Design and Vehicle Implementation of an Adaptive ABS

1Challenge the future

Design and Vehicle Implementation of an Adaptive ABSJ. Tigelaar 13 May 2011

1Design and Vehicle Implementation of an Adaptive ABS


Design and Vehicle Implementation of an Adaptive ABS


Goal

• Research on ABS is part of a larger research objective

• Load based vehicle dynamics control• Decrease in system complexity• Decrease in development cost• Increase in performance

Novel ABS algorithm evaluated on testbench.Limitations: no changing load on tyre

no changing friction coefficient

…Taking the research one step further…

Goal: Evaluate ABS robustness (Fz, µ) and implement in test vehicle



Contents

• Why do we need ABS?• How does it work?

• PART 1 -> Algorithm Design: increase its robustness• Simulation evaluation

• PART 2 -> Vehicle Implementation: implement in test vehicle• Track testing

• Does Load Based Sensing offer improvements in performance?

The next ~25 min



Why does one need ABS?ABS video

Introduction to ABS

This video contains shocking

images and is therefore not suitable for BMW fans.


Why does one need ABS?ABS video

Introduction to ABS

This video contains shocking

images and is therefore not suitable for BMW fans.

To prevent wheel-lock

What happens at wheel-lock?


BrakingForces are generated at tyre-road contact patch

• Many different models of tyre-road interaction

Introduction to ABS

[Jazar, Reza N. (2008): Vehicle dynamics. Theory and application. New York, NY: Springer.]


Braking (Fx)Longitudinal Tyre Forces

( )x zF F

1x

x x

v r rv v

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

1.2

slip

fric

coef

f

()

dry asphaltwet asphaltdry cobblestonesnow

Introduction to ABS

max Fx -> BD

F x/F

z

Slip λ


Braking (Fy)Lateral Tyre Forces

( , )F Fy z

Steerability -> front

Stability -> rear

Introduction to ABS

[Tanelli, M.; Corno, M.; Boniolo, I.; Savaresi, S.M. (2009): Active braking control of two-wheeled vehicles on curves.]

Slip λ

F Y/F

Z


Re-capTrade-off between Fx and Fy

Many different control strategies

to achieve this

Ideal operating range

Prevent wheel-lock in order to:• Maintain

steerability• Maintain stability• Decrease braking

distance

Introduction to ABS

Slip λ

Norm

alize

d fo

rce


How does ABS work?Sensor

Control

Actuator

How does ABS cycle?

Hold/Decrease/Increase pressure

ValvesWheel speed sensor

Control

5 Phase : hybrid wheel deceleration based logic

Introduction to ABS


A 2-Phase exampleA hybrid wheel deceleration based logic

Continuous and discrete states

Algorithm Design

Slip λ

Norm

alize

d fo

rce


ABS 5-Phase algorithm (x1 and x2)A hybrid wheel deceleration based logic

*1x *

2 xx r a

* *1 2 1

1 ( ( ) )xx

x x x av

* *2 1 2 1'( )( ( ) )x

x

bx x x x a uv

2

zRb FJ

Ru TJ

Set of dynamic equations

Algorithm Design

X1 : Wheel slip offsetX2 : Wheel acceleration offset


5-Phase algorithmA hybrid wheel deceleration based logic

Stating that

• Brake torque will either be kept constant or change rapidly

• Switching between torques is triggered by thresholds

• Thresholds are wheel deceleration based

Regulation logic chosen as such to keep unmeasured x1 small

close to 0

Algorithm Design


5-Phase Automaton

*1x

*2 xx r a

The hybrid automatonState transistions governed by guard conditions.Evolution of continuous states is determined by dynamic system.

Algorithm Design

[Pasillas-Lépine, W. (2006): Hybrid modeling and limit cycle analysis for a class of five-phase anti-lock brake algorithms.]


5-Phase 1st Integrals

For constant brake torque

For large torque variations,Approximate first integral is

Approximation error with

2

1 2 1( )ZRI x F xJ

* * * 22 1 2 1

0

1ln(1 ) ( ( ) )2 xI x x x au

Phase-plane (x1-x2) trajectories

22 * * 2 *

2 1 1( ( ) ) ( ( ) )x z xRerror x x a F x aJ

0

1u

Algorithm Design


5-Phase CriteriaThe criteria

2

zRb FJ

Satisfying all 5criteria

Algorithm Design

Stability guarantee

d

Criterion 1 :

Criterion 2 :

Criterion 3 :

Criterion 4 :

Criterion 5 :


Load transferStatic and Dynamic stati

cdynami

c

1

22 cogZ x

hlF mg maL L

,

2,11 ( )2F R

cogZ x

hlF mg ma

L L

2

12 cogZ x

hlF mg maL L

Algorithm Design

[Jazar, Reza N. (2008): Vehicle dynamics. Theory and application. New York, NY: Springer.]


Fz and StabilityThe fifth criteria

• [B] Maintain stability at lower loads -> retune (ε5) thresholds (limited)

• [C] Maintain stability at lower µ’s -> T rate increase in phase 5

,

2,11 ( )2F R

cogZ x

hlF mg ma

L L

static

dynamic

Z

ZF

cog

F Lmgh

Algorithm Design

Friction coefficient µ

Load

tran

sfer

[N]


5-Phase SimulationSimulation results

• 4 wheel car model. Algorithm runs separately on each wheel.

• Straight line braking starting from 150 km/h

• Cost functions are defined as:• Maximum slip

• Braking distance

max( )J

1

0

( ) ( )t

tJ s v t dt

Fy

Fx

Algorithm Design


5-Phase Simulation at high µSimulation results (FL wheel)

• Observed that as µ increases, the algorithm cycles within stable zone.

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.40

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000Long tire force FL vs Slip

Slip

Long

tyre

forc

e [N

]

= 0.4 = 0.65 = 0.9

Algorithm Design

Larger force dropoccurs duringcycling

Long

itudi

nal T

yre

Forc

e [N

]

Slip λ


-40 -35 -30 -25 -20 -15 -10 -5 0 5 10 15-120

-100

-80

-60

-40

-20

0

20

40

60

80

Slip offset x1 [%]

Whe

el a

ccel

erat

ion

offs

et x 2 [m

/s2 ]

ABS Phase-plane evolution (1st initial cycle 1:1900)

Phase 2Phase 3

Phase 1

Phase 2

Phase 5

Phase 2

Phase 4

ABSInitialization

Phase 3

5

4

3

1

2

5-Phase Simulation Limit CycleSimulation results

• 5th thresholddeterminesthe slip levelreached in theunstable zone

Algorithm Design

Wheel slip offset x1 [%]

Whe

el a

ccel

erat

ion

offse

t [m

/s2 ]


5-Phase Simulation at low µSimulation results

• An increased ε5 would be beneficial for higher µ, but adverse for low µ

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.40

500

1000

1500

2000

2500

3000

3500Longitudinal tire forces FL RR vs Slip

slip

LF Fx

LF Fx

RR Fx

RR Fx

A dynamic threshold

ε5 = 50

ε5 = 50

ε5 = 30

ε5 = 30

Algorithm Design

Slip λ

Long

itudi

nal T

yre

Forc

e [N

]


5-Phase Dynamic ThresholdDynamic fifth threshold

• Fifth threshold can be freely defined

• Friction coefficient cannot be measured directly

5, 5 28.527 max 14.218xDynamic

z

FF

( ) X

Z

FF

Algorithm Design

FX/FZ

ε 5


Adaptive 5-Phase ResultsResults - improvements

0.3 0.4 0.5 0.6 0.7 0.8 0.9 160

80

100

120

140

160

180

200

220

240

260

Friction coefficient road

Bra

king

dis

tanc

e [m

]

Braking distance for different threshold e5

e5=30e5=50e5=e5dynTheoretical

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0.05

0.1

0.15

0.2


Slip

Maximum slip for different threshold e5

e5=30e5=50e5=de5

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0.05

0.1

0.15

0.2


Slip

e5=30e5=50e5=de5

• 20 % decrease in braking distance• Maximum slip level is maintained

front

rear

ε5=30

ε5=50

Theoretical

ε5=d ε5

Algorithm Design

Friction coefficient



Brak

ing

dist

ance

[m]


Adaptive 5-Phase VideoResults - improvements

• VIDEOS – steering manoeuvre (2m to the left)

• White car : No ABS• Red car : ABS with standard 5-Phase• Blue car : ABS with adaptive 5-Phase

Lane change during braking from 150km/h

Algorithm Design


Vehicle modifications (VM)BMW 530 (E60)

Vehicle Implementation


VM PrincipleBasic principle

• Original ABS serves as a benchmark, driver can switch

• ECU (sensor input and required control decisions)• Power stage (amplify the control signal)• Hydraulic unit (actuate the solenoid valves)

ABS module

DummyABS

module


How?

Individual wheel control


VM dSpaceBasic principle

dSpace Autobox enables receiving and sending signalsIN: wheel speed (tapped)

brake pressure (installed)OUT: control action valves

control action pump



VM Brake circuitBrake pressure sensors

Intake Valve (#6) Exhaust Valve (#7) Modeopen closed Pressure build-up

closed closed Hold pressureclosed open Pressure decrease

6

7

4

8

12

3 1 Brake pedal 2 Master cylinder 3 Brake booster4 Brake calliper 5 Return pump6 Intake valve 7 Exhaust valve 8 Pressure sensor

5


[Robert Bosch GmbH (2007): Automotive Electrics Automotive Electronics. 5th: Wiley.]


VM T-splitBrake pressure sensors



VM dSpaceBasic principle

dSpace Autobox enables receiving and sending signalsIN: wheel speed (tapped)

brake pressure (installed)OUT: control action for valves (hold and decrease

presssure)control action for pump (increase pressure)



VM ECUThe ABS Module

The ECU

Still beingreverse engineered.



VM ConventionalSystem overview - conventional



VM NovelSystem overview - novel



VM Track testing



ConclusionsAn overview

PART 1• Through the use of load sensing:

Dynamic thresholds can significantly improve ABS

performanceDecreased braking distance by 20%Maintained lateral stability and steerability

PART 2• Vehicle modification allow performance evaluation

Conclusion


?Conclusion


BrakingThe braking process


Is ABS improvement worthwhile?


Further researchTU Delft

• ABS Activation Logic• Different than slip based (e.g. brake pedal)

• Coupling of all 4 wheels• Synchronization on front• A-Synchronization on rear

• Modeling of brake efficiency• Thermal effects

Extra Slides


Future of ABSPossible outcomes

• From motion based to force based control systems

• Electro-Mechanical Brakes• Force modulation is continous (not discrete)• No vibrations to brake pedal (safety)• No toxic brake fluid (leakage)

• From threshold based control rules (wheel deceleration)

move to slip control (multi-applications, ABS, ESP, TCS, …)

• ABS simplificationExtra Slides


ABS Synchronization


0 0.05 0.1 0.15 0.2 0.25 0.3 0.350

10

20

30

40

50

60

70

Time interval occurence

Time steps

Occ

uran

ce

occurencemean time interval

Controller Area Network (CAN)Tests show a deviation in time interval reception, thus large jumps in value

Average time interval over all 4 wheel signals = 60 msec

Whe

el sp

eed

[km

/h]

Occu

rrenc

e

Time [s]

Time steps


Wheel speedActive Inductance Sensors

• Sense change in magnetic field due to incremental ring

• Contains 3 Hall sensors• 1 and 3 serve for velocity estimation• 2 serves for direction of rotation

• Derive wheel deceleration is challenging. Two main methods• Lines-Per-Period• Fixed-position

Extra Slides


Bosch ABS

• Based on:

Heuristics•Requires extensive tuning per vehicle model

Rule-of-Thumb• System is unclear -> contains many control rules

System complexity ever increasing

• Also uses (amongst others) wheel deceleration values


VM PWMBasic principle

• Pulse-Width Modulatedsignal for valve hold controland pump increase control

• Additional I/O signalfor valve release control

50% duty cycle

Documents

Design and Vehicle Implementation of an Adaptive ABS