RE: BMW launches latest ActiveAssist tech
Discussion
TLandCruiser said:
rijmij99 said:
We have lots of automated technology that still needs qualified human oversight so I wouldn't worry just yet.
Now when we automate everything and have no need for humans at all, that's when you worry
Personally it will be a sad day, when you get into the back of your driverless car to take you to the train station, you buy a train ticket at the automated machine, get on the driverless train. When you arrive at the airport you check in at the automated check in to drop your luggage off, you then go to marks and Spencer's where you buy some lunch an pay at te self check out to board your flight with a pilotless airplane.Now when we automate everything and have no need for humans at all, that's when you worry
Obviously that is in the far future but it is a reality at some point and is a great shame as the level of human interaction comes less and less.
RemarkLima said:
The other way to look at this, is the day when people do not need to do menial jobs, scanning items and taking money endlessly, stacking shelves, etc... is surely a great freedom? It'll cause some pain short term, but ultimately, who wants to do a dead end, meaningless job? At that point, your "human interactions" should ideally become more meaningful, as people have more time to read, paint and educate themselves... *ahem* or just watch Jeremy Kyle.
This is getting political I know, but if that happens, what do all the people who previously stacked shelves to earn money do?!Can't belive it'll be calibrated to promote oversteer. Though maybe if you get the car into a situation where the only option is to drift out of danger... can that ever really happen? Surely it could reign things in before needing to actively drift, unless provoked by someone who knows how the system works and is tricking it into drifting.
I wonder how it knows where you should be going... e.g. what if it 'corrects' your trajectory and pilots your towards something you can see in the distance but that the radar hasn't picked up yet. It MUST be clever enough for this not to happen of course
I wonder how it knows where you should be going... e.g. what if it 'corrects' your trajectory and pilots your towards something you can see in the distance but that the radar hasn't picked up yet. It MUST be clever enough for this not to happen of course
It's the same argument as "automated vs manual gearbox" really. If a car can do something better than you can, would you like it to do it? In my case, the answer is generally (but not always) no. I can quite see why for many people the answer would be "yes", though.
Regardless of the fact that I don't want it, it's impossible to deny that the technology required for this sort of system is very, very impressive.
Regardless of the fact that I don't want it, it's impossible to deny that the technology required for this sort of system is very, very impressive.
The fundamental issue with all this assisted driving tech is what do you use as the primary input for the cars trajectory? i.e.
In a purely mechanical car the driver uses their senses (sight, feel, balance etc) to work out where they want to go at any given moment, and apply handwheel and throttle/brake inputs to make the car go in that direction. For most people, most of the time, this is within the purely linear and positive region of control (ie. more handwheel = more turning etc). Once tyre slip at the non steered axle exceeds slip+turn at the steered end, the handwheel direction reverses, and the control becomes very non linear (as the inertia of the car starts to play a large part in the control lag). At this point, most drivers used to crash, but good ones could still operate in this highly non linear region (through experience, training and of course talent!) and continue to make the car follow their chosen trajectory.
Fast forward to the typical car with ESP, and the issue becomes one of deciding where the driver wants the car to go, and helping him go that way. One of the reasons most road ESP systems are thought by skilled drivers to "cut in too early" is because they must try to ascertain the intended vehicle direction before the controls become non linear. Systems that have a high degree of yaw authority and follow the handwheel angle religiously can cause unexpected events during oversteer! ie. the driver has left hand lock on, as they want to corner to the left, but the car oversteers halfway around the bend, so they now put on righthand lock. Should the system now try to turn the car right?
Anyone who's driven a fully "active" WRC car will have learn't to avoid excessive oversteer, because the system really could make the car suddenly reverse it's direction mid corner! As we currently have no other arbitor of intended direction than the handwheel, these systems become a compromise, because they have to work with drivers who's response ranges from "do nothing, just freeze" to "perfect application of opposite lock and drive torque". That is extremely difficult to calibrate for.
So newer systems, like the BMW system shown, now have the capability for full yaw authority as they can now steer themselves with the steering rack. But this means that unless the system somehow uncouples the handwheel from the rack, it cannot use the handwheel angle as it's trajectory request, it must take information from other sources (GPS, radar, laser scanners etc). For such a track test as shown in that video, that is quite easy, as it is not hard to pre-program a simple course through some cones under controlled conditions. However, in the real world, the task of integrating that system with a "real" driver and obeying their requests is very very difficult indeed. In fact, it would be a lot simpler to just cut out the human driver entirely, as that BMW video shows (ie. hands off all the controls, car does the driving!)
Historically, the aerospace industry has show that the issues surround the cases when both the human and the computer are both thinking the other one is "flying the plane" (see AF447!).
In a purely mechanical car the driver uses their senses (sight, feel, balance etc) to work out where they want to go at any given moment, and apply handwheel and throttle/brake inputs to make the car go in that direction. For most people, most of the time, this is within the purely linear and positive region of control (ie. more handwheel = more turning etc). Once tyre slip at the non steered axle exceeds slip+turn at the steered end, the handwheel direction reverses, and the control becomes very non linear (as the inertia of the car starts to play a large part in the control lag). At this point, most drivers used to crash, but good ones could still operate in this highly non linear region (through experience, training and of course talent!) and continue to make the car follow their chosen trajectory.
Fast forward to the typical car with ESP, and the issue becomes one of deciding where the driver wants the car to go, and helping him go that way. One of the reasons most road ESP systems are thought by skilled drivers to "cut in too early" is because they must try to ascertain the intended vehicle direction before the controls become non linear. Systems that have a high degree of yaw authority and follow the handwheel angle religiously can cause unexpected events during oversteer! ie. the driver has left hand lock on, as they want to corner to the left, but the car oversteers halfway around the bend, so they now put on righthand lock. Should the system now try to turn the car right?
Anyone who's driven a fully "active" WRC car will have learn't to avoid excessive oversteer, because the system really could make the car suddenly reverse it's direction mid corner! As we currently have no other arbitor of intended direction than the handwheel, these systems become a compromise, because they have to work with drivers who's response ranges from "do nothing, just freeze" to "perfect application of opposite lock and drive torque". That is extremely difficult to calibrate for.
So newer systems, like the BMW system shown, now have the capability for full yaw authority as they can now steer themselves with the steering rack. But this means that unless the system somehow uncouples the handwheel from the rack, it cannot use the handwheel angle as it's trajectory request, it must take information from other sources (GPS, radar, laser scanners etc). For such a track test as shown in that video, that is quite easy, as it is not hard to pre-program a simple course through some cones under controlled conditions. However, in the real world, the task of integrating that system with a "real" driver and obeying their requests is very very difficult indeed. In fact, it would be a lot simpler to just cut out the human driver entirely, as that BMW video shows (ie. hands off all the controls, car does the driving!)
Historically, the aerospace industry has show that the issues surround the cases when both the human and the computer are both thinking the other one is "flying the plane" (see AF447!).
Max_Torque said:
Lot's of stuff
Generally, in most modern systems, you effectively have a two layer control. You're correct in saying that, until the driver indicates their chosen trajectory, it's simply not possible to take full yaw priority away from the driver.It is possible, however, to control the yaw rate and reduce the severity of yaw gain giving the driver a longer time to respond. This also helps to reduce the severity of the system control as you're giving the car less yaw moment to overcome when the driver does indeed provide some countersteer.
In the latest generation of controllers, if you flick the car in to oversteer, rather than try to automatically correct to the car's original trajectory, it will arrest the yaw gain instead. Subjectively, it's like driving a car down a channel in the road that has rubberised walls that mean the car will only oversteer to a fixed point. In the worst case situation where a driver fails to add any corrective lock at all, the car just progressively scrubs speed off sideways.
To the experienced driver, the car still moves around under them (i.e. it's possible to lift off, get the tail to tuck in and maintain a degree of rear axle slip through the corner), but if they do add corrective lock, it will help them out by resorting to a more usual turning moment from the front inside wheel. By softening the response though, it's less abrupt and more natural.
The Skoda Yeti calibration for countersteer gave the driver a slight nudge in the correct direction in oversteer. It's enough to encourage you to steer in that direction, but not enough to feel disconcerting. If the BMW system was tuned to do something similar but then perhaps add further lock if the driver hadn't reacted enough, you could achieve the same yaw gain control but with no brake interventions at all.
RemarkLima said:
The other way to look at this, is the day when people do not need to do menial jobs, scanning items and taking money endlessly, stacking shelves, etc... is surely a great freedom? It'll cause some pain short term, but ultimately, who wants to do a dead end, meaningless job? At that point, your "human interactions" should ideally become more meaningful, as people have more time to read, paint and educate themselves... *ahem* or just watch Jeremy Kyle.
well, with ever increasing world population and more and more systems becoming automated, phone services, customer service, banking etc etc... You could argue that you require people to do the simple jobs 
unless we have a super large welfare state or huge job share RacerMike said:
Max_Torque said:
Lot's of stuff
Generally, in most modern systems, you effectively have a two layer control. You're correct in saying that, until the driver indicates their chosen trajectory, it's simply not possible to take full yaw priority away from the driver.It is possible, however, to control the yaw rate and reduce the severity of yaw gain giving the driver a longer time to respond. This also helps to reduce the severity of the system control as you're giving the car less yaw moment to overcome when the driver does indeed provide some countersteer.
In the latest generation of controllers, if you flick the car in to oversteer, rather than try to automatically correct to the car's original trajectory, it will arrest the yaw gain instead. Subjectively, it's like driving a car down a channel in the road that has rubberised walls that mean the car will only oversteer to a fixed point. In the worst case situation where a driver fails to add any corrective lock at all, the car just progressively scrubs speed off sideways.
To the experienced driver, the car still moves around under them (i.e. it's possible to lift off, get the tail to tuck in and maintain a degree of rear axle slip through the corner), but if they do add corrective lock, it will help them out by resorting to a more usual turning moment from the front inside wheel. By softening the response though, it's less abrupt and more natural.
The Skoda Yeti calibration for countersteer gave the driver a slight nudge in the correct direction in oversteer. It's enough to encourage you to steer in that direction, but not enough to feel disconcerting. If the BMW system was tuned to do something similar but then perhaps add further lock if the driver hadn't reacted enough, you could achieve the same yaw gain control but with no brake interventions at all.
The other issue with trying to control excessive yaw with a autonomous steering input, is taking it back off again quickly enough, especially if the driver has just shut the throttle and removed the tyre slip! And then the question becomes "do we stop the driver reducing torque that fast" etc etc A very difficult and long winded compromise is necessary as you say. For some drivers, who don't have a "smooth" driving style, they will say the system is holding back the car unnecesarily, and yet some will hardly notice the intervention! The latest systems are certainly way better at filtering the drivers behaviour and adapting too it as necessary.
Which brings me back to completely removing the driver from the control loop.......... ;-)
It is in effect a "talent" mode, but having a button marked "T" on the console is a no no.
So it will have to be hidden in the idrive system,
Which begs the question, will it include the following modes ?
F.D.Q (Full Dieter Quester)
F.H.S (Full Hans Stuck)
Although blaming the car for swapping paint with other traffic is pushing the boundary a bit.
So it will have to be hidden in the idrive system,
Which begs the question, will it include the following modes ?
F.D.Q (Full Dieter Quester)
F.H.S (Full Hans Stuck)
Although blaming the car for swapping paint with other traffic is pushing the boundary a bit.
Max_Torque said:
The fundamental issue with all this assisted driving tech is what do you use as the primary input for the cars trajectory? i.e.
In a purely mechanical car the driver uses their senses (sight, feel, balance etc) to work out where they want to go at any given moment, and apply handwheel and throttle/brake inputs to make the car go in that direction. For most people, most of the time, this is within the purely linear and positive region of control (ie. more handwheel = more turning etc). Once tyre slip at the non steered axle exceeds slip+turn at the steered end, the handwheel direction reverses, and the control becomes very non linear (as the inertia of the car starts to play a large part in the control lag). At this point, most drivers used to crash, but good ones could still operate in this highly non linear region (through experience, training and of course talent!) and continue to make the car follow their chosen trajectory.
Fast forward to the typical car with ESP, and the issue becomes one of deciding where the driver wants the car to go, and helping him go that way. One of the reasons most road ESP systems are thought by skilled drivers to "cut in too early" is because they must try to ascertain the intended vehicle direction before the controls become non linear. Systems that have a high degree of yaw authority and follow the handwheel angle religiously can cause unexpected events during oversteer! ie. the driver has left hand lock on, as they want to corner to the left, but the car oversteers halfway around the bend, so they now put on righthand lock. Should the system now try to turn the car right?
Anyone who's driven a fully "active" WRC car will have learn't to avoid excessive oversteer, because the system really could make the car suddenly reverse it's direction mid corner! As we currently have no other arbitor of intended direction than the handwheel, these systems become a compromise, because they have to work with drivers who's response ranges from "do nothing, just freeze" to "perfect application of opposite lock and drive torque". That is extremely difficult to calibrate for.
So newer systems, like the BMW system shown, now have the capability for full yaw authority as they can now steer themselves with the steering rack. But this means that unless the system somehow uncouples the handwheel from the rack, it cannot use the handwheel angle as it's trajectory request, it must take information from other sources (GPS, radar, laser scanners etc). For such a track test as shown in that video, that is quite easy, as it is not hard to pre-program a simple course through some cones under controlled conditions. However, in the real world, the task of integrating that system with a "real" driver and obeying their requests is very very difficult indeed. In fact, it would be a lot simpler to just cut out the human driver entirely, as that BMW video shows (ie. hands off all the controls, car does the driving!)
Historically, the aerospace industry has show that the issues surround the cases when both the human and the computer are both thinking the other one is "flying the plane" (see AF447!).
Good post.In a purely mechanical car the driver uses their senses (sight, feel, balance etc) to work out where they want to go at any given moment, and apply handwheel and throttle/brake inputs to make the car go in that direction. For most people, most of the time, this is within the purely linear and positive region of control (ie. more handwheel = more turning etc). Once tyre slip at the non steered axle exceeds slip+turn at the steered end, the handwheel direction reverses, and the control becomes very non linear (as the inertia of the car starts to play a large part in the control lag). At this point, most drivers used to crash, but good ones could still operate in this highly non linear region (through experience, training and of course talent!) and continue to make the car follow their chosen trajectory.
Fast forward to the typical car with ESP, and the issue becomes one of deciding where the driver wants the car to go, and helping him go that way. One of the reasons most road ESP systems are thought by skilled drivers to "cut in too early" is because they must try to ascertain the intended vehicle direction before the controls become non linear. Systems that have a high degree of yaw authority and follow the handwheel angle religiously can cause unexpected events during oversteer! ie. the driver has left hand lock on, as they want to corner to the left, but the car oversteers halfway around the bend, so they now put on righthand lock. Should the system now try to turn the car right?
Anyone who's driven a fully "active" WRC car will have learn't to avoid excessive oversteer, because the system really could make the car suddenly reverse it's direction mid corner! As we currently have no other arbitor of intended direction than the handwheel, these systems become a compromise, because they have to work with drivers who's response ranges from "do nothing, just freeze" to "perfect application of opposite lock and drive torque". That is extremely difficult to calibrate for.
So newer systems, like the BMW system shown, now have the capability for full yaw authority as they can now steer themselves with the steering rack. But this means that unless the system somehow uncouples the handwheel from the rack, it cannot use the handwheel angle as it's trajectory request, it must take information from other sources (GPS, radar, laser scanners etc). For such a track test as shown in that video, that is quite easy, as it is not hard to pre-program a simple course through some cones under controlled conditions. However, in the real world, the task of integrating that system with a "real" driver and obeying their requests is very very difficult indeed. In fact, it would be a lot simpler to just cut out the human driver entirely, as that BMW video shows (ie. hands off all the controls, car does the driving!)
Historically, the aerospace industry has show that the issues surround the cases when both the human and the computer are both thinking the other one is "flying the plane" (see AF447!).
These systems are great...for promotional purposes...put it in the real world and then lets have a look at how the sensors determine what the actual path should be...without a clear steering wheel angle to determine path.
Good luck with that.
Also...the autonomous oversteer mode is subject to the same grip limitations as a normal driver, so if the rear axle moves onto a lower grip surface then the slide may not be recoverable. Is that a good strategy for the public road?
Autonomous modes of transport will only ever really work when ALL road vehicles are autonomous. This has been said before but the transition from a semi autonomous road network to a fully autonomous one will be quick as EVERYONE will see the advantage of removing the (frankly s
tty) chore of commuting.
Within ten years I hope to be able to get into a large people carrier/camper (with beds, kitchen, toilet etc.) and have it drive myself and the family down to the south of France without me having to worry about negotiating traffic or staying awake. Instead I can just enjoy the view with my family.
tty) chore of commuting. Within ten years I hope to be able to get into a large people carrier/camper (with beds, kitchen, toilet etc.) and have it drive myself and the family down to the south of France without me having to worry about negotiating traffic or staying awake. Instead I can just enjoy the view with my family.
RemarkLima said:
The other way to look at this, is the day when people do not need to do menial jobs, scanning items and taking money endlessly, stacking shelves, etc... is surely a great freedom? It'll cause some pain short term, but ultimately, who wants to do a dead end, meaningless job? At that point, your "human interactions" should ideally become more meaningful, as people have more time to read, paint and educate themselves... *ahem* or just watch Jeremy Kyle.
Do you work for Google?We already have self driving cars in development that are far more advanced that this. Seems pretty simple compared to what Google, etc are doing. Also I saw BMW drive Clarkson round the Top Gear track SIX years ago using the same technology. They've programmed it to drift now but it's not really anything new. Maybe I'm being a bit critical, but that's the way I see it.
daytona111r said:
I think many people are being very naive here, thinking perfect I can put my feet up on the commute and let the car take over. This will not mean more leisure time, it will mean more time you can be productive ie work. Incidentally, I enjoy driving on the motorway and so do many others, even in something not designed to go anywhere near one.
Eventually driving a car will be like being chauffeured around. And the net result will be that cars will be designed more to chauffeuring it's passengers around than to the whims of the few keen drivers left. Self driving will eventually take over 95%+ of all driving. Sure you will still get track and track orientated cars, but why would a car manufacturer care about steering feel, throttle response, etc if it simply is no longer relevant?
But then if in 20 years time your car's speed is monitored / restricted on every piece of public road, self-driving cars would even make sense to someone like me, simply because there would be no enjoyment left in driving. And when there is no enjoyment left in driving, there is absolutely nothing sexy or exciting about it, that's when cars will become more like white goods than the emotional purchases they still are now.
Sad times ahead.
I completely agree. It's a sad state of affairs when the company that claims to build cars for the keen driver is reduced to making this kind of "ultimate self-driving machine". - I don't want electric steering based on what I've experienced so far, let alone this kind of junk.Eventually driving a car will be like being chauffeured around. And the net result will be that cars will be designed more to chauffeuring it's passengers around than to the whims of the few keen drivers left. Self driving will eventually take over 95%+ of all driving. Sure you will still get track and track orientated cars, but why would a car manufacturer care about steering feel, throttle response, etc if it simply is no longer relevant?
But then if in 20 years time your car's speed is monitored / restricted on every piece of public road, self-driving cars would even make sense to someone like me, simply because there would be no enjoyment left in driving. And when there is no enjoyment left in driving, there is absolutely nothing sexy or exciting about it, that's when cars will become more like white goods than the emotional purchases they still are now.
Sad times ahead.
If they were really as clever as they think they are, they'd make a modern version of the S54 engine - a high revving NA screamer - that can do 30 mpg without resorting to turbos or other experience-dulling compromises. That's something I would be interested in. I'm not against technology but surely it would be better if we used it to enhance the driving experience, not dumb it down and make a fleet of dull, soulless pretend sports / driver's cars like manufacturers seem keen to do at present?
I wouldn't mind "advances" if they didn't result in unnecessary compromises for the keen driver but anyone who thinks the future of the performance car should be about such nonsense as dull turbocharged engines being dressed-up with fake exhaust noises, or cars that drift themselves using computer-controlled electric steering should stick to the PlayStation.
Max_Torque said:
The fundamental issue with all this assisted driving tech is what do you use as the primary input for the cars trajectory?
The sat nav?Stick in your post code, sit back and crack open the stella, read a book, take a snooze. your own virtual chauffer. Execellent.
If you happen to notice the roads look interesting, better hope that you didn't have that stella and switch to manual.
For 90% of my driving time I'd be quite happy to sit back and not be concentrating on the driving. I do enough thinking at work. Currently I could drive to work quicker than the train, but - over 50% of the journey is in 5mph traffic, so I catch the train and get an extra 2 hours snoozing a day done.
Would prefer it if I could let the car drive and not have to listen to middle aged women talking about their friends, daughters , sister who bought a jumper in the sales at Debenhams for £5. Or what happened on TOWIE last night. WHYYYYYY!!!? I feel like offering them a toilet roll and some duct tape and say - you have a hole in the roof of your mouth and all the s**t in your head is leaking out. However a £30 pair of closed back Senheisser ear buds sorted that.
The only thing I have to put up with now is people reading the metro with their oversize elbows and women who wear 1/2 a litre of channel no5 shocking me from slumber. Thankfully not that often.
Gassing Station | General Gassing | Top of Page | What's New | My Stuff