Recompute is a task in which recorded or artificially generated data is provided to a system, such as ETAS Deterministic Middleware Solution applications.
The system processes this input, and its output is then processed or recorded for subsequent analysis and reporting. In the context of ETAS DMS and virtual testing, recompute plays a crucial role in addressing the challenges of validation in the development of automated driving systems. It helps shorten feedback cycles for developers, enables efficient continuous validation, and supports various use cases such as development, forensic analysis, and validation testing.
Recompute is particularly vital in safety-critical environments like the automotive industry, ensuring determinism, reproducibility, and real-time execution in virtual testing scenarios.
Real physical test drives in ADAS / AD development serve to observe misbehavior or bugs and record driving data. These drives offer a dynamic environment for evaluating system performance in real-world scenarios, enabling developers to identify issues not apparent in simulations.
ADAS functions and automated driving systems can behave different for identical driving scenario.
Important is that the systems always exhibit a reliable behavior in terms of functional safety and Safety of the intended function (SoTIF)!
What matters for problem analysis is that each recompute is fully identical compared to its original test drive.
This form of recompute - the so-called deterministic forensic recompute - supports code debugging with back & forth stepping!
ETAS DMS provides different types of recompute. This allows to cover a large set of development tasks - from problem analysis / debugging to function development and simulation-based validation.
The ETAS DMS SDK provides several tools to leverage the benefits of recompute.