Debugging Critical Situations by Replicating Generated Scenarios in the Lab.

• Replicating critical driving situations in the lab
• Minimizing the time “from bug to debugging”
• Scenario database for future use cases

Task: Debugging by Replicating Driving Situations in the Lab

Issues that occurred during real test drives with prototype vehicles are analyzed in the lab by means of simulation

Challenge: Exceeding the Limits of Software

In certain situations, the capabilities of the software are limited. This results in disengagement, which means that the driving function gives control back to the safety driver. We no longer want to only exceed these limits; we want to push them.

Solution: Reproducible Tests with Generated Scenarios

Recorded measurement data (raw data or object lists) makes it possible to replicate the driving situation in the simulation. The reproduction is precise. If required, it even includes 3-D environment models for the physical sensor simulation. Thanks to reproducible tests in the simulation, software additions can be created in the lab and immediately be tested in a reproducible manner. This lets you prove error-free operation.

The situations in which the driving function disengages are usually corner cases, which will also be relevant for the validation of future driving functions. These valuable simulation scenarios can be stored in a scenario database.

Benefits of Scenario Generation by dSPACE and understand.ai

Once the process for the automated import of measurement data into the dSPACE tool chain is set up, the simulation scenario can be created in a matter of hours. This reduces the time 'from bug to debugging' to a minimum.

To make a test meaningful, a scenario must be semantically consistent. This is achieved by extracting scenarios from raw data, which ensures that the scenario does not contain false positives or false negatives.