Safe Speed Forums

Purpose
Mitigate for unstable vehicle behaviour (oversteer/understeer)

Basic Function
Monitor actual vehicle behaviour through yaw rate and lateral acceleration, Calculate demanded vehicle behaviour from steer angle and vehicle speed (etc). Detect oversteer / understeer from comparison of actual and demanded vehicle behaviour. Brake individual wheel(s) appropriately: typically understeer inside rear wheel, oversteer outside front wheel (and a number of more complex strategies besides).

Hardware (typical)
Wheel speed sensors
Hydraulic Valve block + pump (integrated)
ECU
Master Cylinder PRessure Sensor
Yaw Rate & Lateral Acceleration Sensor
Steering Angle Sensor

Links
Other Sites
Rospa Description

For further info & how to comment on this post please refer to active vehicle systems

I'm interested in one of the design problems. Let's say we've got 10 degrees of yaw arisen due to oversteer. We detect the acceleration in the yaw axis and know that we have a problem to deal with. So we apply a brake and induce a restorative yaw in the opposite direction.

But how on earth do we know when we've done enough? We need to apply sufficient 'fiddle braking' to restore zero rate of yaw at yaw angle = 0. But how do we know when yaw angle = 0?

And a second problem - normal cornering provides an acceleration in yaw - how do we distinguish oversteer yaw from normal cornering yaw?

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

simplistically: the control target is yaw rate calculated from vehicle speed & steer angle (etc).. so you apply a brake moment to acheive the yaw rate the driver is demanding.

probably answers ur second question too.

Nope. That doesn't answer it at all. (but thanks!) We haven't yet distinguished between yaw from cornering from yaw from spin.

As you have written it we could be travelling in a straight line with spin to deliver a yaw rate equivalent to steering demand. Even the lateral acceleration could match for part of the spin as sideways motion provides a 'braking' force.

Perhaps the bit I haven't twigged involves a sensor that distinguishes axial travel - although I can't presently think of an inertial way to determine axial travel.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

I am assuming in your scenario, Paul, that you have already opposite locked it and you are effectively crabbing along the carriageway, and you want the system to correct this as you pay off the steering but to eliminate what might be a tail whip as it reaches stability. I think that should be possible shouldn't it just with accelerometers and reference to the steering wheel angle, with appropriate PID of course?

I can't see how we determine the end point of the correction (which is obviously required for PID calculations).

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

My understanding of the stability program on my car is that it works as follows:

1. Steering angle and vehicle speed inputs are used to compute an intended rate of steering.
2. The actual rate of change of heading is measured
3. differential braking inputs are applied until (2) matches the result of (1)

From what I've gleaned here and there...

* The sensor is basically a rotational accelerometer
* The system corrects oversteer by applying the outside front brake
* The system corrects understeer by applying the inside rear brake
* The system only operates under braking (ie it won't correct a powerslide)

I think one point of confusion here is that the system doesn't measure yaw angle, rather the rate of change of direction. In other words as long as the rate at which the car is turning the corner matches the amount of steering applied then it doesn't give two hoots whether the car is doing so with the back end 30 degrees out. That said, the two will approximate to the same.

"Yaw" to me is a measure of the difference between the direction the vehicle is pointing and the direction it is actually travelling - what a sailor would call "leeway". AIUI this is not what these systems measure or act upon.

_________________
CSCP Latin for beginners...
Ticketo ergo sum : I scam therefore I am!

it operates independantly of any driver brake input.
in the case of a powerslide (RWD?) first move would be to cut the throttle obviously!



yaw is the angle of the car relative to a fixed reference (not terribly helpful or easily measurable).
yaw rate is the rate of change of yaw angle (i.e. speed of rotation) and can be measured wth an inertial sensor.

what you describe above is sideslip or slip angle, difference between yaw angle & direction of travel (also difficult to measure directly).

when f(steer angle, speed) - measured yaw rate < threshold

?

I bloody hope not because that could be satisfied with a spin going on if the steering input was compatible with the rate of yaw in the spin.

We could end up with the electronics inducing a spin to match the steering input to the yaw rate!

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Yes, the TCS will divert and then reduce torque in the event of power induced wheelspin, but this is a separate system.

From what I've read, my understanding is that the stability system is only ever active during braking, for instance neither system would intervene if the front axle lost grip whilst powering round a corner.


For the purposes of the debate I would say that what I call "yaw" and what you call "slip" are effectively the same, as I can't see any point in measuring yaw against anything other than the direction of travel!

_________________
CSCP Latin for beginners...
Ticketo ergo sum : I scam therefore I am!

effectively understeer control ?

the relationship between steer angle & vehicle speed is only going to demand a 'reasonable' yaw rate... i.e. trying to model what the driver is expecting (which isnt usually a spin).

a poor algorithm or poor tune could certainly knock the vehicle from understeer to oversteer & vice versa.

usually part of the same system... but ESP will usually request throttle intervention to maintain stability.



ESP does not require any brake input from the driver.



frames of reference....
if you want to use a moving frame of reference that follows the direction of travel then yes.

My rather empirical feeling about this is that the ideal vehicle would linearly map yaw rate to steering input, in other words the driver dials in what rate of change of direction he wants and the vehicle responds.

If we assume that the steering geometry is capable of providing something close to this, then what will actually happen is that the vehicle will respond correctly up until the point where grip is exceeded.

What happens beyond this point depends upon which axle is the first to break traction, with the result being a loss of linearity in steering response - ie the yaw rate suddenly becomes greater than intended (oversteer), or smaller (understeer).

Electronic aids basically restore this linearity, so that (in theory at least) a massive steering input could indeed force the car to spin. But I don't think this will happen for two reasons:

Firstly the system will no doubt only provide a certain degree of correction, but more importantly there is no way a driver would ever apply the amount of steering lock required.

_________________
CSCP Latin for beginners...
Ticketo ergo sum : I scam therefore I am!

I think there's also a speed input so it know the car is moving forward along a curved path rather than going sideways whilst not moving forwards (the "doughnut"!). I guess it would also be possible to look at the rotational (average) speeds of the outside two tyres and the inside two from the ABS sensors. That and the road speed and steering angle and lateral acceleration inputs should ( I guess) be enough to work out how much of the yaw is due to travelling along a curved path and how much of it is due to getting out of line.

I don't KNOW that any of the above is true, but the car will definitely have all the necessary hardware for it!

I also have the following from one major manufacturer on the subject (rear wheel drive vehicle):

Electronic Stability Program (ESP)

ESP is a driving safety system that enhances the technology already present in the antilock brake system (ABS) and the acceleration slip regulation system (ASR) by using a series of additional sensors. These sensors play an integral role in calculating the vehicle’s movement in a skid. In addition to signals input from the ABS and ASR system sensors, the ESP control unit also processes information such as the steering angle, accelerator pedal position, engine torque and transmission gear ratio as well as the specified values for the yaw rate and lateral acceleration permitted in the given driving situation.

If the actual behavior of the vehicle deviates from the calculations of the safe "ideal line of travel", the system intervenes as follows: by applying the brakes on one or more wheels and/or reducing the engine torque.

The ESP system contains various subsystems that are all integrated in the ESP control unit.
ASR
Acceleration Slip
Regulation
EBD
Electronic Brake Force
Distribution
BAS
Brake Assist
ABS
Antilock Brake System
EBR
Engine Braking
Regulation
ESP

Examples of ESP in use:
1 The vehicle breaks away at the rear (oversteer); ESP brakes the
right front (outer) wheel.
2 ESP reduces the engine torque to help stabilize the vehicle.
3 The vehicle breaks away in the other direction; ESP brakes the
left front (outer) wheel.
4 The vehicle is now stabilized.

1 During emergency braking, the steering wheel is wrenched from
the driver's hands: the vehicle starts to become unstable.
2 The vehicle is understeered, and ESP brakes the left rear wheel;
the vehicle can now follow the steering input.
3 Quick countermovement of the steering wheel causes the rear of
the vehicle to break away; ESP brakes the left front (outer) wheel.
4 The vehicle is now stabilized."

But instead of that try thinking from a point of oversteer back to axial travel under ESP. How is the ESP going to know that we have arrived back at axial travel (which may be on a curved path).

With all the yaw accelerometers I can imagine we can meet the curved path yaw demanded with a spin in yaw. On what I know, it just doesn't work. I'm guessing I'm missing an input - something that can confirm axial travel irrespective of a curve in the path. But what sort of sensor can do that?

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

please define axial...!

I may be wrong, but my rather simplistic understanding goes something like this:

Imagine your car parked in the middle of a vast airfield. Now get in and turn the steering wheel 1/4 turn, then let the brakes off and push the car gently. It will describe a large circle until (theoretically at least) it lands back where we started.

So we can see that any given degree of steering lock has a corresponding turning circle assuming no sideways slip of any wheel. That's one variable.

The only other variable we need is speed. Now assume we fix our steering lock and drive (with no slip) at velocity v such that we describe one complete circle in one minute. Clearly our vehicle also rotates at the rate of 1rpm about it's vertical axis.

So for our given steering lock if we are travelling at speed v our computer should compute a yaw rate of 1rpm. If our accelerometer measures a rotation of >1rpm then the car is turning "too fast". This means that the car cannot be following it's ideal path, it is trying to follow a tighter one. Similarly, if the car yaws at a rate of <1rpm then it must be running wide of the correct arc. In either case the computer can obviously apply a correction via independent braking to match the actual rotation rate of the car with that which it ought to be achieving, given the current speed and steering input.

It may still be running wide of the arc, but if it is it does so in a stable state, ie all 4 wheels sliding sideways at an equal rate. This is reasonable, as the driver will be intuitively led to steer tighter to try and regain his intended path, thus generating an understeer or oversteer situation that the system can then deal with.

_________________
CSCP Latin for beginners...
Ticketo ergo sum : I scam therefore I am!

I'm thinking 'axial' when the car's long axis is aligned with the direction of travel. i.e. little or no yaw relative to intended path. (But the possibility remains of considerable yaw from cornering.)

And I think I've solved it - we need front and rear lateral accelerometers. Then we should be able to tell the difference.

Are twin (lateral / yaw) accelerometers fitted to ESP equipped vehicles?

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

to my knowledge only single yaw rate & lat acc.

although 'front' lat acc is a combination of vehicle lat acc & yaw acceleration.