Some LKA observations

  • Thread starter Thread starter hobbit
  • Start date Start date
  • Replies Replies 2
  • Views Views 909

hobbit

Well-Known Member
Having finally had a chance for a long "flatlands run", I not only got to play with cruise-control a little
more [which I won't use in the hills, it's too stupid], I also tried to see if the full-bore "Active LKA" to see
if the combination could approximate self-driving for those very specific conditions. Normally I run it
in the middle setting, which gives the gentle and quite reasonably well-damped nudge when I actually
wander to the edge of a lane; that's occasionally helpful if I'm poking a GPS or the like, but doesn't try
to actively kill me when the lanes go weird.

So it almost works, but still wants *some* input from the driver. With many seconds at a stretch to let
the algorithms play out, I determined what it's actually looking for. It doesn't want to see just any torque
sensor input against where it thinks the wheel should be, it wants to see *variable* input, any at all, within
the timeframe of about 13 seconds.

So the "hang a weight on one crossbar" hack would not work, that's a passive, unchanging torque delta
that's not "alive" enough for the LKA's druthers. Whar worked to eliminate the nagging was to simply nudge
the wheel once, about every 10 seconds. I could keep that up indefinitely, it seemed, tooling along at a
nice rate a quarter-mile or so behind a truck and having the car follow the gentle curves by itself, decently
centered in the lane. Nudge ... nudge ... nudge .... kept it happy.

But this actually grew tiresome after a while, so I took control and went back to my normal "middle mode"
for the rest of the day. I can still respond more smoothly to what's upcoming, because I look farther ahead
than the LKA camera and evaluate what traffic and terrain are doing in the distance. It was worth the brief
experiment to determine more about how the system works, nonetheless.

_H*
 
Back
Top