OpenSimulationInterface / open-simulation-interface

A generic interface for the environmental perception of automated driving functions in virtual scenarios.
Other
267 stars 125 forks source link

add RCS (Radar Cross Section) to VehicleAttributes #251

Open hschoen opened 6 years ago

hschoen commented 6 years ago

I suggest to add RCS (Radar Cross Section) to the message VehicleAttributes in osi_object.proto. The RCS is angle dependent and typically given in 1 degree steps (monostatic RCS). So the RCS of the MovingObject can either be given for all (360) degree steps or just for the respective viewing angles of all radar sensors in operation. I would prefer the second option (just for the sensor viewing angles of the object) as this reduces the data load. The (repeated) values must then be sorted according to the radar sensor IDs (lower IDs first). The viewing angle of the object would be defined as the angle between the viewing angle of the respective radar sensor (sensor orientation) and the center of the object's bounding box. The RCS could instead be assigned to the object by the sensor model, but I believe the sensor model is not the right place to host the RCS database for all kind of objects. Moreover, alternative radar sensor models should use the same RCS data.

Remark 1: The RCS is needed for the phenomenological radar sensor model. A physical sensor model could derive the object's RCS from the given geometry by raytracing.

Remark 2: The RCS of an object is also a radar specific sensor OUTPUT. However, this output value will not necessarily be identical to the input value discussed here due to environment effects (attenuation, occlusion, etc.).

Remark 2: Providing RCS would require to have a database for monostatic RCS of all objects that are detected by the radar sensor. It would be a good start to have at a typical RCS map for every member of the classes for moving and stationary objects (as it is already done by IPG CarMaker for certain objects).

ghost commented 5 years ago

Hi @hschoen, could you describe the derivation of the RCS more concisely? As far as I understand, you propose the transmission of one cross section per sensor, orthogonal to the line of sight of the sensor and the (target) vehicle reference point?

If the sensor model is not the right place to hold the RCS data, why would you transmit it via OSI SD instead of calculating it in user specific code?

jdsika commented 4 years ago

@PhRosenberger this might be solved by completing #367 - I will leave it open for now

PhRosenberger commented 4 years ago

@PhRosenberger this might be solved by completing #367 - I will leave it open for now

Yes, we will solve it there and then close it.