The use of GPS data in an IoT context was illustrated by Dubravko Dolic from Continental Reifen Deutschland GmbH at the Data Festival 2018. Dubravko Dolic, Lead Architect Advanced Analytics at Continental Tires, provided insights into the handling of GPS data at Continental in his presentation “Working with GPS data in IoT environment from the data science perspective”. The presentation included use cases of GPS data – one of the most common data types within IoT-scenarios – and its handling. Moreover, Dolic Dubravko presented measures to be taken to validate and improve the accuracy of GPS data.
Retrieving GPS data from sensors
Tires are both an important and expensive part of a vehicle; therefore, Continental started collecting data from sensors mounted right in the tires or the vehicle in order to develop a tool for fleet management. Two types of data can be collected from a tire sensor: information on temperature and pressure. This data is then enriched with the vehicle’s CAN Bus data, GPS data and acceleration data collected by an IMU sensor installed on the FlexBox, Continental’s data logger.
An in-situation view on temperature and pressure data is provided to the fleet managers, who can thus easily recognize changes in temperature or pressure. In order to extend this current system, Continental implemented a complete system, the so-called integrated fleet system, consisting of an intelligent tire that is connected to a Tire Management Portal and Tire Service.
The detailed function of this tool is illustrated in this video; in a nutshell, it works as follows:
- temperature and pressure data are collected by a by a sensor mounted on the tire inner liner.
- this data is transferred to the so-called FlexBox via the vehicle’s telematic system.
- Once this data is enriched with GPS data or acceleration data, it is either transferred to a third-party backend or the Continental FlexBox backend.
Providing GPS data to analysts
So far, only a limited number of vehicles has been equipped with sensors for data collection at Continental in order to test the FlexBox system. The box is piloted in the automotive department of the company and the retrieved data is then analyzed by Dolic’ advanced analytics team. Nevertheless, the GPS data, IMU and TTM-CPC data of those few vehicles add up to a large amount of information over time. In order to handle this pile of data in an adequate manner, different tools have been tried and are partly still being used.
GPS data: tools used at Continental
PostgreSQL and Tableau
Right at the beginning, the raw data was uploaded to PostreSQL (postgres) – which worked quite smoothly. However, in terms of data analysis, the tool is not particularly suitable, as its functions are limited and it cannot be used to visualize more than one track at a time.
In the next step Tableau was used to visualize and plot tracks – the data was mainly retrieved from CSV files. For the generation of quick plots of tracks as well as conducting quality checks, the use of Tableau is beneficial. However, it is not very suitable to perform in-depth analytics, since data analysis in Tableau requires extensive (programming) know-how. Only that those experts with the necessary know-how prefer to use tools other than Tableau for the purpose of data analysis. Therefore, Tableau was not the number one choice for data management. On the plus side, it can be said that its use does not require any expertise in GIS and can therefore be performed by a broad range of users.
For the use of QGIS, knowledge of GIS (Geographic Information System) is in fact necessary. The tool is a good choice for realizing GIS operations on a single dataset of tracks, such as for simplifying tracks, measuring distances or identifying points of interest. At Continental, it is widely used to create visuals for the engineering department. Data can be accessed in different ways:
- via CSV files, which is almost always a reliable method.
- via PostGIS – access can be granted using the PostgreSQL database via PostGIS.
- via Shapefiles, which offer further geographical information such as maps.
R and Shiny are used for generating special reports and serve as a layer between data science and the business perspective. Business users can work with apps developed by programmers. In the context of GPS data, the programming language R has one major advantage: R supplies a range of geography libraries to choose from and implement into the code. Defined reports visualize the information that engineers need to access. A ready-to-use dashboard and maps with overlays in R and Shiny can be made available to engineers. R and Shiny can be used by accessing the data either via data streaming, HDFS or PostgreSQL.
PostGIS offers several functions for handling GPS data: It can be used to automate GPS data processing, provide geometry objects by creating connection lines between single points to aggregate spatial data or for geometry indexes. In addition, it is a powerful tool to realize simplifications.
Simplification of tracks
Tracks can be simplified by reducing the number of points and by that convert tracks into paths. Data gaps need to be deleted and useful groups of units, such as minutes or days, should be built. Simplification can be achieved using two main methods:
- simplification by sampling (e.g. only use every fifth point)
- simplification by algorithm (e.g. , used only for visualizing tracks on maps in R or QGIS)
This simplification function is important in order to reduce the amount of data and thereby decrease loading time. Simplified tracks can be converted into paths, which are saved and provided to users directly. However, despite all its advantages, simplification is not suitable for a detailed analysis of graphs.
Providing spatial data
Spatial data at Continental is provided in several steps. First of all, GPS data is created, collected and transferred: All raw data is saved on Amazon Glacier in order to conserve it. The data in its unprocessed status is not yet ready to be analyzed, however. The next steps are the distribution and storage of data, but only after the preprocessing stage can the data actually be used. There are several options for preprocessing:
- aggregation of point data.
- alignment of data with an IMU – inertial measurement unit – an electronic device that can measure several data of a vehicle, such as rotation and acceleration: the aggregation of GPS data and acceleration data is important, as only in combined form they reveal interesting results. Aggregation could be realized using a time stamp in R. Further details on aligning data are available in the presentation video below.
- enrichment with weather data: an example of an HTTP is provided in the video.
The preprocessed data is stored on-premises to allow direct access for data scientists in order to analyze it. Finally, the data is provided for mapping.
Sometimes the data delivered by the systems are not fully accurate. However, certain measures can be taken in order to improve the level of accuracy:
- DOP (dilution of precision) values as quality markers: the level of accuracy can be improved by deleting values that exceed certain limits, by scrutinizing the values and keeping only those up to a certain DOP value.
- Which data can be used for gaining speed data? Speed data can be derived from GPS data by calculating values from longitude and latitude coordinates.
- Distance measures: distance measures derived from GPS are often suboptimal, exhibit gaps and underestimate vehicle distance. A solution to this issue is still being developed by Continental – currently the team is working on interpolation to correct the inaccuracies.
- Acceleration: in some cases, the data measuring device is not installed on a plane surface in the car. As a consequence, erroneous data is produced. Especially data visualization in a coordinate system reveals this issue. In order to rectify this, Continental developed a mechanism that identifies incorrect values and corrects them automatically.
Learn more about the use of GPS data directly from the presentation and discover hacks to solve challenges such as data accuracy. Dubravko Dolic’ complete presentation is available here: