Unlimited Class Race Car

[Alastor]

That's what testing sessions are for. As RM says they'd get a rough idea before they show up and then test to get their final setup. The car could sense changing conditions and adjust, wouldn't be all THAT different from maximizing the course that a human driver is setting. Rain or something drastic like that would probably throw it for a loop, though. Eh, adjust to different settings as you tell it to come in to swap tires.

To Alastor's previous question, the line is programmed in. Same way they programmed the Audi to draw circle. That portion of my argument is very much in use at the moment, in that they all run lap-time simulation already to figureout the ideal line to develop their car setup to do rig testing. And the drivers now take that into account when they go to a track.

But as already stated that is not racing. It just following some predetermined set of parameters, it is not making decisions on the fly as the track conditions change due to weather, traffic, vehicle state, etc.

When the car brakes too early or late for a corner will it adjust itself for the next lap? If so, will it just guess based on some previously defined look-up table, or will it calculate the best solution on the fly based on data it collects as it circles the track? It seems to be the latter is what it would take to really run the fastest lap possible.

Now thrown in the some traffic, and the ideal racing line is no longer the ideal racing line, the car needs to resolve a new line on the fly. Likewise, when it rains the car needs adjust its line and speed accordingly.

Because UAVs keep coming up I would add that what we should be talking about is having a UAV run the equivalent of a Red Bull air race with multiple aircraft racing at the same time. Not impossible, but far from simple.

As already stated MANY posts before, it's just too slow and not accurate enough ( even WITH differential ). Vision ( and physical feedback ) is about the only solution to fine tune the best outcome. Some of the issues on timing on the math already been given. Sample rates of video capture needs to be an order of magnitude higher than we have at reasonable cost/weight. QUality needs to be equally as top-notch. So now we're looking at processing 270 degrees of vision at >HD quality at frame rates wll in advance of 30fps and identify objects in those frames. Again, are we even CLOSE to doing that in real time at low res and low frame rates. BARELY.

I worked on a similar system that would combine multiple SD/HD video sources into a single large mosaic. It needed nearly a dozen video devices to cover the full range of view. Each device required a GPU and Network Card. I remember correctly every two cameras required a separate CPU. To meet the ‘near real time’ requirement the system would have to process anywhere between a few hundred megabytes to a few gigabytes worth of data in less than 50 ms. That means capture, process, display, etc….~200-1,500 MBs of data in less than a twentieth of a second. That was challenging enough to just stitch the images together using specialized GPUs, it would have been even harder with additional post-processing.

[RacingManiac]

Monaco, v Spa v Monza are spring and damper settings, ride hieght, downforce, tyre pressures, wing settings. The basic car is the same, thus why we get some chassis are better at some circuits.

Depends on what you define as "basic car". If you were to be anal, this theoretical car with active whatnot might actually need less parts changed between tracks since it will likely to be designed to work in a range of parameter. I'd imagine being able to run wings at Monaco level at low speed corner and trim it out to Le Mans level on the straight. The only thing that might change will be the codes of your program.

Currently cars run at least different floor, front and rear wing and maybe brake duct configuration at a minimum between specific tracks. The physical wing themselves are different from race to race, not just what angles are set at(most if not all rear wings in F1 now are fixed and non-adjustable). The designs were derived long before they arrive at that race weekend. Some cars(Mercedes for example this year), even runs different suspension arms that alters the wheelbase and kinematics between tracks, and track like Monaco at least requires everyone to run different steering rack to deal with the hairpins. Not to mention the whole deal of people putting on and taking off F-duct. So in essence in that respect the only thing that is constant is the tub, and that is not necessarily the case in the past neither. Its only recently that they introduced rule to homologate the chassis design in the beginning and they can't change it....

Taking the whole computer driven car out of the equation, the "ultimate" car shouldn't need track specific setup physically.

Like I said, I don't think wheel to wheel racing is feasible, nor really kinda fits the spirit of competitive sport since at that level you need to be comparing human as opposed to program #1 to program #2. As far as an engineering exercise in the competition to lap around the track as fast as possible(hell, they still race Time Attack and Solo "racing" don't they?), I don't see why this can't be done. The only reason why its NOT done is because there are no reason to do it. Capitalistically if there are no commercial interest and no real scientific gain to be had, no one would do it anyway.

[wwgkd]

NEVER ONCE SAID THEY WERE MAKING CARS.

HAVE said they spend billions on "UAVs" and next gen autonomous.
Every smart bomb is autonomous once launched

Still missing the point and need to think how FAST things happen on a race track. Sonar ? At VERY short distances the error increases the lower the frequency of the pulse and the pulse gaps. So unlike parking which can wait 1/4s in it's control loop then you've hit the car beside you in turn one BY THE TIME YOU'VE WORKD OUT he's there

"remote" can never work as so much of the real time control be a driver is that difficult to quantify "feel". In a light weight car you can feel fuel slosh in a tank in quick esses. It's why any serious race/sprint caterham etc run race fuel tanks in place of the road one. Without that feel then the loop is too slow to react having to wait until the OUTCOME of the movement, THUS why I keep pointing out that thinking flight ( which is where lots of research on autonomous control is on ) isn't useful at all. Because the consequences in flight are nothing. On track at higher race speeds than we now see ... it's a barrier - or the car beside you.

Also there's a feel that "testing sessions" are about sorting out the control system and I don't see that as in the remit of RACING. Sorting out the car setup yes, as that's what happens now. But how coudl it be "racing" ( which is where the remit of the thread went to ) if a team are allowed to effectively BUILD a different car for each track ? ie fundamentally change add/remove control features to match the one track in question. I'd say that's cheating the original premise.

Yes, the military has spent billions on making UAVs, and now civilian companies are taking advantage of much of the research done, or are paralleling it for far less. And in this exercise we can take advantage of that. I’m not suggesting we start over from the stone age.

You’re right, I assumed when you kept saying the military can barely build semi-autonomous vehicles that you were referring to ground vehicles, since we've had fully autonomous aircraft that can even land themselves and designate targets (as accurately as a pilot) for a while now. They've also had aircraft perform terrain following maneuvers at higher speeds and lower altitudes than a human pilot can achieve. I'd say that's fairly close to what you're saying is impossible here.

We're not talking about low frequency sonar, I specifically mentioned high frequency sonar, like what several car companies are using. Yes what their using isn't applicable here, but said anything that shows that a higher cost version couldn't be useful. I'm guessing F1 drivers would appreciate something more than tiny vibrating mirrors to let them know when somebody is in one of the many large areas that they're not able to see at all. You keep saying GPS isn't accurate enough but human drivers are even less accurate.

I have said that wheel to wheel racing would take some time, but there are many forms of "racing" that don't require 20 vehicles wheel to wheel on the track, such as some of the videos you've posted of your own driving. Time attack is seeing who can get their car around the track the fastest, right? This is basically what we’re talking about, but with more nerds involved.

But as already stated that is not racing. It just following some predetermined set of parameters, it is not making decisions on the fly as the track conditions change due to weather, traffic, vehicle state, etc.

When the car brakes too early or late for a corner will it adjust itself for the next lap? If so, will it just guess based on some previously defined look-up table, or will it calculate the best solution on the fly based on data it collects as it circles the track? It seems to be the latter is what it would take to really run the fastest lap possible.

Now thrown in the some traffic, and the ideal racing line is no longer the ideal racing line, the car needs to resolve a new line on the fly. Likewise, when it rains the car needs adjust its line and speed accordingly.

Because UAVs keep coming up I would add that what we should be talking about is having a UAV run the equivalent of a Red Bull air race with multiple aircraft racing at the same time. Not impossible, but far from simple.


I worked on a similar system that would combine multiple SD/HD video sources into a single large mosaic. It needed nearly a dozen video devices to cover the full range of view. Each device required a GPU and Network Card. I remember correctly every two cameras required a separate CPU. To meet the ‘near real time’ requirement the system would have to process anywhere between a few hundred megabytes to a few gigabytes worth of data in less than 50 ms. That means capture, process, display, etc….~200-1,500 MBs of data in less than a twentieth of a second. That was challenging enough to just stitch the images together using specialized GPUs, it would have been even harder with additional post-processing.

Well that gets into definitions. Is racing only racing when a human is making all the decisions? Also, how far ahead of time can the human make those decisions? It could adapt within set parameters to those conditions according to what the human decides, so in effect it is similar to racing with a techie instead of a driver.

Given these budgets I’m guessing the UAV Red Bull Air race wouldn’t be far off, multiple planes is tough though. And not something that even the Red Bull racers do.

Out of curiosity what were you trying to do for that project? Or is it not something you can talk about?

Depends on what you define as "basic car". If you were to be anal, this theoretical car with active whatnot might actually need less parts changed between tracks since it will likely to be designed to work in a range of parameter. I'd imagine being able to run wings at Monaco level at low speed corner and trim it out to Le Mans level on the straight. The only thing that might change will be the codes of your program.

Currently cars run at least different floor, front and rear wing and maybe brake duct configuration at a minimum between specific tracks. The physical wing themselves are different from race to race, not just what angles are set at(most if not all rear wings in F1 now are fixed and non-adjustable). The designs were derived long before they arrive at that race weekend. Some cars(Mercedes for example this year), even runs different suspension arms that alters the wheelbase and kinematics between tracks, and track like Monaco at least requires everyone to run different steering rack to deal with the hairpins. Not to mention the whole deal of people putting on and taking off F-duct. So in essence in that respect the only thing that is constant is the tub, and that is not necessarily the case in the past neither. Its only recently that they introduced rule to homologate the chassis design in the beginning and they can't change it....

Taking the whole computer driven car out of the equation, the "ultimate" car shouldn't need track specific setup physically.

Like I said, I don't think wheel to wheel racing is feasible, nor really kinda fits the spirit of competitive sport since at that level you need to be comparing human as opposed to program #1 to program #2. As far as an engineering exercise in the competition to lap around the track as fast as possible(hell, they still race Time Attack and Solo "racing" don't they?), I don't see why this can't be done. The only reason why its NOT done is because there are no reason to do it. Capitalistically if there are no commercial interest and no real scientific gain to be had, no one would do it anyway.

Well said, but as for the last bit I look at it as racing for people with the know how but not the abilities or bravery/stupidity (admit it, some forms of racing are just dangerous, especially in the past “glory” years.) You’d be comparing engineering and design talent rather than the combination of that and drivers, and that can still be racing. Just not worth the budget. Not that I’d mind of somebody spent it but I don’t pay to go to races as is.

[Kitdy]

First off, I am not a tech expert (especially when it comes to computers and telecommunications) so Matra, don't tear me a new asshole for being really wrong. Also presume we have the "freshest" available tech. I want to make this utterly ridiculous.

Ok, how about you get a driver on a track to set a hot lap (the fastest lap at the circuit), then get the car to mimic this to perfection. I know there will be variances and the car will need to react (there will need to be logic). Then, you focus on cutting off tiny fractions of a second with the computer car - a bit more precise on throttle application than the human, a bit closer to the apex, whatever. You reference this new information with the ideal human hot lap so maybe the calculations for the computer car (if it manges a bit more top speed on the straight due to it being faster on the throttle, or is a bit shifted for some reason - beneficial or not) and maybe the calculations become easier?

Or what if we used the most powerful computer available and had it at the track and used some very fast connectivity to transmit information to the car and for the car to decipher it (with a powerful computer onboard) in nearish real time? It would seem as though the car could not process the vast information that the SC gave to it in time to react though.

GPS is no good, and I think Matra said other types of referencing - RADAR or some EM shit of some kind may not be good enough.

I hear Matra about to tear my ideas apart in my head - something like this: "Yae cannae break the laws of physics Captain!"

This is practically impossible it seems (when I think of the relatively unadvanced state of driverless cars though, I guess it's no surprise) even - when you make things quite impractical it may not be possible.

Maybe though Matra is wrong? Probably not, but maybe.

[ruim20]

Like i said before, the cars must be piloted by a human driver, drones don't have the same feeling has watching a human performing. drivers have faces we relate to, if they can't take the Gs, make something in the car that could help with that, be it mechanical or electronic. Active or passive.

[Matra et Alpine]

They've also had aircraft perform terrain following maneuvers at higher speeds and lower altitudes than a human pilot can achieve. I'd say that's fairly close to what you're saying is impossible here.

As I'd said .. NOTHING like it
In a plane a 10 foot variance is NOTHING and the control system has tons of time to respond to the unexpected. In a race car in some circuits 10 INCHES is a wall SO we are clearly ORDERS OF MAGNITUDE more complex requiring speed and decision making.

We're not talking about low frequency sonar, I specifically mentioned high frequency sonar, like what several car companies are using.

Current car sonor brings it limits ( the laws of physics yes ).
It takes 30milliseconds for a pulse to travel 1m and 30 milliseconds back.
So 60milliseconds to get an indication of the presence or position.
THEN the system has to react.
This is the basic problem in any equation based control system, it waits for an input and then acts by calculation large numbers of inputs and then determines an output.
THe brain of a driver uniquely projects forward using "feel" and "instinct" and not equations. We've not yetr managed to approach this complexity in computing and certainly NO WHERE NEAR the speed and "vaguness" of the process in Alonso's brain.

The mirrors aren't the real issue with the cars, it's the safety high sides which prevent the driver being able to LOOK by moving his eyes/head
Mirrors just then make it worse.
Of course any alternative THEN has the ergonomics issue oif HOW to inform the driver. See we've develoepd a cortical system over millions of years of evolution and a brain which processes much of it "on the fly" and adaptively. So most of the time you're not aware of peripheral vision or what you see. But simple test. There aren't many red cars on the road now. Youve' not seen one for years. You will see one today They were always there, just the "brain" is discarding the vision input that spotted it.
THAT is what I keep coming back to and is the intractable issue all advanced computer manangement systems face. HOW do we "make a person" thinking/reacting machine.

You keep saying GPS isn't accurate enough but human drivers are even less accurate.

Not thinking like a driver.
An F1 driver can place a front wheel to within millimetres on EVERY lap.
We use our visual system to provde OUR positioning.
Nobody gives a toss on a race circuit which lattitude and longtitude you're at and how accurate it is ( which is GPS role ) BUT use vision ( and feel ) to determine position WAY more accurate than we can ever even DREAM about getting systems like GPS to use. As I'd alluded to before, you COULD add lots of sensors and signals aroudn a track to try to get that close. But it ain't GPS and I'd say "cheating" as the track is being used to imrpve the car Thus why only vision has a chance of being a valid someday workable solution. ( hmm maybe phased radar arrays, but then the car would need a transmitter license and be kept away from airports AND weigh a couple of tons )

Time attack is seeing who can get their car around the track the fastest, right? This is basically what we’re talking about, but with more nerds involved.

THe "other cars on ther track" is jsut an added complexity.
JUST track position and is enough of an issue WHEN DRIVING VERY FAST - whcih is our remit in this mental masturbation exercise.

Also, how far ahead of time can the human make those decisions? It could adapt within set parameters to those conditions according to what the human decides, so in effect it is similar to racing with a techie instead of a driver.

And techie's can't match Alonso
Human decision making is horrendously complex because EVERY time we identify "error" in our model we on the fly adapt it.
Thus why equations will struggle at those limits because at those limits lots of VERY SMALL variables can have massive inputs. I've already said, this is the realsm of chaos-theory math and not undergrad simultaneous equations stuff.
Adaptign within small well defined limits will no doubt get a car moving.
I jsut contend when we think of the number of varibales that get added the faster we go that the problem is currently intractable. Every ONE varibale introduced increse the problem by the power of the varibales already present. Thus why before when 500 data acquisition inputs in F1 was cited I'd pointed out that in sim-eq world that has given us 500 to the power 500 and that is one HUGE number ( online power calc jsut barfed and said "infinity" )
YES, an over simplification of the complexity, but even if I'm out by a thousand orders of magnitude it's still HUGE

Given these budgets I’m guessing the UAV Red Bull Air race wouldn’t be far off ...And not something that even the Red Bull racers do.

I agree to a degree, because you can easily be 50 feet out and then get it all back in line for the "gate" AND the gates have a failry large ( >10 ft ) spare space. Repeat, in car race circuit THAT is a concrete wall the vehicle just hit.
However, when I then try to consider how FAST I want to do it and that wind and more importantly tuurbulence from last time through that bit of space introduces variables that I contend we are back at the having to fly slower so enough time to process and handle. eg go back to the multi rotor vid I posted. Closed room MUCH simplified environment !

First off, I am not a tech expert (especially when it comes to computers and telecommunications) so Matra, don't tear me a new asshole for being really wrong.

Never the intent unless I find repeating somethign for the tenth time and clear the reader is ignoring it

Ok, how about you get a driver on a track to set a hot lap (the fastest lap at the circuit), then get the car to mimic this to perfection. I know there will be variances and the car will need to react (there will need to be logic).

The remit was to drive faster and as cars are on physicl limits on grip at times then the "variances" happen very often and very quickly. So we can't avoid the "elephant in the room" .... getting it roudn is easy. Getting it round faster than a driver is the issue.

Then, you focus on cutting off tiny fractions of a second with the computer car - a bit more precise on throttle application than the human, a bit closer to the apex, whatever. You reference this new information with the ideal human hot lap so maybe the calculations for the computer car (if it manges a bit more top speed on the straight due to it being faster on the throttle, or is a bit shifted for some reason - beneficial or not) and maybe the calculations become easier?

Certainly I'd already assumed we had the faster reaction bit down pat and had mentioned it.
But WHEN and HOW MUCH on each lap becomes the real task.
Driving fast is about getting somethign JUST over the limit on each lap and bring it back and have that as a goal to then work around to find alternative strategy to push it slightly more each time.

Or what if we used the most powerful computer available and had it at the track and used some very fast connectivity to transmit information to the car and for the car to decipher it (with a powerful computer onboard) in nearish real time? It would seem as though the car could not process the vast information that the SC gave to it in time to react though.

THat's not really the "car" then
You'd built the car in that case much as the human brain works.
So the car woudl have some very fast control loops in place to give immiediate reaction stuff and clear and fairly simple control inputs the SC woudl be handling the higher intelligence and feedback. All the issues on vision etc all disappear whn you let an external system handle it. Put a stereo HD camera on the car and broadcast at 6-10GHz to an array of video processing systems dedicated to ONLY work out position. HUGE processing needs. Again, if we put markers then you coudl simplofy things, but I wans't wanting it to be a race as fast as a car on a track modified to give the computer an advantage eg like the little balls used on motion capture
Still needs to be damn quick but the SC would treat the car like a slave. Tell it what to do knowing it will do it .. and avoid doing anything stupid. Tell it often enough and you'll be fast. Kind of like Scalextric Yes, but that takes us off to where we werne't ... eg a "car" able to race as fast.

I hear Matra about to tear my ideas apart in my head - something like this: "Yae cannae break the laws of physics Captain!"

KNow you meant that with love
But you're right.
But it's only TODAYS laws of physics.
20 years ago we had to have dedicated computers and HUGE complxity in programming to recognise speech even after training it to one speaker for days. Today you get it for any person with only a few minutes training on your WIndows7 for FREE
Computers are MUCH faster and more capacity, but in that domain most of all was a realisation of the way we form speech sounds. ( I had a research student doing speech back in the 80s and used recognition almost daily since the 90s )
Vision is where speech was back then.
IN time, we'll grasp more efficient ways to encode and decide on video. At the moment we're seeing some of it coming in very specific areas - like face rocognition. Until then it will take VAST computing to brute force number crunch ( the "current" laws of physics cap'n )

[Matra et Alpine]

Maybe though Matra is wrong? Probably not, but maybe.

Hey, I always start with the premise that what I knew yesterday was wrong and seek to increase understanding and knowledge. It's why I worked in HP R&D for 30 years I met people there who tore MY concepts apart and I thanked them for it. KNowledge and improving it can only be a good thing. Any idea of "Protecting an idea" is pointless.

Thus why I'd gone and foudn the clips and items on the 4 rotor autonomous research and the audi bit. If you don't attack your own ideas how can you defend them

I'm loving (most of) this as we've a flow of ideas and questions.

Each one shodl be bringing each of us to question the fundamental, to explore the issue and possibility and to evlauate the concept openly.

Well I hope so.



Just when I do that I post my opinion and with the reasoning to get there.

Sorry some if that seems "expert" or "arrogant".

The only thing I've ever been good at is questioning the obvious and then looking for possible answers and evaluating them.

[wwgkd]

As I'd said .. NOTHING like it
In a plane a 10 foot variance is NOTHING and the control system has tons of time to respond to the unexpected. In a race car in some circuits 10 INCHES is a wall SO we are clearly ORDERS OF MAGNITUDE more complex requiring speed and decision making.

I’m talking terrain following, not 30,000 feet up. At very high speeds below 100 feet you don’t have a 10 foot variance, and certainly the trees are coming at you faster than those concrete walls. Like I said this was performance that the best pilots can’t match, or do you think that F1 drivers are so far beyond the best pilots?

Current car sonor brings it limits ( the laws of physics yes ).
It takes 30milliseconds for a pulse to travel 1m and 30 milliseconds back.
So 60milliseconds to get an indication of the presence or position.
THEN the system has to react.
This is the basic problem in any equation based control system, it waits for an input and then acts by calculation large numbers of inputs and then determines an output.
THe brain of a driver uniquely projects forward using "feel" and "instinct" and not equations. We've not yetr managed to approach this complexity in computing and certainly NO WHERE NEAR the speed and "vaguness" of the process in Alonso's brain.

1. That was a response to your bashing the use of low frequency radar as inaccurate when it had never been mentioned.
2. I think I could handle a whopping 30 milliseconds considering the various nerve impulses required for a human to react aren’t a whole lot faster.

The mirrors aren't the real issue with the cars, it's the safety high sides which prevent the driver being able to LOOK by moving his eyes/head
Mirrors just then make it worse.
Of course any alternative THEN has the ergonomics issue oif HOW to inform the driver. See we've develoepd a cortical system over millions of years of evolution and a brain which processes much of it "on the fly" and adaptively. So most of the time you're not aware of peripheral vision or what you see. But simple test. There aren't many red cars on the road now. Youve' not seen one for years. You will see one today They were always there, just the "brain" is discarding the vision input that spotted it.
THAT is what I keep coming back to and is the intractable issue all advanced computer manangement systems face. HOW do we "make a person" thinking/reacting machine.

So if a car was able to sense what was behind/to the side of it, then it’s already ahead of a human driver. I wasn’t talking about the way a brain works, I was talking about abilities. Abilities that the drivers would like to have. You don’t have to make a car that thinks or reacts like a person, because all this would be done in a different way than trying to imitate a person. Trying to make a car drive itself by imitating the way a person does it is like trying to make a robot that looks like a person, there’s really no point function wise.

Not thinking like a driver.
An F1 driver can place a front wheel to within millimetres on EVERY lap.
We use our visual system to provde OUR positioning.
Nobody gives a toss on a race circuit which lattitude and longtitude you're at and how accurate it is ( which is GPS role ) BUT use vision ( and feel ) to determine position WAY more accurate than we can ever even DREAM about getting systems like GPS to use. As I'd alluded to before, you COULD add lots of sensors and signals aroudn a track to try to get that close. But it ain't GPS and I'd say "cheating" as the track is being used to imrpve the car Thus why only vision has a chance of being a valid someday workable solution. ( hmm maybe phased radar arrays, but then the car would need a transmitter license and be kept away from airports AND weigh a couple of tons )

Again, GPS can be accurate to 2mm. Global Positioning System - Wikipedia, the free encyclopedia

If human drivers were that accurate corner after corner, lap after lap, then they’ve been out there trashing millions of dollars of machinery on TV just for the fun of it.

THe "other cars on ther track" is jsut an added complexity.
JUST track position and is enough of an issue WHEN DRIVING VERY FAST - whcih is our remit in this mental masturbation exercise.

And techie's can't match Alonso

For charisma? You should meet some of the techies I know.
For “feel?” No.
For technical know how and understanding? Absolutely.

Human decision making is horrendously complex because EVERY time we identify "error" in our model we on the fly adapt it.
Thus why equations will struggle at those limits because at those limits lots of VERY SMALL variables can have massive inputs. I've already said, this is the realsm of chaos-theory math and not undergrad simultaneous equations stuff.
Adaptign within small well defined limits will no doubt get a car moving.
I jsut contend when we think of the number of varibales that get added the faster we go that the problem is currently intractable. Every ONE varibale introduced increse the problem by the power of the varibales already present. Thus why before when 500 data acquisition inputs in F1 was cited I'd pointed out that in sim-eq world that has given us 500 to the power 500 and that is one HUGE number ( online power calc jsut barfed and said "infinity" )
YES, an over simplification of the complexity, but even if I'm out by a thousand orders of magnitude it's still HUGE

Again we’re not trying to do things the way humans do. If we always did that then calculators wouldn’t be any faster than humans and would be so expensive as to not be worthwhile. Instead they solve large equations as a series of simple tasks performed very, very quickly and are often handed out for free. They can’t do everything in the math world that a human can, but what they do they do very well. That’s all we’re trying to (theoretically speaking) do with a car here. Driving a car autonomously has happened. Variables are added when you drive the car at its limits, and that’s happened. More variables are added when driving at extremely high speeds certainly, but not 500^500.

I agree to a degree, because you can easily be 50 feet out and then get it all back in line for the "gate" AND the gates have a failry large ( >10 ft ) spare space. Repeat, in car race circuit THAT is a concrete wall the vehicle just hit.
However, when I then try to consider how FAST I want to do it and that wind and more importantly tuurbulence from last time through that bit of space introduces variables that I contend we are back at the having to fly slower so enough time to process and handle. eg go back to the multi rotor vid I posted. Closed room MUCH simplified environment !

Flying a plane quickly, in an environment where there is very little space/time to react has been accomplished. It can react to turbulence and such or it wouldn’t have been accomplished. This would mostly be an issue of adapting it to a plane with different capabilities. Which given someone with the money and the desire, could be done.

[Matra et Alpine]

I’m talking terrain following, not 30,000 feet up. At very high speeds below 100 feet you don’t have a 10 foot variance

erm, YES THEY DO.

and certainly the trees are coming at you faster than those concrete walls. Like I said this was performance that the best pilots can’t match, or do you think that F1 drivers are so far beyond the best pilots?

Even the BEST pilots don't go below 50ft unless they have LONG clear sight.
Dad used to go lower when he was fighter training in Canada in WW2 and his pilot log is full of reprimands for wheeat in the undercarriage

2. I think I could handle a whopping 30 milliseconds considering the various nerve impulses required for a human to react aren’t a whole lot faster.

We're talking about COMPUTER response times to complex inputs using a non-deterministic approach to chose outcomes.
And if you think you can respond to a complex input in 30msec then you are not human It is 10 TIMES THAT on average.
Even our fastest - the blink on touch response is 100msec.
See for a VERY SIMPLE visual reflex test and databse of results ... http://www.humanbenchmark.com/tests/...time/index.php

That's one fifth of a second. At 150mph you've just gone 50 FEET !!
Now let's factor in that it's not jsut a SIMPLE response, but is one which takes in to account a vehicle and the decision is about whether to steer, throttle, brake or just "hang on" ? The math really is THAT simple and why I'd said earlier that the best drivers have automatic learned responses that we dont' understand ( and someonhow have to capture in any control loop hoping to better them )

So if a car was able to sense what was behind/to the side of it, then it’s already ahead of a human driver. I wasn’t talking about the way a brain works, I was talking about abilities.

No point KNOWING somethign is there.
ACTING on it is necssarya dn that's the whoel point of all the areas I'm pointing us to concentreate on.
MEASURING is relatively easy - though expensive and heavy.
MONITORING is viable but not for very fast and larege numebrs of points
MANAGEMENT is horrednously complex and especially whne trying for "soft skills".
This "3-loop" is a capability core to HP and Agilents systems and an area we spent millions in research across 4 sites

Trying to make a car drive itself by imitating the way a person does it is like trying to make a robot that looks like a person, there’s really no point function wise.

Getting hung up on the PHSYICALITY.
NOT what I'm talking about as THE challenge in doing the task at hand.

The challenge is to make a car drive faster.
Now given the SAME CAR then it comes down to operating the same inputs.
HOW does a driver decide whether to accelerate, brake or turn ?
Well at low speeds it's obvious - at high speeds then it's not.
OTHERWISE, we'd all be able to drive F1 cars ? Do you think you coudl ????

Again, GPS can be accurate to 2mm. Global Positioning System - Wikipedia, the free encyclopedia

You've not studied that
2 PARTS PER MILLION (ppm no mm) is their error spread on a 1-3cm accuracy on a system which relies on having VERY accurate clocking and stationary objects and referring back to known HIGHLY ACCURATE positions on a regular basis !
Please research ideas before adding them as ity's jsut wasting time and others efforts to respond to googling efforts. bad

If human drivers were that accurate corner after corner, lap after lap, then they’ve been out there trashing millions of dollars of machinery on TV just for the fun of it.

Sorry but you have missed 100% of the issue.
THEY can position it. BUT what *it* is varies on every lap.
The optimum turn in changes, the grip at point B changes, the temp in the tyre changes.
No poitn me repeating all the REALITY of driving fast if anyone thinks that only the LINE matters. ONLY IN console arcade games does that BS work

Again we’re not trying to do things the way humans do. If we always did that then calculators wouldn’t be any faster than humans and would be so expensive as to not be worthwhile. Instead they solve large equations as a series of simple tasks performed very, very quickly and are often handed out for free. They can’t do everything in the math world that a human can, but what they do they do very well. That’s all we’re trying to (theoretically speaking) do with a car here. Driving a car autonomously has happened. Variables are added when you drive the car at its limits, and that’s happened. More variables are added when driving at extremely high speeds certainly, but not 500^500.

^COmpletely wrong assumptions.
Not wasting my time explaining why as it's common sense and has been covered many times.
I can't make horses drink.

Flying a plane quickly, in an environment where there is very little space/time to react has been accomplished. It can react to turbulence and such or it wouldn’t have been accomplished. This would mostly be an issue of adapting it to a plane with different capabilities. Which given someone with the money and the desire, could be done.

No point carrygin on.
Those planes are NOT flying ALL the time to mm accuracy.
THey are pitching 10 feet MINIMUM.
I'm out of here as it's pointless.

[Matra et Alpine]

I'm out of this guys. Enjoy the rest and I look forwrd to the outcome -- hopefully WOuter will keep this thread open for 18 years so I can come back and go "told you"

[f6fhellcat13]

The No-Holds-Barred Race Car Trapped Inside Gran Turismo 5

When Gran Turismo 5 finally arrives, it will include the Red Bull X1 prototype, the answer to the question of "What if you designed a race car with an unlimited budget and no regard for rules?" Gas turbine, engage.

While we wait and wait for GT5, the specs on the Adrian Newey-shaped virtual machine just add to the pain of patience: 1483 hp and 527 lb.-ft. of torque from a gas turbine, moving 1,201 lbs. of carbon fiber to a top speed of 249 mph. Someday maybe you too can try to best the Suzuka F1 record by 20 seconds.

[Matra et Alpine]

Gas turbines have even less torque than the RX-8 at low revs.
Intrigued to see how many gears they need to make it workable.
EDIT: aha, found another link that it will be CVT.
So I suspenct the belts and variable pulleys are assumed to be made of unobtanium and mithril
Sexy look tho' and fantastic product positioning as ever by Red Bull !!!

[RacingManiac]

I guess this will be unlimited but somewhat grounded in reality and self-imposed constraint...

Cool RBR and Adrian Newey tie-in though...

[RacingManiac]

[ame="http://www.youtube.com/watch?v=S_eVE6KQ4Jg&feature=relmfu"]http://www.youtube.com/watch?v=S_eVE6KQ4Jg&feature=relmfu[/ame]

Prius, our future unmanned race car.....

I'd be interested to see how consistent the laptime it'll do....

[csl177]

^^^ If it's "reading" the cones, I'm in... how much to install that rig in my next autocrosser? Guaranteed FTDs!