AI generates driving scenarios from measurement data
In order to be able to reliably safeguard autonomous or semi-autonomous vehicles, many thousands of scenarios are needed that are as realistic as possible and also include rare events. The manual creation of so-called “rare events” in special editors is enormously time-consuming. With a service for scenario generation, dSpace now wants to transfer the complexity of the real world to the simulation and thus enable the safeguarding with critical simulation scenarios.
The Scenario Generation Service from understand.ai and dSpace draws on existing data sets recorded during measurement runs. In a highly automated process, understand.ai’s AI-based annotation solutions extract the relevant information from the raw data of the vehicle sensors. In this way, realistic and consistent simulation scenarios are created. Optionally, data from object lists can be used for scenario generation.
The scenarios are used to generate exact reproductions of real driving situations in the simulation in order to recreate events from test drives in the laboratory or to compare simulations of sensor models with measurement data. The generation of logical simulation scenarios makes it possible to create many new, previously unknown “corner cases” via scenario-based testing by simulation and thus to test autonomous driving functions in a variety of relevant and critical situations.
The road model required for the simulation can also be modelled by the Scenario Generation Service on the basis of the sensor data. Alternatively, the use of supplied HD maps is possible for this purpose. In addition, detailed 3D models of the vehicle environment can be generated for the physical sensor simulation.
The generated scenarios enable immediate use of the scenarios in the ASM simulation environment of dSpace and the existing infrastructure for SIL and HIL applications. In addition, the scenarios are provided in OpenScenario and OpenDrive so that transferability to other simulators is possible.
More information: https://www.dspace.com/en/pub/start.cfm?select=1
Related articles:
AVL, Rohde & Schwarz join forces for XiL tests systems
PSA taps Altran’s expertise in the validation of automated cars
Development tool enables energy-efficient electric powertrains
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
