Week 3 meeting - Jelc-sys/AV-control-handover-in-critical-situation GitHub Wiki

Meeting 19-9-2024 notes

Present: Luca and Jelco

  • We could make our experience like an experiment (log a keypress for example)

    • Does it feel safe to cross?
    • Also record a baseline, so you can indicate whether something is statistically significant
    • Idea: Press a key when you feel nervous (as the emergency vehicle is coming by)
  • Change the interior of the car to make it more human-centered

  • Humans have limited cognitive ability, so when distracted we must draw their attention ahead of time

  • US/EU focuses on vehicle 'heritage', AV cannot differ too much from current cars. China does take bigger leaps in changes.

  • Idea: Implement platooning: Copy many cars and have them follow each other. This is to improve traffic flow and reduce emissions.

    • Point: What if the AV is manually overridden and the driver stops? How does the rest of the platoon stop?
    • This is a problem to tackle for programming, but will also be a problem in real life when platooning becomes reality.
  • Will the AV be in an evolutionary or revolutionary scenario? What will the environment look like:

    • Evolutionary:

      • Steering wheel in place
      • A lot of driver feedback, as AV is not yet fully autonomous (big screen on dashboard, ADAS etc.)
      • There will be non-autonomous drivers on the road mixing with AVs.
      • Road signage in place to assist manual drivers, as well as traffic calming measures
    • Revolutionary:

      • No driver input (steering wheel, gas pedal etc.)
      • Less driver feedback (perhaps only show waypoints, current speed etc.)
      • Only AVs, so less complex implementation
      • Less signage and no traffic calming measures (roundabouts, speed bumps)
      • Cityscape is changed - less parking spaces, so more room for greenery etc.
  • In a revolutionary scenario, what if the car has no more windows etc. because it's too heavy? Then people would wear AR goggles to simulate being able to see outside, whilst sitting in the dark metal box that is now the car.

  • Can we implement the idea of SLAM being present in the simulation? How to let the experiencer know?

    • Show map on the dashboard with outlines of environment (like Tesla does now)
    • Add AV infrastructure to the map (communication poles, LIDAR and radar on car)
  • Perhaps use XBOX controller to also implement haptic feedback

  • Situation for handover:

    • There has been an accident on the road, there is also an emergency vehicle coming -> Car hands over control
    • Could already give heads up far ahead that there is an accident -> Put away whatever you're doing and pay attention
    • When the car gets to the accident, an emergency vehicle approaches and the car opts to give you back control
      • If you don't respond within 10 seconds, the car will pull over

AP's

  • Getting an AV to stop when another vehicle is ahead - Luca
  • Pedestrian and driver press button when you feel trust - Luca
  • Add a 'Tesla-style' screen for evolutionary interior - Luca
  • Haptic feedback implementation for the driver - Jelco
  • Handover process - Jelco
  • Emergency override - Jelco
  • Car interior change - Marthe
  • Implement visuals on the dash for warnings or to indicate handover (such as timer) - Luca
  • Environment - accident up ahead and the emergency vehicle approaches - Marthe