Learning-based motion planning for AMZ Driverless

แชร์
ฝัง
  • เผยแพร่เมื่อ 29 ต.ค. 2024

ความคิดเห็น • 7

  • @sebastiannicolasgiles3659
    @sebastiannicolasgiles3659 6 ปีที่แล้ว +17

    Hi guys, are you getting close to the car's power/traction limits with your system? if not, then what is your number one bottleneck? Is it sensing, planning or control?

    • @jurajkabzan86
      @jurajkabzan86 6 ปีที่แล้ว +3

      We are getting to the traction limit in corners. In the fast corners, we are taking advantage of the aerodynamics package. At this point, the bottleneck looks to be combined. Not a single package but rather the fusion of all of them.

    • @sebastiannicolasgiles3659
      @sebastiannicolasgiles3659 6 ปีที่แล้ว +2

      ​@@jurajkabzan86 Wow, so the actual trajectory it's taking already reaches physical limits.
      Why does it often run wide? Does it fail at holding the planned trajectory or is that indeed the trajectory it computed as optimal?
      (PS that guy on LikedIn copied my comment :) )

  • @fredericsidler
    @fredericsidler 6 ปีที่แล้ว

    Amazing. Could you use a drone view to improve it?

  • @matel9985
    @matel9985 6 ปีที่แล้ว +3

    Good job! Is it beating your human lap times?

    • @xxemerica
      @xxemerica 6 ปีที่แล้ว +1

      Nope its not. Not yet - next car is coming soon.

  • @ireminmon
    @ireminmon 6 ปีที่แล้ว

    Good work.
    The car seems to be somewhat afraid of cutting corners. I don't doubt you can get this thing to drive faster without human driver. It just needs some more confidence =).