Sensor Models for Simulation

BASELABS Models for virtual test of data fusion systems

Virtual test and validation of ADAS systems is an established procedure to reduce the effort for test drives with real vehicles. Current simulation environments provide virtual sensors for the evaluation of data fusion systems. Typically, ideal sensors with error-free detection characteristics are provided. This is a drawback to the realism of the evaluation, as real sensor data is abundant with errors like measurement noise and missing detections. BASELABS Models inject typical errors of ADAS sensors into the simulation, thus increasing the realism of the virtual test.


Sensor models for virtual ADAS sensors
BASELABS Models are sensor model plugins for the use in virtual environments. The models increase the realism of the simulated sensors for ADAS and automated driving systems evaluation. They can be used in established simulation environments like PreScan or CarMaker, but also in proprietary environments. It is also possible to adapt the models to customer specific requirements. BASELABS Models are generic sensor models which mimic important characteristics of typical ADAS sensors. The models are not physical sensor models, but take realism in sensor simulation a significant step further. The following image gives an impression of the state-of-the-art in virtual sensor models and their availability.

Integration in validation environments
BASELABS Models can be used in in-the-loop setups. This includes Model-in-the-Loop (MiL), Software-in-the-Loop (SiL), Hardware-in-the-Loop (HiL) and Vehicle-in-the-Loop (ViL). The integration into evaluation tool chains based on Matlab Simulink is also possible. The implementation for PreScan supports the integration into Simulink out of the box.

Reproducible results
Simulation runs with BASELABS Models can be done either in a non-reproducible way to cover the broadest variety of possible sensor errors. Or the simulation can be repeated with exactly the same modelling of sensor input with the same distribution of injected errors to evaluate a critical constellation in a reproducible way.


Loading the player ...

Usage of BASELABS Models in PreScan with simulation of true and false detections.


Principle and usage of BASELABS Models
BASELABS' expertise originates from the development of data fusion systems which can handle the error-prone detection characteristics of real ADAS sensors. For BASELABS Models, we invert this knowhow to inject errors into the simulation that are comparable to the characteristics of real sensors. The user of BASELABS Models can modify this characteristics by setting parameters for the injected errors to adjust the behavior of the plugin to his or her specific needs.


Infusion of typical errors of radar sensors

Measurement noise

False Negatives

False Positives


Smart Camera

Infusion of typical errors of camera sensors

A camera sensor naturally may determine only the position of objects in the image in the camera directly. Based on the observation of those object positions over time dynamic properties of the objects are determined. However, these methods are subject to typical errors, which are modelled by the Smart Camera sensor model:

  • The filter algorithms for the estimation of the object dynamics based on a sequence of object positions, which are applied inside Smart Cameras, are subject to statistical errors as they are based on model assumptions.
  • The motion of the vehicle, such as pitching on braking, affects the position of objects in the image of the Smart Camera and therefore also the output of the filter algorithms.


Validation of the simulation results
The plausibility of the BASELABS Models approach of fault injection to achieve a more realistic simulation approach has been empirically proven by comparing the result of the fault injection with the theoretically expected distribution of the corresponding faults in real sensor data.

Validation results

Measurement noise

False Negatives

False Positives

Contact & Downloads

Top of page
Evaluate now!

evaluate now!

Please contact us for an evaluation (+49 (0)371 3371 51 0) or leave your contact data below.

* Required fields