Proposal: Standardize Sensor Mounting Position as FMI Input Parameters

[ClemensLinnhoff]

Describe the feature

Currently, it is unclear, where in a co-simulation the mounting position of a sensor is defined. By talking to different people of OEMs, Tier1s and tool vendors, mainly three possibilities were mentioned:

1. Set the mounting position where the SensorView is generated and transfer it to the model with the SensorView.
2. Hardcode the mounting position in the sensor model and if necessary transfer it with a SensorViewConfig request.
3. Set the mounting position via proprietary FMI parameters in the sensor model and if necessary transfer it with a SensorViewConfig request.

All three methods are used in different companies. This limits the interchangeability of the model FMUs.

Describe the solution you would like

All three methods have their pros and cons. Here is my take:

Solution	Pro	Con
1	No SensorViewConfig request necessary to set the mounting position. Since to my knowledge no simulation tool currently supports the config request anyways, this would circumvent problems in that regard.	In a potentially automated test with multiple scenarios and variation of mounting positions, this automation needs to be done with proprietary tooling where the SensorView in generated.
2	The mounting position is clearly defined and can be validated for the model. In a production ready simulation, this makes it easier to configure the sensor setup in the simulation, since the naming of the model (e.g. corner_radar_front_left_vehicle_xy) makes it clear, which sensor it simulates.	The mounting position cannot be changed after compiling the model. Variations cannot easily be simulated, if desired. And, depending on the model, the SensorViewConfig request is necessary.
3	The mounting position can be defined e.g. with SSP. It can be set freely in the co-simulation and can also be varied from simulation run to simulation run. No tool proprietary mechanisms are needed, since setting FMI parameters is defined in SSP.	Currently FMI parameters to set the mounting position are not standardized. Every model has a different naming. This limits the interchangeability of the models. And, depending on the model, the SensorViewConfig request is necessary.

I am personally leaning towards solution number 3 for the named benefits. But I would standardize the naming of the FMI parameters here in OSMP. I would suggest to introduce 6 scalar variables, e.g.

• mounting_pos_x
• mounting_pos_y
• mounting_pos_z
• mounting_orient_yaw
• mounting_orient_pitch
• mounting_orient_roll

With "mounting position" I am referring to the physical mounting position of the sensor. If we need to set the virtual mounting position also this way is up to discussion.

[ClemensLinnhoff]

Any takes on this?

@jdsika @LukasElster @HendrikAmelunxen @raue @hschoen @pmai @ThomasNaderBMW

[raue]

I'd prefer option 2 for use cases, which deal w/ validated models (e.g. customer projects). For others option 3 is fine.

[sterres]

I prefer option 3 as it gives most freedom to user.

We do not expect to see SensorViewConfig mechanism work with the typical simulators out there.

[jdsika]

I like to consider Option 3 as it puts some positive pressure on the usage and implementation of SSP in simulators.