Automated driving features present their own unique development challenges. For example, collaborative vehicle communication systems can offer significant safety advantages through direct communication with nearby vehicles, providing location information on movements that are outside of visual range due to obstacles. However, according to testing specialists FEV, the highly complex and unpredictable environment in which these systems operate presents complex development and validation requirements.
“Collaborative embedded systems (CES) in vehicles create new challenges such as increased security risks and heightened performance requirements to enable safe and effective operation in dynamic and unpredictable environments,” said Elmar Börner, senior group director for ADAS (advanced driver assistant systems) and AD (automated driving systems) development at FEV. “The unpredictable operating conditions for such systems present a nearly limitless combination of potential use cases, which makes comprehensive physical testing impractical.”
To overcome these hurdles, FEV says that a virtual test environment, empowered by co-simulation is required. It explains that the term ‘co-simulation’ refers to a virtual development approach that integrates multiple simulation tools, enabled by a co-simulation master such as FEV’s real-time software for virtual experimentation: xMOD.
Traditionally, model inputs would be compiled and directly imported from slave applications, for example via a Functional Mock-up Interface (FMI). However, certain inputs, such as 3D environmental models developed in tools like CARLA, are bound to their executing platform and thus cannot be imported into a co-simulation master.
To address this challenge, FEV explains it has utilized a new Distributed Co-Simulation Protocol (DCP) for time-synchronized execution of models on distributed platforms. It claims that DCP support in xMOD enables easy integration of new simulation models, and even hardware, into co-simulations that would be incompatible with previous integration methods. Using DCP allows for both definition of operating models in hard real-time, soft real-time, or non-real-time as well as their combining with existing model exchange solutions. The end result, states FEV, is the ability to undertake complex cyber-physical simulations, like multiple collaborating vehicles in a common 3D environment, controlled via CES.
“Virtual testing is mandatory for automated driving applications since on-road validation is not feasible,” noted Börner. “Complex cyber-physical system simulations and co-simulators play a key role for virtual testing, and FEV is proud to be the first user of the DCP co-simulation standard within an automated driving context globally. Our approach enables the user to integrate and handle different simulation platforms available on the market and to deal with the highly dynamic situations occurring in real traffic environments.”