-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
[uk] UltraSonic Sensors removal/TV replacement performance
- Thread starter randompixel
- Start date
MrBadger
Badger out
I work with measuring hundreds of objects per second potentially in close proximity to each other. We need to measure location to a few cm or so over distances that may be many 10’s of meters.
We trial multiple sensor types simultaneously so can compare the relative performance of them.
Guess what. Pure camera vision is the worse performing and the only reason why we might consider using them is that they are ‘cheap’ in comparison with some of the other technologies. .
Some sensor technologies give us accurate locations bus inferior visual properties than pure vision cameras that might give better visual interpretation but poor location. AI perception of objects tells us what we are looking at ragardless of technology.
We work as a partner with various sensor manufacturers. Some pure vision based, some not.
tl;dr if you want something to look pretty, you use a vision based camera, if you want to accurately know where it is, you choose something else.
There may be a reason why SpaceX use non vision technologist in some situations…
Last edited:
We are parking a car on planet earth not landing a probe in Mars!
interbear
Active Member
Interesting debate but surely the simple answer for cars is that a mix of camera based vision and sensors is the best option for accurate and safe operation. In order of all round effectiveness in all weathers, night and day, I’d imagine it’s:
1. Cameras + Sensors
2. Sensors only
3. Cameras only
I don’t see any argument other than cost saving or irrational blind (no pun intended) optimism to conclude, as Tesla have, that only using cameras is the best possible option.
Obviously it’ll be interesting to see the real-life experiences and evidence over time but I hope that my sensor equipped Model Y continues to use them in parallel to cameras, rather than getting hobbled by a future software release, which I fear will be the case as I can’t see Tesla wanting to maintain multiple streams of software.
1. Cameras + Sensors
2. Sensors only
3. Cameras only
I don’t see any argument other than cost saving or irrational blind (no pun intended) optimism to conclude, as Tesla have, that only using cameras is the best possible option.
Obviously it’ll be interesting to see the real-life experiences and evidence over time but I hope that my sensor equipped Model Y continues to use them in parallel to cameras, rather than getting hobbled by a future software release, which I fear will be the case as I can’t see Tesla wanting to maintain multiple streams of software.
I would expect that the software will test for the presence of USS hardware and use it. If the test fails it will use cameras. They could even put in an on/off toggle for the USS in the UI. It would surely only take a short piece of code.
MrBadger
Badger out
Whilst SpaceX may have aspirations of flying to Mars, I am not aware they have yet done so.
But they do use LiDAR when parking their capsules against other space objects.
The argument, that I'm not remotely qualified to offer an opinion on, is that each type of sensor is prone to a degree of false detection, and safety means you should always err on the side of caution so it therefore increases the number of false detections e.g. tells you to STOP or applies brakes. If you can achieve the required level of safety with one type of sensor then the other adds no value.
I also have USS and they have never detected a kerb, and I (once) damaged my alloys on such a kerb, as I would suggest most people here have done. While I haven't driven into a wall, I'm not very convinced that's down to the USS as it seems to spend most of it's time telling me to STOP. Would I trade losing service in some weathers for kerb detection? maybe. might be nice to have a toggle.
I drove 50 miles to work this morning on the M5 through heavy rain and spray. the rear camera is very blurred out, covered in muck and water; car says Park assist unavailable, yet when I reversed to park in my car park, the park assist stuff all worked perfectly, It even saw the curb I was reversing up to, it told me to STOP at the appropriate time. I don't have USS, just tesla vision. So far, it seems to work fine and is better than not having anything for the last few months.
That's the kind of false reading you'd expect.. it sees something lighter in front of it and assumes it's daylight. It would be fun to park in front of a mirror..
Don’t say it works the mob here will throw bricks because they want this to fail
Did you not read my original post for which he said about SpaceX? You don’t right.
Last edited:
Chimpy McChimp
Member
I don't want it to fail. I desperately want it to work and work well. I just downloaded last night and had a very quick try out this morning. So far not very impressed. Parking between two cars the display jumped from saying 24inches to stop with very little movement from the car. The 24 inches was probably accurate and the car maybe moved a further 2 inches before telling us to stop.
Also at home we pulled up to a fence and it said 24 inches and we were around 8 inches from the fence.
That said, it does spot kerbs and edges very well which is great for ensuring I don't kerb the alloys, or drive on my grass at the side of my drive (can I get it installed on my dad's Kia as he seems to drive on my grass every time he parks in our drive?
)
I'm hoping it improves over time. Better than nothing, but needs work.
Also at home we pulled up to a fence and it said 24 inches and we were around 8 inches from the fence.
That said, it does spot kerbs and edges very well which is great for ensuring I don't kerb the alloys, or drive on my grass at the side of my drive (can I get it installed on my dad's Kia as he seems to drive on my grass every time he parks in our drive?
)I'm hoping it improves over time. Better than nothing, but needs work.
The model 3 with USS I have mostly say 32 inches when I reverse park when it is roughly around 12 inches. I’ve never found that difficult to guess after the initial couple of attempts. If accurate sensors is the must in you car needs you may have to start with that in the top of the list when you are narrowing down your car options.
Last edited:
Zilla91
Active Member
Don't say it doesn't work because the true believers will throw bricks because they don't want this to failDon’t say it works the mob here will throw bricks because they want this to fail
I dont think its a case of it working or not. Under certain circumstances Vision park assist seems to work really well, and in others it does not work at all. All engineering solutions have pros and cons, including USS. But at the moment it seems for Vision Park Assist is that there are more cons at the moment.Don’t say it works the mob here will throw bricks because they want this to fail
So its not for me and I have cancelled my '23 Model Y delivery.
Similar threads
