AWD without differentials?

Mikebergy

1 W
Joined
Nov 15, 2015
Messages
61
Location
San Luis Obispo, CA, USA 93401
Curious if anyone knows of any differential-less AWD or even 2WD drivetrains? I'm curious if anyone has used "torque-vectoring" or whatever the electronic control technology may be to control the L-R variation while turning, loss of traction,etc without the use of a mechanical differential. I've read a couple very technical articles but nothing that I have seen industrialized.

Moderators: I just noticed I started this in the wrong forum category, can someone move it for me? Probably into EV general discussion
 
I read recently, although I can't remember where, that someone found that there was no particular problem with not being able to use torque vectoring on an AWD EV. I suspect it might have been about a formula student race car whose team had run out of time/funds/capability to program it in. The car still ran without handling weirdly.
 
No differentials works great. It's a nice way to save wasting a bunch of power at cruise from churning oil.
 
liveforphysics said:
No differentials works great. It's a nice way to save wasting a bunch of power at cruise from churning oil.

Hub motor on all four wheels. Electric motors aren't driving at constant speed, they are driving at constant torque (OK really a torque curve but that's beside the point. would be about constant through a short curve). Turn, and both motors are still applying torque, even if one is going slower. The only use of a differential is to drive more than one wheel with one motor, but with electrics, it would be better to drive each independantly.

I'm using 2WD on a tadpole trike, the principle works there as well.
 
Inboard motors direct drive with a CV shaft is also ideal in applications where low unsprung wheel mass is considered critical.
 
Yeah, hub motors would not be my choice. I don't think applying the same torque on both inside and outside wheels during cornering would be ideal, since at some point the inside will begin to spin, but as long as the wheels don't get chewed up when cornering that should be alright.

Llile, no issues when cornering your trike hard? I don't imagine you can corner well with that thing anyway.
 
Mike, read what llile said again..
Motors drive at constant torque, so a slight speed change between inside and outside wheels wont cause any slip.
The same thing accurs with ICE motors driving separate wheels, or even multiple motors driving the same wheel .!
 
Yes, I understood. What I was curious about was how to deal with the weight transfer and the inevitable inside wheelslip. As one corners, the weight transfers to the outside wheel, and then you want to distribute the torque unevenly left to right, biasing heavy on the outside where the grip is higher. With one throttle, a way to control that torque distribution is needed. That is what limited slip diff's do; modern AWD systems do the same thing, but both systems have traditionally been mechanical. I guess my first concern, you guys have helped me with- given identical throttle, the motors will function like an open diff up to the point of slip, I'd just need to set the steering and suspension up correctly to account for Ackerman. But it looks like torque vectoring/electronic differential has not yet been utilized anywhere outside of the lab experiments I read about.
 
llile said:
The only use of a differential is to drive more than one wheel with one motor, but with electrics, it would be better to drive each independantly.
I don't think think this is universally true, as a well-designed mechanical differential would not really be any worse than the set of reduction gears that would be needed if a diff was not used.
 
But it looks like torque vectoring/electronic differential has not yet been utilized anywhere outside of the lab experiments I read about.
Electronic (torque vectoring) differentials have been used in competition cars ..race and rally etc,..for several years, and i believe several road cars (Mitsubishi evo, Nissan GTR, etc) are also using similar electronic systems.
The Mercedes SLS EV has full electronic 4 wheel torque vectoring....
http://www.caranddriver.com/news/2014-mercedes-benz-sls-amg-electric-drive-photos-and-info-news
 
Torque controls is a great way to go, with anti-slip detection and torque backoff.

It is easy enough to reduce the torque on the inside wheels if that is desired. For most situations it is probably not needed, but for extreme turning at high drive torque it can reduce slip. But if anti-slip techniques are employed then the torque will be reduced automatically when the wheel slips, so why prematurely reduce torque? Do whichever or all of the above.
 
Hillhater said:
But it looks like torque vectoring/electronic differential has not yet been utilized anywhere outside of the lab experiments I read about.
Electronic (torque vectoring) differentials have been used in competition cars ..race and rally etc,..for several years, and i believe several road cars (Mitsubishi evo, Nissan GTR, etc) are also using similar electronic systems.
The Mercedes SLS EV has full electronic 4 wheel torque vectoring....
http://www.caranddriver.com/news/2014-mercedes-benz-sls-amg-electric-drive-photos-and-info-news
All the examples you mentioned (except for the AMG SLS) all use electronically controlled mechanical systems. That isn't what I am looking for. I am curious about the system used in the AMG, as it looks interesting, thanks.
 
Alan B said:
Torque controls is a great way to go, with anti-slip detection and torque backoff.

It is easy enough to reduce the torque on the inside wheels if that is desired. For most situations it is probably not needed, but for extreme turning at high drive torque it can reduce slip. But if anti-slip techniques are employed then the torque will be reduced automatically when the wheel slips, so why prematurely reduce torque? Do whichever or all of the above.
It is not just a reduction of torque, it is a redistribution of the torque that requested through the throttle input. If say you request x-amount of torque, then the electronics would vary the throttle input to the drive wheels based on the situation, not just reduce the throttle to one of the wheels, but reduce one and increase the other. Maybe it is easy, but I am curious about it from a homebuilt perspective, and I've not yet come across anything that I could personally use if I wanted to design and build a small vehicle with that feature. From the papers I read, it is not easy to implement.
 
Doesn't rimac use the torque vectoring your describing?
https://youtu.be/bD2Do1gAuog
 
Mikebergy said:
Alan B said:
Torque controls is a great way to go, with anti-slip detection and torque backoff.

It is easy enough to reduce the torque on the inside wheels if that is desired. For most situations it is probably not needed, but for extreme turning at high drive torque it can reduce slip. But if anti-slip techniques are employed then the torque will be reduced automatically when the wheel slips, so why prematurely reduce torque? Do whichever or all of the above.
It is not just a reduction of torque, it is a redistribution of the torque that requested through the throttle input. If say you request x-amount of torque, then the electronics would vary the throttle input to the drive wheels based on the situation, not just reduce the throttle to one of the wheels, but reduce one and increase the other. Maybe it is easy, but I am curious about it from a homebuilt perspective, and I've not yet come across anything that I could personally use if I wanted to design and build a small vehicle with that feature. From the papers I read, it is not easy to implement.

Unless your vehicle is pushing traction limits, you don't need to bother with torque vectoring.

If you are using torque type controllers (most sinusoidal controllers are torque controlled) then it is a fairly simple matter to make a box that reduces the signal to one and increases it to the other. The harder part is to decide when to do that. A steering angle sensor, or accelerometer could be used.

It is still probably simpler and safer to do anti-slip torque reduction detection on each wheel. This would reduce torque to the inside wheel automatically just as much as needed. Then if the driver wants more torque in the corner they can increase throttle if they want it. Otherwise it would have the unsettling effect of trying to cause the outside wheel to slip when the inside wheel starts slipping if it increased torque on the outside wheel. That might not be a good automatic response to slipping. The driver is already selecting the max torque for whatever situation, so increasing it could be bad. But you can make whatever algorithm you choose once you have the infrastructure to support it.

If you use PWM type controls then the whole problem gets much harder. They fight the slowing of the inside wheel with nonlinear increase in torque and essentially fight the turn. If you start with torque controllers it is much easier, since they will already compensate for the slowing of the inside wheel, so you don't have to do that too in your box.

From a homebuilt perspective you can use either analog or digital. Digital gives more choices and allows more algorithmic possibilities. A small micro such as an Arduino with an ADC and a pair of DACs can read the throttle and output throttle signals to two controllers. Or you can use the internal ADC and PWM with filters to make it cheaper but less precise and lower bandwidth. A pot on the steering angle can give you data there, or an accelerometer. Monitoring pulses from the motors can give you wheel speed, or magnets on the wheels can be used. Multiple magnets might be needed to detect wheel slip quickly enough. Motor pulses are more frequent and easier if you have hall sensors you can use those. If you pick up all three hall sensors you will get faster detection of slip.

There are lots of little micro boards these days that are not expensive or hard to program. Arduino Nano, Teensy, Feather, Pro Trinkets, Pro Mini, many others. Some of them have ARM chips for higher power, but are still easier to do real I/O than a graphics oriented Raspberry Pi.

Then you can put whatever algorithm into the throttle system you want. Make sure you have an emergency stop in case the throttle program gets stuck. Program carefully, limit power during testing to gain confidence.
 
Like Alan said, unless you're dealing with big traction issues torque vectoring is adding complexity for nothing.
 
In the same way suspension does more than simply soak up road bumps,..TV is more than just traction control.
Torque vectoring is also used as a handling (cornering control) feature at high speeds to adjust turn in , slip angles , etc.
But unless you are in that high performance /competition league,..it would be hard to justify other than as a development project.
 
Thanks for the replies. Alan gives some some really interesting ideas about the hardware, and I do think that it is an added complexity, but then so are a lot of systems going into vehicles. Nobody needs auto drive or back up cameras or battery cooling systems, or etc... But maybe a system like this could actively distribute the torque between the motors in such a way that the system as a whole was more efficient, wasting less energy over the course of a day of driving. Who knows. If it is easier to drive the wheels of a vehicle independently, then why did Tesla decide to use a mechanical differential? Just interesting topics for discission, thanks for participating. Are there any topics out there that discus the use of micro PCs (like Rasp Pi) to control inputs to motor controllers and deliver information to the driver in a novel way?
 
First, I wanna say that I think the phrase "torque vectoring" is a very funny phrase and it doesn't really make sense. Torque is already a vector, so you could say that every electric motor already has "torque vectoring". In fact, I would argue that the phrase "torque scalaring" is actually more accurate although fairly obscure about its meaning. I prefer to call this sort of thing "torque balancing", because that is what you're actually doing: controlling the balance of total torque to the two wheels. You're not multiplying the torque by some vector (which would change its axis, certainly not really possible with a real shaft or wheel), it's actually just a scalar.

I think the phrase came about from "thrust vectoring" on fighter aircraft where you actually are changing the force's axis! That, and it sounds pretty cool if you don't really know what a vector is.

As far as its implementation with differentials goes, it seems like nearly all systems actually just use each wheel's brake to reduce torque on the desired side, possibly with a calibrated increase in throttle to remain with the same overall torque (although this seems unlikely). Remember that an open differential splits torque evenly, so to have less torque reach the wheel on one side, just use the brake on that side. The other side still gets the same torque as before.

With electric motors it's just a bit easier to translate control effort into a reaction in the system because the electric motors let you control torque on each wheel directly (current control for synchronous motors) instead of having to guess how much braking effort turns into torque.

Now, why did tesla only use 1 or 2 motors in their cars? Well, both cost and efficiency comes to mind actually. When you double the number of motors, you need double the number of motor controllers, motors, wires, reductions, etc, but lose the differential. All four wheels still need brakes so you can use the same stability control system as ICE cars. It helps if you don't need to entirely develop a new control algorithm and test and verify it all.
Now I can't comment on what actually turns out to be more efficient, but I suspect that having 1/2 motors and a differential is more efficient and cost effective. Tesla is all about efficiency so I can only assume that is the case since that's what they chose.
 
The system changes the magnitude of the torque across the axle in order to change the direction of the vehicle. Sounds reasonable enough to describe it as vectoring?
 
I agree with alexaus,..the term is technically incorrect as its implies that the torque is being "vectored"...which it is not.
However, i guess its one of those "every day language" accepted phrases now so no point in fighting the flow.
It would just as easily be called "torque steer".. just as the same , but uncontrolled, effect is called when applied to front wheel drive vehicles.
 
Mikebergy said:
Yeah, hub motors would not be my choice. I don't think applying the same torque on both inside and outside wheels during cornering would be ideal, since at some point the inside will begin to spin, but as long as the wheels don't get chewed up when cornering that should be alright.

Llile, no issues when cornering your trike hard? I don't imagine you can corner well with that thing anyway.

You kidding? This bike corners like a ferrarri. This is the best maneuvreing bike I've built. It's got a wide wheelbase (as I said, tadpole trike with motors in the front wheels) and in tests I could handle a curve at speeds I wouldn't try in a car. Really, the wheels don't go at much diffreent speed around a curve unless you are turning a really tight curve, which you'd only do when, say, parking.
 
Back
Top