scottf200
Well-Known Member
It doesn't make any sense to me why there is so many spline nodes there. NOTE that in the other lane that there are 52 spline nodes.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
AP2.0 Cameras: Capabilities and Limitations?
- Thread starter lunitiks
- Start date
I on the other hand know what root is and would like to root my car so I can have full access to the systems
Even as someone who works in tech, there are moments when I get to snicker about usage of the term root.
Urban Dictionary: root
PaulS
Member
Anybody know what AP does (or will do in the near future) if there is an obstacle in the road, like something that feel off a truck or a rock too be to let a tire hit?
And if she reads this, you'll be driving yourself to your new appartment.
Well, right and wrong. Remember 186,000mi/sec, it's the law! Drones don't just send the commands 3,000mi line of sight, but to space and back. Adds a little distance. I'm sure they have some sort of algorithm to reduce the latency to nearly zero, but if they can do it, Elon can do it.
Easy big guy. As a professional airline pilot, I take offence to that one.
Run into it. At least for the foreseeable future. Eventually it may avoid or come to a stop but now.
Trolling the people who already bought FSD?
Middle of next year we will be 1/5th the way there, for sure!
Middle of next year we will be 1/5th the way there, for sure!
The take off and landing are done by a local drone pilot then when they are high enough up it switches to the pilot on the other side of the planet. Since they can fly for a long time it doesn't take a lot of local pilots to handle all of the take off and landings.
Three months maybe, but definitely six months for certain!
This is a minor thing, but you are confusing two concepts - you are talking about spline nodes/points, but highlighting on the screenshots the number of elements in Local Speeds list. Those two does not have to be related. There are 18 and 46 nodes on those splines, which is still a lot.
I don't know why exactly there are so many nodes and speed points, but I may have one hint. In both those cases the spline needs to be divided into segments of equal sizes. For spline nodes I think it is needed to simplify the calculation of the position along the spline. For speeds list this is just how this list works. So if for any reason the entity creating this map decides, that it needs a short segment to properly describe the geometry of the road or the distribution of speed, than it needs to divide the whole spline into equally short segments.
pilotSteve
Active Member
PaulS
Member
Ouch
Do you suspect that the HW2 platform can capture enough info for future s/w algorithm enhancements to achieve avoidance? I really wonder if HW2 will achieve full autonomous travel... many times we engineers don't know what we don't know, until we get out there and try... I suspect HW2 is a huge step forward, yet possibly Try2 of a couple more HW iterations to come over the next several years as new challenges are uncovered by real-world testing. Somebody has to do it and have loyal customers willing to assist along the way... keep up the great work Tesla![ btw, sorry for the typos in the original question... child distracted at home... kind of like child distracted while driving at times... all the more reason for AP assistance ]
Purely speculation at this point, since FSD is speculation at this point, but I wouldn't see why the camera's and/or radar couldn't "see" enough to mitigate or avoid this type of collision. The NN "just" has to be trained to identify road debris, but probably more challenging is the driving policy has to be taught when to avoid it (no obstacles in adjacent lanes) or just run over it, if that is the lesser of evils so to speak. As we all well know, sometimes you just don't have a choice but to run over debris if there aren't any other options.Ouch
[ btw, sorry for the typos in the original question... child distracted at home... kind of like child distracted while driving at times... all the more reason for AP assistance ]
They are still struggling to get AP2 just to see road markings, let alone obstacles or understanding the road layout.
AP2 was a step backwards, and I doubt it will ever do FSD.
AP2 was a step backwards, and I doubt it will ever do FSD.
btw I was exploring this whole cars display in the past couple of day and it seems that AP itself is pretty aware of the cars in adjacent lanes, just not communicating this info to IC. But it all changes when you try to change a lane, then it sends the relevant data there and you can see it on IC.
Internally this can be see in debug variables (ic/cid side) as:
So here we can see that there's a car to the right that IS relevant to control and then ape communicated its existence to us.
I realize there's nothing groundbreaking as it was known for awhile the IC displays adjacent cars when you are changing lanes. Just showing some underlying mechanics. vehicle IDs are interesting, I am sure there were no 74 cars around me at just that moment, so there's probably some sort of memory in the system for that? Sides are still blind spots too as right about when this snapshot was taken the car tried to change lane into a car to the right of me that was not seen by the front cameras.
Internally this can be see in debug variables (ic/cid side) as:
Code:
DAS_autoLaneChangeState,InProgressR
...
DAS_cruiseLeadVehicle2Distance,51.000
DAS_cruiseLeadVehicle2ID,63
DAS_cruiseLeadVehicle2LatPosition,-1.400
DAS_cruiseLeadVehicle2RelVelocity,2.000
DAS_cruiseLeadVehicle2RelevantForControl,false
DAS_cruiseLeadVehicle2Type,Car
DAS_cruiseLeadVehicleDistance,20.000
DAS_cruiseLeadVehicleID,10
DAS_cruiseLeadVehicleLatPosition,-0.400
DAS_cruiseLeadVehicleRelVelocity,2.000
DAS_cruiseLeadVehicleRelevantForControl,true
DAS_cruiseLeadVehicleType,Car
...
DAS_cruiseRightVehicle2Distance,@invalid
DAS_cruiseRightVehicle2ID,@invalid
DAS_cruiseRightVehicle2LatPosition,@invalid
DAS_cruiseRightVehicle2RelVelocity,@invalid
DAS_cruiseRightVehicle2RelevantForControl,false
DAS_cruiseRightVehicle2Type,Unknown
DAS_cruiseRightVehicleDistance,29.500
DAS_cruiseRightVehicleID,74
DAS_cruiseRightVehicleLatPosition,2.800
DAS_cruiseRightVehicleRelVelocity,-2.000
DAS_cruiseRightVehicleRelevantForControl,true
DAS_cruiseRightVehicleType,CarSo here we can see that there's a car to the right that IS relevant to control and then ape communicated its existence to us.
I realize there's nothing groundbreaking as it was known for awhile the IC displays adjacent cars when you are changing lanes. Just showing some underlying mechanics. vehicle IDs are interesting, I am sure there were no 74 cars around me at just that moment, so there's probably some sort of memory in the system for that? Sides are still blind spots too as right about when this snapshot was taken the car tried to change lane into a car to the right of me that was not seen by the front cameras.
hm... I was examining a snapshot I accidentally made just a minute or so before the above data and noticed an interesting thing...
So I looked around and found a snapshot from a couple of days ago that had an even more interesting one
The timestamp is different from all the other cameras (a lot more in the past) so ape realized something is amiss I guess.
So I looked around and found a snapshot from a couple of days ago that had an even more interesting one
The timestamp is different from all the other cameras (a lot more in the past) so ape realized something is amiss I guess.
Can you explain what is interesting, I’m not seeing it =(
Similar threads
- Article
