mrkisskiss
Member
Actually, I find this a new behaviour in 17.17.4... it seems to move over depending on what the ultrasound sees. If you're driving on a dual carriageway here in the UK, in the right hand lane (i.e there's the road barrier on your right, and you're passing cars that are on your left), the ultrasound sees this, and the car moves away from the barrier - towards the middle of the road (and the cars you're passing). Notably, the car centres itself in the lane if there's a car next to you (i.e you're sandwiched).
The bias away from the barrier is dangerous, and makes for very uncomfortable driving. The problem is that the ultrasounds are slow to pick up cars next to you - and sometimes don't at all... so you end up overtaking VERY close to the cars, and coming up behind them very close - you're almost over the lane line at this point. It does show the bias on the display (the car is rendered very close to the centre line)
I've taken over numerous times because of this - it makes autosteer somewhat useless on dual carriageways with a barrier/wall relatively close to the roadside.
Hopefully they'll be able to figure out the path delimiters using vision, rather than relying on ultrasound, and make a judgement call on whether it's safer to centre in the road, bias away from cars, or away from the barriers, depending on the surrounding road situation.
Complex stuff -- it's moments like this where I ponder if a vision-first system is actually capable of keeping us safe in the interim.
The bias away from the barrier is dangerous, and makes for very uncomfortable driving. The problem is that the ultrasounds are slow to pick up cars next to you - and sometimes don't at all... so you end up overtaking VERY close to the cars, and coming up behind them very close - you're almost over the lane line at this point. It does show the bias on the display (the car is rendered very close to the centre line)
I've taken over numerous times because of this - it makes autosteer somewhat useless on dual carriageways with a barrier/wall relatively close to the roadside.
Hopefully they'll be able to figure out the path delimiters using vision, rather than relying on ultrasound, and make a judgement call on whether it's safer to centre in the road, bias away from cars, or away from the barriers, depending on the surrounding road situation.
Complex stuff -- it's moments like this where I ponder if a vision-first system is actually capable of keeping us safe in the interim.