New Geographic Information Systems Toolbox
by Andreas Kuhn
From now on there is a first beta version of ANDATA's new Geographic Information Systems Toolbox for selected customers and projects available.
The Geographic Information Systems Toolbox is a tool for labeling and specifically analysing trajectories from Naturalistic Driving Data, Field Operational Test Data (FOT), fitness trackers, GPS trackers, ... and assigning geographical and spatial context information to that data.Hereby sensor data and arbitrary tracking data can be combined with additional context information for and with Stipulator to prepare that data in an easy way for any further analysis like Data Mining and Machine Learning procedures within Brainer. The main field of application is primarily (but not limited to) design, development and analysis of behavioural models of arbitrary traffic participants. Such kind of models and analysis are necessary for various solutions within the development and evaluation of automated driving functions and for traffic automation. The toolbox basically is a kind of dynamic geographical information systems, where the information is not only used for static analysis but also in time series and prediction problems.
Analysis of Context Senstive Driver Behaviour
This new toolbox has many different use cases for different purposes. An example is as data source for context-sensitive driver behavior: how does a driver react when approaching a stop sign or a traffic light, or even an unsignalized intersection with a specific geometry? One can use the Geographic Information Systems Toolbox to add this kind of geographical context to the naturalistic driving data, which enables the development for much more powerful behavioural models of traffic participants. One could also use the street inclination data to model the driver reactions to ascending and descending slopes in the motorway. In the next plot one can see a trajectory where the color and altitude over the map represent the speed of a given vehicle.
Calculation of Topological Attributes
It is even possible to calculate and plot different attributes simultaneously. For example, the next plot shows the speed of a trajectory using a color map and its distance to the next intersection using the z axis.
Development of Filters and Functions for Sensor Fusion
Commercially available GPS devices can have quite large measurement errors at times. In various scenarios, e.g. in areas with dense traffic networks, this could mean that the closest way to the measured GPS trajectory is not the actual way where the driver was currently driving. Because of this, the toolbox offers smart compensation techniques enabling to correct the trajectory using information about the street network. Thus the toolbox functions find the most plausible way, instead of the closest graph.
That way tools for the development and assessment of filters and fusion algorithms on odometrie and tracking data are available.
Integration of Fitness Trackers
Of course the functionality of the toolbox is not limited to motorized vehicles. One can also use these geographical features to model behavioural patterns of pedestrians or bicycle rides. The next plot shows the GPS track of a hiking tour with the altitude as color.
The new GIS toolbox natively supports the FIT protocol, a file format explicitly created for the purpose of collecting and sharing data which stems from fitness, sport and health devices. The following picture for example shows speed readings plotted over a GPS track - both recorded via a Garmin action cam action camera, which was mounted at a motorcycle. The FIT format for example also allows an easy integration of data from fitness watches, which deliver data of heart beat and stress levels for naturalistic driving studies.
Analysis of Floating Car Data
The toolbox also makes it possible to analyze Floating Car Data (FCD) and put it in a meaningful geographical context. One can use the plot functionalities of the toolbox to analyze aggregated speeds measured from mobile phones or fitness watches, and to generate heat maps that show which streets and regions are more crowded. For example, the next plot shows a possible analysis for analyzing a database of FCD trajectories. The yellow segments correspond to areas with higher FCD densities, such as metro stations and commercial areas.
The GIS toolbox can also be used to obtain an overview of the vehicle speeds for dedicated regions. In the next plot, red, yellow and green correspond to low, medium and high speeds with respect to the maximum allowed speed (speed reserves). This means that not only single tracks can be analysed. An arbitrary number of moving objects can be evalueated and analysed with respect to regional and spatial effects.
Post-Processing for Mircro Simulations
Floating Car Data can also be generated by virtual vehicles instead of real ones, as a result of traffic micro-simulations (such like VISSIM, sumo, AIMSUM, or similar). That way the toolbox becomes a specific post-processing tool for micro simulation. A possible analysis may be the evaluation of necessary densities of FCD vehicles and/or Car2X communication data for the sufficient identification of the traffic situation or the development of sophisticated traffic control applications.
Finally, the toolbox is also useful as a statistical analysis and data mining tool to find relevant, interesting or unusual parts of the street net, such as motorway sections or intersections, in order to assess the representativeness of scenario-based simulations. Using these tools one can see for example that Vienna has a much larger proportion of 4-way crossings than any other city in Austria, where T-intersections are much more common. Such analysis can help to sort out the most relevant street topologies for the systematic simulation based development of automated driving functions. A further exemplary question may be the information about the real occurance of so-called dog curves (street curves with decreasing radius of curvature) and the distributions of their radii.
Availability and Further Developments
The toolbox will first be available for selected projects and customers as beta version. It will continuously be improved and extended with new functions. The development will primarily be driven from the lead project Connecting Austria.
Here is a brief summary of the already available and planned functions:
|Trajectory localization and correction||available|
|Extraction of trajectory attributes (street type, street curvature, elevation and grade, distance to next intersection/traffic light/pedestrian crossing/traffic sign/...)||
|Customable map and trajectory plotting||available|
|Integration with other ANDATA Tools (Stipulator, Brainer, Expectator ...)||available|
|Compatibility with Openstreetmap||available|
|Compatibility with GIP||available|
|Compatibility with FIT files||available|
|Generic module for extracting statistics and data mining||in development|
|Live access to Openstreetmap data||planned|
|Further data formats (Google, Bing, HERE, etc.)||planned|
This side contains some company and brand names of 3rd party organisations. These are linked to the initiators.