Caporegime
Could a computer drive an F1 car around a track faster than Hamilton?
- Thread starter taximike
- Start date
Caporegime
In a word
NO
NO
Soldato
No, a computer could not a drive a car.
A computer could be used to process the software written to take in all of the parameters from the sensors recording every aspect of the cars physical state and produce appropriate parameters to be fed back into control systems that make the car move.
I know that seems pedantic, but there's a huge gulf between the OPs question and what actually would happen, to the point where the original question is moot and needs to be rephrased so that it reads "could a human design the software algorithms that would process the raw data from available sensors and covert them into appropriate results for currently available control systems".
In which case, I would have to say yes, a computer simulation could perform better than most human drivers, this is going on the success of automated control systems for aerospace applications.
However it is also likely that the human counterpart would always be able to adapt to the point where they can beat the system, unless the programmer allowed for the possibility of the cars destruction due to reckless (i.e. when it's uncertain whether a control result will cause the car to crash or get through a corner very quickly) driving, or not.
In short, no probably not, but never say never!
A computer could be used to process the software written to take in all of the parameters from the sensors recording every aspect of the cars physical state and produce appropriate parameters to be fed back into control systems that make the car move.
I know that seems pedantic, but there's a huge gulf between the OPs question and what actually would happen, to the point where the original question is moot and needs to be rephrased so that it reads "could a human design the software algorithms that would process the raw data from available sensors and covert them into appropriate results for currently available control systems".
In which case, I would have to say yes, a computer simulation could perform better than most human drivers, this is going on the success of automated control systems for aerospace applications.
However it is also likely that the human counterpart would always be able to adapt to the point where they can beat the system, unless the programmer allowed for the possibility of the cars destruction due to reckless (i.e. when it's uncertain whether a control result will cause the car to crash or get through a corner very quickly) driving, or not.
In short, no probably not, but never say never!
Caporegime
I think the biggest hurdle is finding a physical system that can steer as accurately and as quickly as a human does.
Electronic throttle ? yeah no problem, likewise with changing gears, its all done via electronics
But steering is still human, and i dont think you could find anything that could do it quick enough and accurate enough.
Electronic throttle ? yeah no problem, likewise with changing gears, its all done via electronics
But steering is still human, and i dont think you could find anything that could do it quick enough and accurate enough.
Soldato
ah you'd be surprised how quickly actuators etc. work. Take the wheel off, replace it with an accurate motor of some kind. Think how quickly stepped motors you see in mechanical hard drives work, and more importantly how accurately!
Mechanically speaking, replacing a human with motors and other controls is surprisingly easy. Even more so if you do it at the very base level before all linkages etc. are added to give the human a big wheel to grab onto!
Just have a look at the efforts seen in events like the DARPA challenge.
The problem is making sense of all the data, which will undoubtedly be very noisy as well (i.e. require filtering and therefore reduce its immediate accuracy) and turning that into a complex set of output variables.
Mechanically speaking, replacing a human with motors and other controls is surprisingly easy. Even more so if you do it at the very base level before all linkages etc. are added to give the human a big wheel to grab onto!
Just have a look at the efforts seen in events like the DARPA challenge.
The problem is making sense of all the data, which will undoubtedly be very noisy as well (i.e. require filtering and therefore reduce its immediate accuracy) and turning that into a complex set of output variables.
I think a 'computer' would be faster, sensors etc would be so much quicker than a driver's reaction time.
Steering would be no different than any other operation as manic man has quite rightly said, you could eliminate the steering wheel and other components.
I'd imagine you could use aeroplane auto-pilot software designs as a basis for something similar?
Steering would be no different than any other operation as manic man has quite rightly said, you could eliminate the steering wheel and other components.
I'd imagine you could use aeroplane auto-pilot software designs as a basis for something similar?
Last edited:
Nonsense, of course it's possible. Just look at fly by wire fighter planes etc..
Heck, a steering wheel is a simplistic device so that a human can operate it. A computer would have no problem steering all 4 wheels independently. Likewise it can modulate the braking on each wheel independently, and could precisely control the diff for better traction. I think a well made computer driver would be faster without a doubt.
Caporegime
begs the question why hasnt it been done ?
Caporegime
not even just an F1 car, any car really.
The BMW posted above is good, but nowhere near the required speed to drive an F1 car at the limit, let alone beat hamilton.
The BMW posted above is good, but nowhere near the required speed to drive an F1 car at the limit, let alone beat hamilton.
Because there is no need to spend billions on making a self driving car faster than a racing driver. It would be cool but it would serve no real purpose. No one would watch a sport that consisted of computer cars trying to better each other, it would be even more boring than F1 is now!
It would be a mammoth task, not really the sort of thing a few PhD students could knock out in a couple of years
The investment is far better off in the avenues it's in now, self driving cars that can drive safely on public roads, vastly more useful abilities in real terms.
Associate
- Joined
- 13 Aug 2008
- Posts
- 410
I think if you look at the technology in a modern cruise missile (which when compared to just something like a current Power7 or Xeon has a pathetic amount of power) - you begin to realize just how good computers can be at given tasks with the right programming.
Other examples would be F15/Eurofighter style fly by wire systems that can automatically trim out a fair amount of damage to the airframe. Or autopilots for Jets with landings - or the ESA space truck thing that can auto-dock with IIS.
I think that if computers can handle the above (more specifically the math for that many variables in real-time) - enough money could build the perfect race driver computer - but f1 is not the defence/space sector - and there budgets are not comparable
Make sense??
Other examples would be F15/Eurofighter style fly by wire systems that can automatically trim out a fair amount of damage to the airframe. Or autopilots for Jets with landings - or the ESA space truck thing that can auto-dock with IIS.
I think that if computers can handle the above (more specifically the math for that many variables in real-time) - enough money could build the perfect race driver computer - but f1 is not the defence/space sector - and there budgets are not comparable
Make sense??
Comparisons with aircraft don't really stand up as you have another dimension which offers a level of flex in the parameters you wouldn't have in an F1 car. While the speeds and dynamics are immense, 6 inches out at 50K feet is not the same as 6 inches out at Eu Rouge. Today I don't believe it would be possible as the parameters needed to cater for differing conditions would always be a compromise but it is absolutely feasible.
I think given a reasonable budget, say £10m it would easily be possible to do it.
The technology to physically control the throttle, steering and brakes is there, a servo for instance could react quicker than a human could move the wheel.
Most of the "feedback" from the car could be detected by sensors.
There is also the technology to place the car accurately via GPS and via imaging.
The most tricky thing I imagine would picking out standing water and dampness variations, finding the "dry" line.
The technology to physically control the throttle, steering and brakes is there, a servo for instance could react quicker than a human could move the wheel.
Most of the "feedback" from the car could be detected by sensors.
There is also the technology to place the car accurately via GPS and via imaging.
The most tricky thing I imagine would picking out standing water and dampness variations, finding the "dry" line.
I think that's vastly vastly underestimating the task.
Wait one guy suggests 10 mill and we are all branded? Harsh.