDD steps in the ITS domain, such as map-matching, trajectory partitioning, flexible routing algorithms, and map coverage analysis.
To show the effectiveness of these components, we report also on how we applied it on a real-world massive trajectory dataset.

All the components we developed are open-source and freely downloadable, as we hope that they could further foster the data-driven ITS research.

Contents
This item includes supplementary materials and a KNIME workflow to reproduce our experiments measuring the road-network coverage achieved by a fleet of taxis in the Municipality of Rome using the custom KNIME extension we developed.

The OSRM-Ubuntu-20.04LTS.ova file is a virtual machine image which can be executed using VirtualBox. The image comes with an OSRM instance capable of serving routing requests over the Municipality of Rome. The credentials are:

Username: User
Password: user

After login, to start the OSRM instance it suffices to open a terminal and run the following command from the home directory:

osrm-routed --max-matching-size 3000 OSRM-data/rome/rome.osrm

The Rome-Taxi-Analysis.knwf file is a KNIME workflow exported using KNIME 4.3. Before opening the workflow, make sure that the custom KNIME nodes we developed are properly installed.

The taxi_february.txt file is a csv dataset of Floating Car Data recorded in the Municipality of Rome. This file must be specified as an input to the CSV Reader node in our workflow. For more information, see:
Lorenzo Bracciale, Marco Bonola, Pierpaolo Loreti, Giuseppe Bianchi, Raul Amici, Antonello Rabuffi, CRAWDAD dataset roma/taxi (v. 2014‑07‑17), downloaded from https://crawdad.org/roma/taxi/20140717, https://doi.org/10.15783/C7QC7M, Jul 2014.

The rome.osm file is the OpenStreetMap file we used for Map Matching. This file is to be supplied to the Map Matcher node. If you use a different version, coverage results might be slightly different.

The knime-nodes.zip archive contains the KNIME nodes we developed. These nodes can be installed by copying the correspoding JAR files in the /dropins directory of a KNIME installation. Check out the dedicated repository at https://luistar.github.io/knot/ for newer releases.

Instructions

To reproduce our results, we recommend the following procedure:

1. Install KNIME 4.3, the install the custom nodes we developed as instructed in https://luistar.github.io/knot/.
2. Download and run the OSRM-Ubuntu-20.04LTS.ova virtual machine, the start the OSRM instance as shown above.
3. Download the taxi_february.txt dataset and the rome.osm file.
4. Make sure that the nodes are configured properly:
   1. The CVS Reader node should read the taxi_february.txt file;
   2. The Map Matcher node should be configured to read the rome.osm file, and be given a correct address to send https requests to the OSRM instance running in the virtual machine. In a simple configuration in which the Virtual Machines runs on the same machine running KNIME, the ifconfig command can be run within the VMto find out its local IP address, and that local IP address can be specified in the Map Matcher configuration dialog.