Instance discussion - Driver/Passenger vs Left/Right

[erikbosch]

There have been some criticism on the VSS change some time ago to use DriverSide/PassengerSide as reference rather than Left/Right. I can see values in supporting both notations, or that at least OEMs can decide which notation they prefer.

This PR discuss/propose a possible solution. In short it is based on having something like this in VSS:

Seat:
  type: branch
  instances:
    - Row[1,2]
    - ["DriverSide","Middle","PassengerSide", "Left","Right"]

We can also use this PR to capture the discussion/decision on whether we actually need to do something.

Some previous discussion available in #642

[doganulus]

I have seen the mention of reusable definitions in #642 (comment). I think this can be a good solution. Similarly, the JSON Schema has $ref keyword for reusable definitions. An example would look like this:

vehicle.seat.row1_left:
  $ref: #/definitions/seat_definition
vehicle.seat.row1_right:
  $ref: #/definitions/seat_definition
vehicle.seat.row2_left:
  $ref: #/definitions/seat_definition
vehicle.seat.row2_right:
  $ref: #/definitions/seat_definition

definitions:
  seat_definition:
    properties:
      row: 
        type: integer
      pos:
        type: string

Now, above, I have duplicated each signal name. This might be enough, but maybe we can look at key expressions, too. Yet I am not sure about that.

vehicle/seat/row[1-2]_{left,right}:
  $ref: #/definitions/seat_definition

[alanmapleburl-au]

Here is a perspective from an ontologist. TL/DR: I support - ["DriverSide","Middle","PassengerSide", "Left","Right"]. It uses anyOf (inclusive or) logic and provides the knowledge required rather than using oneOf (exclusive or) logic to constrain it.

In Basic Formal Ontology, which is a philosophically rigorous upper ontology, the lateral or y-axis position (Left, Middle, and Right) would be deemed "qualities." They are only dependent on the thing they describe, such as the seat position. They will always be able to describe the seat position regardless of processes or circumstances.

Driver and Passenger, on the other hand, would be deemed "roles." They not only depend on the thing that they describe, but also on some process--driving a vehicle in this case. The problem becomes clear when you consider a fully autonomous vehicle in which there are passenger seats but no driver position.

In a fully-articulated model, the position of the driver controls would be a property of the vehicle, not a seat. Everything necessary could be reported by vehicle.driverLateralPosition and seat.Lateralposition. Given the tree model and the notion of allowed values, if we're going to constrain with oneOf, then [Left, Middle, Right] is the stronger option. But I see no reason that we can't have it both ways by including both the lateral position and the occupant role.

[erikbosch]

My conclusion so far:

• No-one really seems to object supporting both left/right and driverside/passengerside
• The question is if we need to add syntax to VSS so that implementations (like VISS server, Kuksa) automatically knows that they are synonyms, or if we leave that to implementations

Maybe I will create a PR for a "simple" solution

[erikbosch]

Closing this one for now as there likely will be no progress within the next months