I'd be interested in knowing how well it does on the other part of the Piʻilani Highway. Ulupalakua to Kalepa Bridge (HI-31) via Kaupo.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD discussion
- Thread starter Buckminster
- Start date
Quite a few people ITT are seemingly not understanding the implementation by Mercedes here. Imagine all the functionality of Autopilot but it can flip into a Level 3 Traffic Jam Assist in the specified conditions: traffic slows below a certain threshold and the roads are mapped in the system.
You activate this system on the highway and it typically functions like Autopilot. But then you hit a traffic jam on the highway and bam it flips into Level 3 mode, you can pull out your phone, tablet/laptop, documents etc and actually stop paying attention to what the vehicle is doing while in traffic.
If the traffic jam starts clearing, the system warns you to start paying attention again before it accelerates above the threshold. You put down whatever you were doing and go back to focusing on the road while accelerating up to highway speeds.
To say this has no utility? I don’t agree, and I think Tesla would be wise to start working on stuff like this with a focus on an excellent human-machine interface to seamlessly blend everything together. But I think generalized Level 4-5 robotaxis are 10+ years away.
You activate this system on the highway and it typically functions like Autopilot. But then you hit a traffic jam on the highway and bam it flips into Level 3 mode, you can pull out your phone, tablet/laptop, documents etc and actually stop paying attention to what the vehicle is doing while in traffic.
If the traffic jam starts clearing, the system warns you to start paying attention again before it accelerates above the threshold. You put down whatever you were doing and go back to focusing on the road while accelerating up to highway speeds.
To say this has no utility? I don’t agree, and I think Tesla would be wise to start working on stuff like this with a focus on an excellent human-machine interface to seamlessly blend everything together. But I think generalized Level 4-5 robotaxis are 10+ years away.
Daniel in SD
(supervised)
I guess some people just don’t drive in traffic on the interstate. Or maybe they’re already watching movies while using Autopilot already.
AP/EAP is intended for use on highways. I've never been on Highway 37 south of Pukalani. I've used it to Pukalani from the north, where it's an excellent divided highway and it works like a dream. I have no idea what it's like south of there. My experience of EAP is that if the road is not too curvy and the lane lines are clear, it can handle it well. I would never dream of engaging AP on South Kihei Road. And I have used it on curvy mountain roads in the past without incident, but I stopped doing that because it sometimes comes closer to the lane edge than I like. Anyway, it's not intended for that kind of road. When used as intended it's wonderful. And it's intended for use on many more roads than Mercedes, which won't work anywhere on Maui.
LOL, probably similar to how I haven't been to Pier 39 in ages. But I do go on the Hana Highway/Piʻilani Highway route once every 4-5 years. Generally I'll go clockwise because you're not on the cliff side. This was the last time I did it (Piʻilani Highway section/past Kīpahulu starts at 10:33):
I've got GPS coordinates at the bottom of the video so should be pretty easy to location everything on Google Maps.
Didn't see this on these forums:
www.consumerreports.org
Maybe it's expectations vs. reality (when something is called Full Self Driving, you'd expect it to be self driving, but that wasn't evaluated). Queuing list of posts which state CR is anti-Tesla soon.
Ford’s BlueCruise Remains CR’s Top-Rated Active Driving Assistance System - Consumer Reports
Consumer Reports reviews and ranks active driving assistance systems, including Ford BlueCruise, General Motors Super Cruise, and Tesla Autopilot.
Maybe it's expectations vs. reality (when something is called Full Self Driving, you'd expect it to be self driving, but that wasn't evaluated). Queuing list of posts which state CR is anti-Tesla soon.
willow_hiller
Well-Known Member
Didn't see this on these forums:
Ford’s BlueCruise Remains CR’s Top-Rated Active Driving Assistance System - Consumer Reports
Consumer Reports reviews and ranks active driving assistance systems, including Ford BlueCruise, General Motors Super Cruise, and Tesla Autopilot.
Maybe it's expectations vs. reality (when something is called Full Self Driving, you'd expect it to be self driving, but that wasn't evaluated). Queuing list of posts which state CR is anti-Tesla soon.
No need to accuse them of being anti-Tesla. Just need to accuse them of a poorly conducted test. And all the evidence is in the article you linked.
1. They only evaluated lane keep and cruise control, and didn't account for any EAP or FSD Beta features: "Additional features such as automatic lane changes or reacting for traffic lights were not evaluated in this test."
2. Only 1/5 of their evaluation criteria was actually how the system performed. The rest was what they deemed to be important for driver monitoring safety: "CR testers evaluated the way each of the 12 systems performed within five specific categories: capability and performance, keeping the driver engaged, ease of use, clear when safe to use, and unresponsive driver."
3. Tesla tied for first place under capability and performance, and was only scored lower for those other subjective safety metrics. They're important, but certainly not worth 4/5th of the entire score of the system.
Driver engagement monitoring and the human-machine interface will be massively important for a long time and are what will prevent drivers from becoming overly reliant on ADAS that is far from infallible, which will help prevent accidents resulting from complacency or intentional misuse. I think many road users, and especially the regulators, will put a lot of importance on ensuring people using these systems actually have their focus on the road ahead of them and the vehicle itself is working to ensure that.
I would feel better if all vehicles came with similar monitoring systems whether or not cruise control/autopilot is being used, there are way too many distracted drivers on the road. If people want to put themselves in danger that's there prerogative, but engaging in that type of behavior puts other road-users at risk and ADAS with insufficient driver monitoring is contributing to that risk.
willow_hiller
Well-Known Member
I don't think anyone is disputing that driver monitoring is important. But to say Tesla falls from 1st place to 7th based solely on varying dimensions of driver monitoring is asinine.
Just evaluating base Autopilot's capabilities, it beats every other system. If you add FSD Beta into the mix, no other commercially available system even compares.
Ha! I started to watch that and just watching the video started to make me car sick, so I stopped. I once drove about 5 miles along the Highway to Hana (to a hiking trail) and just in those 5 miles I got horribly car sick, even though I was driving.
Didn't see this on these forums:
Ford’s BlueCruise Remains CR’s Top-Rated Active Driving Assistance System - Consumer Reports
Consumer Reports reviews and ranks active driving assistance systems, including Ford BlueCruise, General Motors Super Cruise, and Tesla Autopilot.
Maybe it's expectations vs. reality (when something is called Full Self Driving, you'd expect it to be self driving, but that wasn't evaluated). Queuing list of posts which state CR is anti-Tesla soon.
FSDb is not widely available, which may explain why CR didn't evaluate it. As to not evaluating regular FSD, I'm going to speculate that it's because FSD in the city requires so much driver intervention that CR regarded it as too unsafe to test. Just a guess.
I read the CR report. I can understand their concern for safety measures such as direct driver monitoring, but it fails to provide sufficient information for drivers who do not need to be bullied into paying attention. It's probably the best way to evaluate for the majority of drivers, who are idiots. People will do everything other than drive when they're driving, and a system that does not force them to pay attention to the road is unsafe for many people.
I also feel that they give insufficient weight to where a system can be used: A driver-assist system that can only be used on limited-access freeways is far less useful than one that can be used on any divided highway, and that's less useful than one that can be used on non-divided highways, and that's less useful than one that can be used wherever lane markings are present. That's where Tesla would score very high on "capabilities."
I won't be trading in my Tesla for another make any time soon. But I agree with some of their criticisms of autopilot. I regard it as safe for me because I do pay attention when I'm driving. (I have ridden with drivers who keep taking their eyes off the road to look at me. This in cars without any driver-assist systems. That scares the *sugar* out of me. I never look at my passenger while driving.)
Newbie question. Does it matter if anyone agrees? I.e., Does any of this filter up or is this just mutual massage?
From what I'm gleaning in this assessment, it's really only looking at highway systems. Right now the highway side for Tesla is Autopilot/EAP, FSD Beta is urban and really mostly Autosteer on City Streets at least until the stacks are merged and that might also improve highway performance. There are other assessments that compare Autosteer on City Streets to Waymo, Cruise, and other systems aiming to be urban robotaxis.
Didn't see this on these forums:
Ford’s BlueCruise Remains CR’s Top-Rated Active Driving Assistance System - Consumer Reports
Consumer Reports reviews and ranks active driving assistance systems, including Ford BlueCruise, General Motors Super Cruise, and Tesla Autopilot.
Maybe it's expectations vs. reality (when something is called Full Self Driving, you'd expect it to be self driving, but that wasn't evaluated). Queuing list of posts which state CR is anti-Tesla soon.
Agree with the test being pretty superficial. The most interesting category might be Capabilities and Performance where Mercedes took 10/10 but the test is of questionable value.
There's a pretty subtle distinction here in where a system can be used, this is out of the current Model Y owner's manual:
Particularly amusing with Autosteer, go to the Navigate on Autopilot page and you see this
So you leave a controlled-access highway on NOA, and the system reverts back to Autosteer that is intended for use on the type of highway you just left?
But the distinction here is mostly that Tesla doesn't disallow use of the system anywhere, the default is to allow use and then disclaimers are buried in the owner's manual that actually say it's not really intended for use in these spots or just straight up shouldn't be used. If it's not intended or shouldn't be used on these roads, why is it allowed? Much higher risk tolerance on Tesla's end I think, the other car companies could allow this but they are more conservative in their implementation.
This was a big criticism that was levied at Ford's BlueCruise when Sandy Munro put it on their YouTube channel. Oh no it doesn't handle sharp curves, that's terrible! And then you dig into Tesla's Autopilot manual and it clearly says do not use on winding roads with sharp curves, it's just that Autopilot doesn't lock it out and allows you to take the risk while covering their butts in the manual. Whether or not that's a sound approach, I think you'd get arguments for and against it.
And I'd argue that if you're willing to take on more risk by not locking this stuff out for consumers, driver engagement monitoring and the human-machine interface is all the more important.
Last edited:
I totally agree that autosteer is no good when there are sharp turns or the road is slippery. And I always disengage when there are pedestrians or cyclists. But a lot of people are more cavalier, which is why I always qualify that AP makes the car safer when used as intended.
The purpose of adas is to be used as it is much safer than when it's not used. this has been proven over and over again. A bunch of criteria and overbearing driver monitoring that restrict the ease of use or location of use prevents people from using this safety feature which results in the typical amount of crashes the US sees today. Your position is always from the "human drivers are perfect, only AP is not perfect so we should monitor people to death until it's perfect" perspective like all critics of these systems. The truth is people are terrible at lane keeping, falls asleep at the wheel, texting, arguing, and doing all sorts of shennigans that will cause crashes. I trust AP more with someone literally sleeping inside the car vs someone awake and reading instagram while driving the car in manual mode.There's a pretty subtle distinction here in where a system can be used, this is out of the current Model Y owner's manual:
View attachment 901013
View attachment 901008
Particularly amusing with Autosteer, go to the Navigate on Autopilot page and you see this
View attachment 901016
So you leave a controlled-access highway on NOA, and the system reverts back to Autosteer that is intended for use on the type of highway you just left?
But the distinction here is mostly that Tesla doesn't disallow use of the system anywhere, the default is to allow use and then disclaimers are buried in the owner's manual that actually say it's not really intended for use in these spots or just straight up shouldn't be used. If it's not intended or shouldn't be used on these roads, why is it allowed? Much higher risk tolerance on Tesla's end I think, the other car companies could allow this but they are more conservative in their implementation.
This was a big criticism that was levied at Ford's BlueCruise when Sandy Munro put it on their YouTube channel. Oh no it doesn't handle sharp curves, that's terrible! And then you dig into Tesla's Autopilot manual and it clearly says do not use on winding roads with sharp curves, it's just that Autopilot doesn't lock it out and allows you to take the risk while covering their butts in the manual. Whether or not that's a sound approach, I think you'd get arguments for and against it.
And I'd argue that if you're willing to take on more risk by not locking this stuff out for consumers, driver engagement monitoring and the human-machine interface is all the more important.
AP + human drivers doing illegal things while driving(drunk, texting, sleeping) = safer
No AP + human drivers doing illegal things = hazard/instant death
AP causes humans to do more illegal things = lol that's hilarious ->this is the position of most critics...as if people are not doing crazy illegal things already in abundance. Even if AP causes 2x more people doing illegal things, it'll still be safer than no AP once accounted for the math.
You know what happens when a tired person who wants to use bluecruise but can't because it sees that the person is falling asleep? The person falls asleep forced into driving the car in manual mode. Yeah that's safer for everyone right?
Last edited:
EVNow
Well-Known Member
Road slippery = don't engage AP/FSD.
Other than that FSD has come a long way. It is quite good even at sharp turns. I've a few winding roads with nearly 90 degree turns and FSDb used to struggle early last year. Now it can handle those smoothly.
I think all vehicles should and eventually will come with robust driver monitoring that is active at all times regardless of whether ADAS is being used, but it’s a balancing act in the real world. Someone who’s overly fatigued shouldn’t be on the road period, and some new vehicles rolling off the assembly line are already equipped with stuff like this.
In general though this is not binary, it’s not allowing a free-for-all or allowing nothing. There are ways to build in systems that properly meld the human-machine interface to reduce risks to the highest degree in all capacities while eliminating weaknesses, and that should be the goal. Perfection isn’t a realistic goal right now, humans aren’t perfect and these systems aren’t perfect but maybe they can offset each others weaknesses to create something that is closer to perfection together than either could be individually.
If such a system detects that someone is overly fatigued or has lost consciousness, which could happen for a variety of reasons ranging from fatigue to a medical emergency, it should pull over safely and put in the flashers. In the UN ADAS regulations, that’s called a minimum-risk manoeuvre. Plowing ahead with zero regard for the driver’s state is a medium-risk manoeuvre at best and likely a maximum-risk manoeuvre in many people’s eyes.
There are people who spend their entire educational and professional careers researching and working on this stuff, I’m not one of them. But I’m sure those people would have responses for all your thoughts here, and I’m sure they’d be full of nuance and complexity.
Last edited:
Someone who can't help but to text, watch movies, drunk, fatigued, and in emotional distress shouldn't be on the road...Period. But apparently no one gives a crap about your period.
So far either you make the system reliable enough so it wouldn't crash or allow the system to self drive and pull over which is not exactly easy given people are doing 90mph on the hwy so these maneuvers are questionable at best. Blue Cruise will shut itself off and put you in manual drive as you have fell asleep or drunk...that's the wrong answer. But hey the monitoring works!...in a deadly fashion.
When people are model drivers fully paying attention, crash and injury rates drop like a rock. FSD Beta has proven this pretty well. We are pretty certain that FSD beta is not more reliable than a human driver at driving. However due to people being more alert using FSD beta, the injury rate every 100M miles is near zero while it should be over 120 per statistics. People seem to forget as they fight Tesla's adas system to be some kind of death trap that the purpose of it is so to prevent injuries/death from people NOT doing the right thing while driving. So instead of preventing death due to people not doing the right thing, people wish it would lock them out from using it. Slow clap for failing to see the purpose.
So far either you make the system reliable enough so it wouldn't crash or allow the system to self drive and pull over which is not exactly easy given people are doing 90mph on the hwy so these maneuvers are questionable at best. Blue Cruise will shut itself off and put you in manual drive as you have fell asleep or drunk...that's the wrong answer. But hey the monitoring works!...in a deadly fashion.
When people are model drivers fully paying attention, crash and injury rates drop like a rock. FSD Beta has proven this pretty well. We are pretty certain that FSD beta is not more reliable than a human driver at driving. However due to people being more alert using FSD beta, the injury rate every 100M miles is near zero while it should be over 120 per statistics. People seem to forget as they fight Tesla's adas system to be some kind of death trap that the purpose of it is so to prevent injuries/death from people NOT doing the right thing while driving. So instead of preventing death due to people not doing the right thing, people wish it would lock them out from using it. Slow clap for failing to see the purpose.
Last edited:
Have I even mentioned BlueCruise here aside from the Munro criticism? I don’t recall but might have forgotten. Perceived shortcomings in one system don’t excuse shortcomings in another, and I won’t be goaded into an emotional response.
But I drove a rental 2021 F150 Lariat for work two years ago that had simple lane keeping, and I was doing long highway hauls. Even with a basic implementation like that, it would detect you weaving between lane lines. First detection it would shake the wheel. Second detection it would shake the wheel and start throwing warnings that you should pull over. In a more advanced implementation that is likely headed our way, such a system would pull the vehicle over. The way you’re portraying the system, I doubt you‘re doing it justice.
I’ve joked on here before that Tesla could make FSDBeta the default drive mode and the human would be there to monitor. Accident rates would plummet, psychiatric bills and anxiety med usage would skyrocket. Reality is full of nuance and trade-offs, and Autopilot continues operating today despite the criticism levied at it. I’ve never argued that it should be 1 or 0.
Similar threads
