Physical Models
High-Fidelity Models
Radar and lidar models using ray tracing
Utilizing the Powerful SMDL
Sensor models with the highest requirements need to be physically accurate. Don't test AI with AI. Use physical simulation instead. We develop high-fidelity radar and lidar models using our powerful Sensor Model Development Library. Inside the sensor models, a full 3D environment is build up based on standardized interfaces. A custom ray tracing with wavelength dependent material interaction is performed. The ray tracing results are used to generate the realistic, high-fidelity sensor output.
Fully Standardized Interfaces
The input of the sensor models is fully standardized. The main input is an ASAM OSI SensorView message. This message contains the ground truth information of all moving and stationary objects. Furthermore, it contains references to ASAM OpenMATERIAL 3D compliant 3D assets. They contain the geometry and material information needed for the ray tracing.
Parameterizable to any Real-World Sensor
The radar and lidar models can be parameterized to any real-world sensor. The parameters are either taken from the datasheets of the sensors or derived from measurement data. This includes a variety of different parameters and effects including timing effects. Every ray in the ray tracing is shot with an individual time stamp. The beam pattern can be changed in any simulation time step, facilitating for example complex beam patterns of MEMS lidars.
FMCW Lidar Simulation
With the SMDL ray tracing as a back bone, we are one of the first companies to offer FMCW lidar sensor models. FMCW lidars have the big advantage over conventional time-of-flight lidars to directly measure radial relative velocity. Our models are capable of simulating the velocity component on every hit point in the 3D environment, including velocities on individually moving parts like wheels and the arms and legs of a human. In the image you can see a point cloud of an FMCW lidar colored by relative radial velocity.
