The assessment of real-world effectiveness (i.e. the impact on traffic safety) of advanced driver assistance systems (ADAS) or automated driving functions (AD) is becoming increasingly important. Virtual simulations are one of the most comprehensive methods to assess multiple research questions regarding this topic now and in the future. The simulation should be able to run a variety of traffic scenarios which depend on human behavior as well as on system behavior. Stochastic algorithms are main requirements to vary situations or to represent uncertain decision making processes. Nonetheless, the simulated situations must be reproducible for an individual in-depth analysis of critical situations. Lastly, a highly flexible interface is necessary to cope with interactions of to-date unknown models.
Further information regarding methodological background is provided, for example, by the P.E.A.R.S. initiative.
Eclipse sim@openPASS provides a software platform that enables the simulation of traffic situations to predict the real-world effectiveness of advanced driver assistance systems or automated driving functions.
Safety relevant aspects of traffic, such as infrastructural elements or behavioral characteristics, are investigated with this platform. The platform is able to run entirely flexible multi-agent simulations which include stochastic variations as well as reproducibility. The flexibility is achieved by outsourcing to dynamic libraries (so called modules) all models defining the traffic scenario or behavior of traffic participants. Additionally to the platform, demonstrator modules are provided, which comprise simple traffic situations and behaviors of traffic participants as well as a collision model to answer basic research questions. Furthermore, demonstrator applications will be given to post process the simulation results. sim@openPASS users are able to incorporate their own developed modules and post processing applications; this allows them to answer a wide variety of their specific research questions. These questions are related to development, testing, and rating of vehicle functions.
Possible users can be automotive OEMs, suppliers, public transport authorities, consumer protection organizations, insurance companies, academia, and legislation.
The platform is compatible with standards such as OpenDrive, OpenScenario and FMI. This project is part of the Eclipse Working Group openPASS (https://www.eclipse.org/org/workinggroups/openpasswg_charter.php).
Eclipse sim@openPASS provides a software platform that enables the simulation of traffic situations to predict the real-world effectiveness of advanced driver assistance systems or automated driving functions.
The platform sim@openPASS will mainly consist of a GUI and a simulation core interacting with openPASS modules as well as external programs for post-processing (Figure 1).
Figure 1: sim@openPASS top-level architecture
In the sim@openPASS GUI, the users will conveniently select traffic scenarios, traffic participants (including vehicles, drivers and systems to be tested) as well as post-processing of the simulation results. The GUI will access module information and enables the parametrization and linking of the comprised models. Furthermore, scenarios available in the OpenDrive format or OpenScenario format can be loaded. The entire configuration of the simulation will be saved in an xml-format which allows re-simulation or batch processing.
The simulation core will import the configuration, load the configured modules and run the specified number of multi-agent simulations. This mainly consists of managing the virtual world, setting up the traffic participants by instantiating their modules as well as scheduling and calling the module interface functions. Additionally, the distribution of simulations over network will be available; this is important if a high number of stochastically varied situations should be simulated, e.g. in a Monte-Carlo simulation.
The openPASS modules will contain the entire functionality of the traffic participants. SpawnPoint modules are responsible for the creation of new traffic participants. Component modules comprise models assigned to the traffic participants, hence including the human behavior and/or system behavior. A general wrapper module to incorporate models using the FMI standard will be provided as one of the demonstrator modules. Observation modules observe the traffic scenario and log general or specific simulation results.
The post-processing of the simulation results will be carried out by separate applications including evaluation or visualization.
sim@openPASS cannot be used standalone. To run the provided program, additional Qt libraries licensed under the LGPL will be required.
With the migration to open source development, the sim@openPASS platform will become available for many users including automotive OEMs, suppliers, public transport authorities, consumer protection organizations, insurance companies, academia, and legislation. This is important for a standardized and thus comparable assessment of real-world effectiveness of ADAS/AD functions.
sim@openPASS aims to provide an initial code base in Q4 / 2016 and to have the first official release sim@openPASS v1.0 ready by the end of Q2 / 2017.
- BMW AG
- Daimler AG
- Volkswagen Group of America
- ITK Engineering AG
- Log in to post comments