The problem here is what you mean by "FSD". NoA/EAP are designed mainly for freeways, though EAP can be used on the streets, but only as a rather dumb driver assist. Then Tesla added "FSD preview" which was really just a hack that added traffic signal awareness to automatically stop the car (mostly) on city streets. Only FSD beta can truly navigate in an urban environment (well, in theory), and this is more correctly called "City Streets".
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD vs Stopped School Bus
- Thread starter mmyers300
- Start date
Yes while we all call it FSD Beta, Beta or FSDB the correct nomenclature following Tesla's would be "Autosteer on City Streets Beta". Of course that is a double mouthful. Overall Tesla has done a poor and confusing job on naming, identifying and standardizing the various driving modes and driving packages they offer and sale.
Or any merging traffic.
Yesterday my Model 3 would not let a Tesla Model S merge into the freeway as I was cruising in the right most lane. Had to accelerate manually to avoid a side swipe. The Model S driver was suitably pissed, and ended up passing me very fast to get in front of me. We both were taking the next off ramp.
It's the merging vehicle's reponsibility to do so safely, either by slowing down or speeding up.
Last edited:
I suppose you also think the triangular signs marked "Yield" mean something having to do with right of way too?
_Redshift_
Member
That’s true. However, when the vehicle is not doing that, it’s the AP’s responsibility to smoothly and defensively drive, to avert a collision. Like a human would. You know, with only two eyes.
AP has ten eyes. Eight ultrasonic sensors. Mine still has a radar.
One feature I would like to see added is to move over a lane when several cars are merging into the highway instead of trying to merge with them. Like a human does. I have seen it behave that way but rarely.
This falls under a broad category of quibbles I've had with AP on the highway that's like.. think a step ahead or second-order decisions making.
Another example of this I find annoying is when traffic has slowed considerably off in the distance but still very clearly visible and coming up quickly. The car will basically cruise along at speed (or even speed up if it was below set speed and a space opened up).. right until it needs to hard brake.
Similarly in some stop&go traffic situations, as car ahead creeps forward it may let a larger than normal gap grow.. and then jerk forward then hard brake. I've seen this happen far too often. Rather than sort of coasting slowly up to speed and then back down..
Feels like the software has a binary black&white brain, and only decides to GO or STOP, without a lot of nuance to how hard to GO or how hard to STOP.
There's a certain lack of smoothness. If your Uber driver did this you wouldn't rate them well.
Reminds me of old yellow cab riding in NYC when you know the guy is on his 6th 12 hour shift of the week...
_Redshift_
Member
Yeah but the neural net… it’s supposed to be learning.
Although I feel it’s been playing hooky for a long time now.
Have a BMW iX on the way.. will report back on how the highway automation compares
TresLA
Member
Not sure why you are referring to you expect FSD beta to handle this situation at this time. Tesla didn’t communicate that it can, and in fact has communicated that it can’t handle a lot (hence L2). Pretty sure no one will legally hold Tesla to list every situation FSD beta doesn’t handle yet, otherwise we’d be listening to a long pharmaceutical commercial.
So we shouldn't expect FSD to handle __ situation, because Tesla didn't tell us that it did.. but also we shouldn't expect Tesla to list all the situations it can't handle because the list would be too long?
This is quite the cop-out.
Similar threads
