Vehicle Dynamics and Control

Jan Åslund
[email protected]
Associate Professor

Dept. Electrical Engineering

Vehicular Systems
Linköping University
Sweden

Lecture 3

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 1 / 34

Longitudinal control: Cruise control

“RunSmart Predictive Cruise: How it Works

Unlike standard cruise control, where the truck tries to maintain a set
speed regardless of the terrain ahead, RunSmart Predictive Cruise looks up
to one mile ahead of the truck’s location and anticipates road grades by
using GPS and 3D digital map technology. The system adjusts the actual
speed of the truck for maximum fuel efficiency based on the terrain while
staying within 6 percent of the set speed.”
Press release from Freightliner Trucks, March 19 2009

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 2 / 34

 Scania Active Prediction - presentation December 2011
Longitudinal control: Cruise control

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 3 / 34

 controller (CC)

Cruise controller (CC): ∆fuel = −8.17 % Look−ahead controller (LC):

29.72 dm3/100km, 162.7 s ∆time = −1.78 % 27.29 dm3/100km, 159.8 s

Jan Åslund (Linköping University)

50

40

Altitude [m]
30
0
58.77° N, 17.01° E 500 1000 1500 2000 2500 3000 3500
58.75° N, 16.97° E

90

80
LC
70

Velocity [km/h]
CC
0 500 1000 1500 2000 2500 3000 3500

LC acc 1
LC brake
CC acc
CC brake 0.5
Level [−]
Longitudinal control: Cruise control

Vehicle Dynamics and Control

0
0 500 1000 1500 2000 2500 3000 3500

LC gear 1
CC gear
12 0.5
LC fuel use
11
Fuel use [dm3]

CC fuel use
Gear number [−]

0
0 500 1000 1500 2000 2500 3000 3500
Position [m]
Comparison between Look-ahead Controller (LC) and Conventional

Lecture 3
4 / 34
Longitudinal control

Some examples

CC Cruise Control
ACC Adaptive Cruise Control
CA Collision avoidance
ABS Anti-Blockier-System

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 5 / 34

ACC

Use radar or other senors to measure the distance to other vehicles.

Control brakes and acceleration
Three different modes
Cruise control
Keep distance to a vehicle in front of you
Brake to avoid collision

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 6 / 34

ACC

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 7 / 34

ACC: String stability
In a long caravan with ACC in all vehicles, string stability is important

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 8 / 34

ACC String stability

Consider a caravan where xi , i = 1, 2, . . . the positions of the vehicles

Define
δi = xi − xi−1 + Ldes
where Ldes is the desired distance.
A simple longitudinal model of the vehicle

ẍi = ui

where the acceleration ui is the control signal

Assume that the following control strategy is used

ui = −kp δi − kv δ̇i

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 9 / 34

ACC String stability

It is possible to show that the transfer function relating the spacing errors
of two consecutive vehicles is
δi (s) kv s + kp
G (s) = = 2
δi−1 (s) s + kv s + kp

The gain is s
kp2 + kv2 ω 2
|G (iω)| =
(kp − ω 2 )2 + kv2 ω 2
p
and it is straightforward to show that |G (iω)| > 1 if ω < 2kp . This
means that the amplitude of a low frequency oscillation increases when it
is transferred backwards in the caravan.
Source: Vehicle Dynamics and Control, Rajesh Rajamani

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 10 / 34

Longitudinal dynamic

From the previous lecture:

Ftot = ma

Equation of longitudinal motion:

dV
m = F − Rr − Rg − Ra
dt

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 11 / 34

Position as independent variable

In the cruise control application, the grade resistance Rg is a function of

position. In this and many other cases it is natural to use position as
independent variable.
The right-hand side of differential then becomes

dv dv dx dv m d(v 2 ) d(mv 2 /2)

m =m = mv = =
dt dx dt dx 2 dx dx
and we obtain
m d(v 2 )
= F − Rr − Rg − Ra
2 dx

Note that the previously introduced models for Rr and Ra are linear
functions of v 2 .

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 12 / 34

Kinetic energy

It follows that

d(mv 2 /2) = (F − Rr − Ra ) dx − mg dh

where

d(mv 2 /2):change of kinetic energy.

(F − Rr − Ra ) dx: work
mg dh = mg sin θs dx: change of potential energy

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 13 / 34

Stopping distance: First case
Given an initial speed V , the objective is to determine the stopping
distance S. From the previous slides:

m d(v 2 )
= −Fb − Rr − Rg − Ra
2 dx

Before the general case is analyzed, some special cases will be considered
First case: Neglect all forces except Fb . In this case we get:

m
d(v 2 ) = −Fb dx
2
Calculate the integrals from start to stop
Z 0 Z S
m
d(v 2 ) = − Fb dx
V2 2 0

Note how the intervals of the integrals were chosen!

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 14 / 34
Stopping distance: First case

The results is
mV 2
= Fb S
2
i.e.
Initial kinetic energy = Stopping distance × Brake force
and
mV 2
S=
2Fb

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 15 / 34

Stopping distance: Second case

The second case includes the grade resistance mg sin θs :

Z 0 Z S
m
d(v 2 ) = − (Fb + mg sin θs ) dx
V2 2 0

and the result is

mV 2
= Fb S + mg sin θs S
2
i.e.

Initial kinetic energy =Stopping distance × Brake force

+ Change in potential energy

and
mV 2 /2
S=
Fb + mg sin θs

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 16 / 34

Stopping distance: General case

m d(v 2 )
= −Fb − mg sin θs − fr mg cos θs − Cae v 2
2 dx
It is a separable differential equation
0 S
d(v 2 )
Z Z
m
=− dx
2 V2 Fb + mg sin θs + fr mg cos θs + Cae v 2 0

and
Fb + mg sin θs + fr mg cos θs + Cae V 2
 
m
S= log
2Cae Fb + mg sin θs + fr mg cos θs
Cae V 2
 
m
= log 1 +
2Cae Fb + mg sin θs + fr mg cos θs

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 17 / 34

Brush model

From Lecture 1

lc

lt

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 18 / 34

Brush model: Normal pressure

It was assumed that the normal pressure was constant in the contact
region.

According to Figure 1.15 a parabola shaped distribution seem more

reasonable, i.e.,
dFz
= Cx(lt − x)
dx

In the adhesion region:

dFx
= kt · i · x
dx
and
dFx dFz
< µp
dx dx

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 19 / 34

 Brush model: Normal pressure

Sketch the normal and longitudinal force distributions:

dFz dFx
dx dx

x x

The longitudinal force Fx is the area under the curve to the right.

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 20 / 34

Brush model: Sliding friction

Figure 1.16 shows the longitudinal force as a function of slip.

It can be seen that the force reaches a peak value and then decreases.

Assume that the sliding friction µs is lower than the friction µp in the
adhesive region, and that the normal force distribution is constant in the
contact region.

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 21 / 34

 Brush model: Sliding friction

Sketch the normal and longitudinal force distributions:

dFz dFx
dx dx

x x

The longitudinal force Fx is the area under the curve to the right.

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 22 / 34

Estimation of coefficient of friction µ

The coefficient of friction has been an important part when alakyzing the
braking and acceleration performance. Now, one example will be presented
on how to estimate the coefficient.
The approach is based on the approximation
Fx
= K (µ) · i
W
where it is assumed that the gradient K is a function of µ. If we first find
an estimate of K , then we can calculate µ.
To be able to calclulate K , we will first estimate
Longitudinal force Fx
Normal force W
Longitudinal slip i

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 23 / 34

Friction: Longitudinal force Fx

Assume that longitudinal acceleration is measured.

Longitudinal model
ma = Fx − Ra − Rr − Rg
The longitudinal force is now given by

Fx = m(a + g sin θs ) + Ra + Rr

where
m is estimated mass
a + g sin θs is measured by the accelerometer
Ra and Rr is calculated using empirical models

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 24 / 34

Friction: Normal force W

Normal force
l2 h
Wf = mg − (Ra + m(a + g sin θs ))
L L
l1 h
Wr = mg + (Ra + m(a + g sin θs ))
L L

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 25 / 34

Friction: Longitudinal slip i

Assume that the car is front-wheel driven.

Then there is no slip at the rear-wheel:
Vx
ir = 1 − =0
ωr rr
Modern cars have sensors measuring angular speed with high precision,
since this information is needed by the ABS-system.
The sensors at the rear wheels can therefore be used to calculate Vx and
then the slip at the front wheel can be calculated
Vx
if = 1 −
ωf rf
Now, the Fxf , Wf , and if at the front wheels are known and it possible to
estimate K and µ

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 26 / 34

Brake force distribution

From lecture 2

Kbf l2 + hµ
=
Kbr l1 − hµ
Kbf + Kbr = 1

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 27 / 34

Electronic Brake-force Distribution

In this case we get

1
Wf = (Wl2 + h(Fb + Fr ))
L
och
1
(Wl1 − h(Fb + Fr ))
Wr =
L
If we neglect aerodynamic resistance we get

Fb + Fr = m(a + g sin θ)

Assume that we measure the longitudinal acceleration. How can we use

this information to distribute the brake forces so that all wheels start
sliding at the same time, without knowing µ?

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 28 / 34

ABS: Introduction

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 29 / 34

ABS

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 30 / 34

ABS: A simple control strategy

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 31 / 34

ABS

When the wheels locks you loose

Brake force
Stability
Ability to control the vehicle

The objective of the ABS system is to prevent wheel lock-up

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 32 / 34

ABS: Detecting wheel lock-up

If the wheels do not slide, then

ω̇r ≈ a ≤ µg

This can be used to detect when the wheels are locking, either by
measuring the acceleration a or using an estimate of µ.

Another option is to use the skid

rω
is = 1 −
V
where V is estimated.

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 33 / 34

ABS

Jan Åslund (Linköping University) Vehicle Dynamics and Control Lecture 3 34 / 34