Model-Classification

ADAS sensor modelling settling between the environment simulation and perceptual psychology. FrontMod classifies the sensor models based on the criterias Interface, Domain, Process und Application.

Sensor Model classification

Application describes the working area of the sensor model. Sensor models can be used for SIL and HIL applications. The requirements for the sensor models used are in the other model categories pictured.

Interface describes what kind of data is fed into the model and get out. "Interface" connections can be between the environment and sensor simulation, as well as in individual submodels of sensor simulation. The data processing speed of the sensor model plays a crucial role and is reflected in the category Process.

Process describes the way sensor models work. If simulated or measured data is replayed will called it Re-simulation. During Re-simulation no new data is generated. Through Re-simulation the data-flow could be observed to find out where in the model which decisions have been made. If input data is injected into the model will call it Data-processing. Through Data-processing new output data is generated which could be compared with other output data to find out if the model quality was increased regarding accuracy, etc. Data-processing can be used for both SIL and HIL applications.

Domain describes the ability how fast models are processed. Domain has the following property values:

Dynamic simulation can be scaled in time. Either these are pending from the simulation framework or on the dependence of the hardware used (GPU). Dynamic simulations range from Real-Time (RT) applications where a fixed time-step is used, up to flexible calculation time-steps that depend on the size of the calculated scene, e.g. Video Games.

Quasi-static calculation-speed describes simulations where model states are waiting for the result of other submodels. As a result model states detached from time limits are calculated. Classic examples are finite elements simulation tools that can calculate model states very accurately but needed several minutes or even hours computation time.