...
- Agenda discussion - anything that needs to be added?
- Prioritized Topics:
- Decision on old open PRs
- Open Pull Requests VSS
- Open Pull Requests VSS-Tools
- Issues VSS
- Issues VSS-Tools
- Discussions VSS
- Discussions VSS-tools
- Prioritized topics for next meeting
Meeting notes 2024-04-18 (April 2024 AMM Working Session on Raw vs. Derived signals):
- The definition of derived signals is not as straightforward as the definition of raw signals
- Distraction levels of 0% and 100% are clear but what does 66% or 34.5% distracted really mean?
- Should distraction really be represented as a percentage? Or, perhaps an enum?
- Units are one way in which clarity can be achieved
- Requires introspection (ability to ask the model) support in the API or equivalent
- Even if enums are used to "bucketize" or categorize the values, interpretation of what each bucket means is still important or not necessarily clear
- For example, what is a crash and severity of crash can have different interpretations depending on culture, context, etc.
- Same for aggressiveness
- How could we make available information on how a signal is derived?
- Where it may be needed or useful
- How can this be expressed to the user?
- Does it need to be digital?
- Margins of error and calculations may vary in time
- Remaining distance based on fuel consumption varies over time
- Some signals from which a derived signal is generate may become unavailable or become available again
- The rate of incoming signals may affect the derivation
- How do we plan to repesent this?
- Even an OEM may not know how some signals are derived
- The Tier-1 may have their own way of doing the derivation and this is not available to the OEM
- What is a raw signal?
- Is there such a beast?
- Clearly there are some, but perhaps the great majority of signals are derived?
- Would adding all the information discussed above impact simplicity?
- One main goal of VSS is simplicity
- Perhaps less precise definitions of signal values is actually simpler
- Should we define the set of parameters to evaluate in order to decide if a derived signal should be added to the signal catalog?
- Is portability of applications (where they can rely on a "standard" definition of a signal) a concern?
- And therefore should the signal catalog only contain clearly defined signals as opposed to leaving definition to interpretation?
- Options to consider:
- Determine if the signal should be made part of the catalog or left for manufacturers to add using overlays
- Require that signal definitions refer to a standard
- Allow liberal interpretation, but mention in signal definition comments that interpretation is left to the implementation
Meeting notes 2024-04-09:
- Crash detection/events - car immobilized
- Stefan: Do we have anything?
- Adnan: PRs welcome
- Erik: Sometimes we create a temporary *.md file and add it to PR for discussion.. PPB to put down initial thoughts in *.md
- AMM
- No regular VSS meeting next week
- Erik has asked Paul to delete
- VSS Deep Dive being prepared by Sebastian Schildt
- Head& Eye
- VSS-tools
...
{"serverDuration": 275, "requestCorrelationId": "dafe28990dfd8c80"}