Download - Canudas de Wit Innsbruck
-
7/27/2019 Canudas de Wit Innsbruck
1/73
Control Design for
X-By-Wire Components
C. Canudas-de-Wit
LaboratoiredAutomatiquede Grenoble, FRANCE
CNRS-INRIA Projet
-
7/27/2019 Canudas de Wit Innsbruck
2/73
P / 2
Class of Systems
Air-scooters
BalancingScooters
Systems witha Driverin the
loopDrive-by-Wire
Systems
-
7/27/2019 Canudas de Wit Innsbruck
3/73
P / 3
A Motivating Example
Stabilization of
not
not explicitly
regulated
Balance Scooter is
similar to a cart-
pendulum
Differences:
-
7/27/2019 Canudas de Wit Innsbruck
4/73
P / 4
How motion is produced ?
Forward motion is induced as a
reaction of the feedback system to
changes in the tilt angle arising from
forces produced by the rider
-
7/27/2019 Canudas de Wit Innsbruck
5/73
P / 5
Control Paradigm
Seville home madeBalancing scooter
CommercialSegway scooter
How to design controllers resulting in:
Comfortable,
Safe,
Pleasant riding
-
7/27/2019 Canudas de Wit Innsbruck
6/73
Outline
1. IntroductionChallenges
Product EvolutionBreakthroughs
2. ExamplesClutch SynchronizationSteer-by-Wire
3. Conclusions
-
7/27/2019 Canudas de Wit Innsbruck
7/73
Outline
1. IntroductionChallenges
Product EvolutionBreakthroughs
2. ExamplesClutch SynchronizationSteer-by-Wire
3. Conclusions
-
7/27/2019 Canudas de Wit Innsbruck
8/73
P / 8
Conceptual Challenges
Different driver skills:
Different levels of: perceptual abilities, physical skills, and
technological understanding,
Decision partition between Driver & Automated system:
Little control: Increases workload
Decreases drivers awareness,comfort & safety
Too much control Confidence excess: false safety perception
A notion of driving pleasure, whose primary
goal is to safely promote comfort
(Goodrich & Boer 01)
Fun-to-Drive:
-
7/27/2019 Canudas de Wit Innsbruck
9/73
P / 9
Technical Challenges
Lack of well defined metrics
How to measure comfort ?,
How to define safety ?
Transfer of standard control notions is not always
straightforward
How to integrate in the control specs the comfort metrics (ifany) ?
Are all stable systems safe ?
How to select/define a few tuning parameters with
subjective meaning (pleasant, hard, nice, tiring, etc.) ?
u
+-
Mechanical
Subsystem
Inner
feedback
loop
Driver
y
YH
-
7/27/2019 Canudas de Wit Innsbruck
10/73
P / 10
Problem
Main difficulty:
How to translate subjective vehicle specs into control
specs?
Subjective
Vehicle Specs:
Maneuverability,
Perceived quality,
& security,
Brio,
Etc.
Control Specs:
Precision,
Damping,
Stability,
Passivity,
Etc.
-
7/27/2019 Canudas de Wit Innsbruck
11/73
P / 11
Examples of Vehicle Control Problems
Implying Jointly Comfort & Safety Specs
Automatic ClutchSmooth synchronization (C),
Engine stall (S).
Steer-by-wire & Electric Power Steering SystemsSteering wheel vibrations, Tire/road sensations (C),
Driver interaction with an active system (S).
Chassis ControlReject road vibrations (C),
Vehicle stability (S).
Adaptive Cruise Control (ACC)Smooth distance/velocity regulation (C),
Vehicle front collisions (S).
Etc
-
7/27/2019 Canudas de Wit Innsbruck
12/73
P / 12
Car Product Design Evolution
ESP, ABS, TC,Airbags, etc.
By base functions:Suspension,
Lighting, etc.
By field of expertise:Acoustics,
Passive safety, etc
By customer-perceived performance:Drivability,
Quality, etc.
Distinctive featuresFun-to-Drive
(something more)
Basic featuresPerceived as a due
T d ff d t th I t d ti
-
7/27/2019 Canudas de Wit Innsbruck
13/73
P / 13
Trade-offs due to the Introduction
of Distinctive Features
Solutions to increase torque clutch capacity:
Higher pre-constraint forces
Larger friction plates
Multi clutch plates
Example: High torque diesel engines
Clutch torque = 1.2 Peak Engine Torque
0
100
200
300
400
500
600
700
Torque [Nm]
Laguna 2.2 D(1994)
Laguna 2.2 DT
(1996)Laguna 2.2 dCi(2002)
Laguna 2.0 dCi(2005)
Ferrari Enzo
T d ff d t th i t d ti
-
7/27/2019 Canudas de Wit Innsbruck
14/73
P / 14
Trade-offs due to the introduction
of Distinctive Features
Solution 1: Higher pre-constraint spring force Fn
Constraints:
Driving ergonomy(for passenger comfort)
Passenger space
(for passenger comfort)
Crash test foot injuries
Jv
kt
t
Je
e c
Jg
Side effects:
Increase of pedal effort
Lengthening of pedal
travel
Fn
T d ff d t th i t d ti
-
7/27/2019 Canudas de Wit Innsbruck
15/73
P / 15
Trade-offs due to the introduction
of Distinctive Features
Solution 2: Larger plate disks
Constraints:
Constraint on power-trainpackaging
Under hood dead volume (~20cm)
(Thatcham & Danner crash tests)
Gear shifting quality
Jv
kt
t
Je
e c
Jg
Side effects:
Increase of effective
volume
Increase moment ofinertia
Trade offs due to the introduction
-
7/27/2019 Canudas de Wit Innsbruck
16/73
P / 16
Trade-offs due to the introduction
of Distinctive Features
Solution 3: Multi disk clutch
Constraints:
Constraint on power-train
packaging
Gear shifting quality
Jv
kt
t
Je
e c
Jg
Side effects:
Increase of lateral
vehicle volume
Increase momentumof inertia
-
7/27/2019 Canudas de Wit Innsbruck
17/73
P / 17
Breakthrough Solutions
Integrate a force assistance device
Clutch-by-wire System
O
-
7/27/2019 Canudas de Wit Innsbruck
18/73
Outline
1. IntroductionChallenges
Product Evolution
Breakthroughs
2. ExamplesClutch SynchronizationSteer-by-Wire
3. Conclusions
O tli
-
7/27/2019 Canudas de Wit Innsbruck
19/73
Outline
1. IntroductionChallenges
Product Evolution
Breakthroughs
2. ExamplesClutch SynchronizationSteer-by-Wire
3. Conclusions
-
7/27/2019 Canudas de Wit Innsbruck
20/73
P / 20
Clutch-by-Wire Benefits
Clutch-by-Wire:
Improves comfort
Increase engine duty life
Avoid motor stall,
-
7/27/2019 Canudas de Wit Innsbruck
21/73
P / 21
Setups
Hardware Setups:
Clutch-by-Wire (CbW), and
Automated Manual Transmission (AMT)
inexpensive and efficient solutions
Control setups:Both engine and clutch torque as control inputs
Clutch torque as the only controlled input, considering the
engine torque as a known input
CbW
AMT
-
7/27/2019 Canudas de Wit Innsbruck
22/73
P / 22
Clutch & Driveline Components
k
Je
e kdmf
dmf
Jdmf
Jtl Jwl
tl
tl
d/2 wl
tireLuGre
JwrJtr
ktr
trd/2
wr
Differential
d Fxl
Fxr
tire
LuGreFn
c
Jg Vehicle
Clutch & Dual Mass
Flywheel with
nonlinear stiffness
Separate left and righttransmission branches
Tire/Ground contact
Engine
Engine torque
(known input)
Clutch normal force
Control input:
-
7/27/2019 Canudas de Wit Innsbruck
23/73
P / 23
Simple Model for Control
Jtl Jwl
ktl
tl
d/2 wl
tireLuGre
JwrJtr
ktr
trd/2
wr
Differential
d
M
Fxl
Fxr
Je
e c
Jg
Fn
kdmf
dmf
Jdmf
tireLuGre
Assuming right-left symmetry
Ignoring DMF dynamics
Ignoring tire dynamics
Equivalent moment of inertia and
stiffness
Si l M d l f C t l
-
7/27/2019 Canudas de Wit Innsbruck
24/73
P / 24
Simple Model for Control
Jtl Jw
ktl
tl
w
tireLuGre M
Fx
Je
e c
Jg
Fn
kdmf
dmf
Jdmf
Assuming right-left symmetry
Ignoring DMF dynamics
Ignoring tire dynamics
Equivalent moment of inertia and
stiffness
Si l M d l f C t l
-
7/27/2019 Canudas de Wit Innsbruck
25/73
P / 25
Simple Model for Control
Jtl Jw
ktl
tl
w
tireLuGre M
Fx
Je
e c
Jg
Fn
Assuming right-left symmetry
Ignoring DMF dynamics
Ignoring tire dynamics
Equivalent moment of inertia and
stiffness
Si l M d l f C t l
-
7/27/2019 Canudas de Wit Innsbruck
26/73
P / 26
Simple Model for Control
Jtl Jv
ktl
tl
Je
e c
Jg
Fn
Assuming right-left symmetry
Ignoring DMF dynamics
Ignoring tire dynamics
Equivalent moment of inertia and
stiffness
Simple Model for Control
-
7/27/2019 Canudas de Wit Innsbruck
27/73
P / 27
Simple Model for Control
Jv
ktl
tl
Je
e c
Jg
Fn
Assuming right-left symmetry
Ignoring DMF dynamics
Ignoring tire dynamics
Equivalent moment of inertia
and stiffness
Engagement problem
-
7/27/2019 Canudas de Wit Innsbruck
28/73
P / 28
Engagement problem
1. Synchronization
3. Avoid residual
Oscillations(Comfort)
Engine
speed
Vehicle & Gearbox
speed
2. Avoid Engine Stall (safety)
4. Minimize the
Dissipated Energy
Slipping time
Synchronization & Comfort
-
7/27/2019 Canudas de Wit Innsbruck
29/73
P / 29
Clutch Engagement:
No-Lurch condition (Glielmo SAE 2000):
y
necessary conditions
JeJe Jv
Engine Vehicle
Synchronization & Comfort
-
7/27/2019 Canudas de Wit Innsbruck
30/73
P / 30
conditions
JeJe Jg
Engine Gearbox
Jv
Vehicle
Clutch Engagement:
No-Lurch condition (Glielmo SAE 2000):
Extended ideal synchronisation & comfort conditions:
Optimal trajectory planning
-
7/27/2019 Canudas de Wit Innsbruck
31/73
P / 31
Problem:
Analytic Solution: Under the hypothesis of constant the TPBVPhas an analytic solution
Energy Minimization: Embedded in the optimization problem viathe weighting matrices
Optimal trajectory planning
Under:
Weighting parameters choice
-
7/27/2019 Canudas de Wit Innsbruck
32/73
P / 32
ensuring Safety
Engine Stall: A posteriori Search of Safe Trajectory as a function ofthe weighting matrices, i.e.
Critical Speed
Results (Simulation & Experiments)
-
7/27/2019 Canudas de Wit Innsbruck
33/73
P / 33
Results (Simulation & Experiments)
Simulations
Standing start on a flat track
Experiments
Engine Speed
Engine Speed
GearboxSpeed
GearboxSpeed
RPMRPM
Results (Experimental setup)
-
7/27/2019 Canudas de Wit Innsbruck
34/73
P / 34
Results (Experimental setup)
Clio II AMT
1.5 dCi 85hp
equipped with dSpace
Tests performed on thetrack test at Renault
Lardy Technical Centre
(South of Paris)
Nov. 2005
Experiments (Pietro in Action)
-
7/27/2019 Canudas de Wit Innsbruck
35/73
P / 35
Experiments (Pietro in Action)
Italian
Start
Starting with
Optimal Control
Inside
Summarizing
-
7/27/2019 Canudas de Wit Innsbruck
36/73
P / 36
Summarizing
Issues:Oscillation reduction below the human perception threshold,
Open-loop trajectories, implemented in closed-loop
Clutch Friction estimation
Applied to standing-start, but also useful for gear shifting
Clutch Synchronization assistance (CbW)
Outline
-
7/27/2019 Canudas de Wit Innsbruck
37/73
Outline
1. IntroductionChallenges
Product Evolution
Breakthroughs
2. ExamplesClutch Synchronization
Steer-by-Wire
3. Conclusions
Trade-offs due to the introduction
f Di ti ti F t
-
7/27/2019 Canudas de Wit Innsbruck
38/73
P / 38
of Distinctive Features
Possible solutions :
Hydraulic Power Steering + Electronic Pomp Assist
Electric Power Steering
Active Front Steering
Example: Advanced Power Steering System
Electric actuators
Steering
Assisted
Power Steering
PIGNON
Steering Gear
BIELLETTE
Variable force amplification
Variable gear ratio
Active safety systems
Packaging and Environmental constraints
Trade-offs due to the introduction
of Distincti e Feat res
-
7/27/2019 Canudas de Wit Innsbruck
39/73
P / 39
of Distinctive Features
Solution 1: HPS + Electronic Pomp Assist
Constraints:
CostConstraint on power-train
packaging
Environmental
Active safety not easy
Side effects:
Pomp assistance
dimensioning and control
Increase of effective volume
Presence of fluid
No control of steering angle
ECU
Trade-offs due to the introduction
of Distinctive Features
-
7/27/2019 Canudas de Wit Innsbruck
40/73
P / 40
of Distinctive Features
Solution 2: EPS
Constraints:
Constraint on power-trainpackaging
No optimal customer benefit
Side effects:
Steering column
Separate Torque Feedback
(comfort) or Active Steering
(Safety) are not possible
Steering
PIGNON
Steering Gear
BIELLETTE
DC motor
Trade-offs due to the introduction
of Distinctive Features
-
7/27/2019 Canudas de Wit Innsbruck
41/73
P / 41
of Distinctive Features
Solution 3: Active Front Steering (AFS)
Constraints:
Constraint on power-trainpackaging
Trade off in customer features
(comfort and active safety) andCost
Side effects:
Increase of effectivevolume
Passive or Active
Torque Feedback
and Active Safetypossible
Trade-offs due to the introduction
of Distinctive Features
-
7/27/2019 Canudas de Wit Innsbruck
42/73
P / 42
of Distinctive Features
Breakthrough Solutions
Integrate a force assistance and a steering
devices
Steer-by-wire System
Steer-by-Wire: Benefits
-
7/27/2019 Canudas de Wit Innsbruck
43/73
P / 43
Steer-by-Wire benefits:
Steering columns disappears
Easy-to-package (left & right models)
Improves active vehicle control
Steer-by-Wire: Technology
-
7/27/2019 Canudas de Wit Innsbruck
44/73
P / 44
Steer-by-Wire: Control Set-up
-
7/27/2019 Canudas de Wit Innsbruck
45/73
P / 45
Components:
Steering column
Steering gear
Driver impedance
Tire/road impedance
Controller
Control Specs:
Steering Comfort & Safety
-
7/27/2019 Canudas de Wit Innsbruck
46/73
P / 46
Steering Comfort & Safety
Comfort (Transparency):Ability of minimizing the extent towhich the steer-by-wire system alters the sensation felt by thedriver,
Comfort (Steerability): Provide power amplification to theSteering wheel to make it easy to steer,
Safety (Passivity):A passive system is inherently safer since itdoes not generate energy, but only stores, dissipates and releases it
(Li & Horowitz 99).
EH
Control Specs: Alternatives
-
7/27/2019 Canudas de Wit Innsbruck
47/73
P / 47
Comfort (Transparence, Steerability):Separate Force/position scaling,
Hybrid formulation: impedance/admittance shaping
Impedance
Admittance
Nonlinear maps
Safety (Passivity):Bilateral/Unilateral
Coupled/Uncoupled
Control alternatives:Exact Matching
Approximate matching (Multi-criteria optimization)
Idealizing Comfort:
Power Scaling
-
7/27/2019 Canudas de Wit Innsbruck
48/73
P / 48
Power Scaling
Power Scaling
EnvironmentImpedance
Separate Force/Position Scaling
Apparent
ImpedanceSeen by
the Driver
Power Scaling
Idealizing Comfortcont.
Impedance (Admittance) Shaping
-
7/27/2019 Canudas de Wit Innsbruck
49/73
P / 49
Two port ideal
Impedance model
peda ce ( d a ce) S ap g
Hybrid Force/Position Specs
AdmittanceRepresentation
Equivalent Admittance model
EApparent
Impedance
Idealizing Comfortcont.
Nonlinear admittance
-
7/27/2019 Canudas de Wit Innsbruck
50/73
P / 50
Two-Stage model from physics laws (structure)
Driver-perceivedperformance:Auto-alignment torque,
(virtual, or true)
Steering wheel friction
Torque amplification
(booster stage)
Hydraulic-like behavior
Speed vehicle dependency Torque booster
Mechanics
ApparentAdmittance
Idealizing Comfortcont.
Nonlinear admittance
-
7/27/2019 Canudas de Wit Innsbruck
51/73
P / 51
Subjective Parameters Assessment a-fastness
b-softness/hardness
(flow leakage)
c-amplification
Cmax-Saturation
b-dependencyc=f(a, Cmax)-dependency
Torque booster stage
Enforcing Safety (linear setting)
Passivity
-
7/27/2019 Canudas de Wit Innsbruck
52/73
P / 52
y
Uncoupled (Bilateral) Passivity
PassiveApparent
Admittance
PROs: very safe
Allows connection with
non-passive environments
CONs:Conservative design
Power amplification
forbidden
Enforcing Safety (linear setting)PassivityCont.
-
7/27/2019 Canudas de Wit Innsbruck
53/73
P / 53
y
PROs:Less
conservative
CONs:Knowledge of :
E, H
Coupled Unilateral Passivity
Coupled Bilateral Passivity
Control Problem: Exact Matching(Linear setting)
-
7/27/2019 Canudas de Wit Innsbruck
54/73
P / 54
PROBLEM: find K(s) such that
Exact idealized comfort is meet,
The closed-loop apparent admittance results in one of theselected passivity constraints:
Uncoupled bilateral passivity
Coupled bilateral passivity
Coupled unilateral passivity
Limitations: Solutions may not always exist.
Relaxation of this problem via optimization
Control Problem: ApproximatedMatching (Linear setting)
-
7/27/2019 Canudas de Wit Innsbruck
55/73
P / 55
Problem 1: Uncoupled bilateral passivity
MatrixTransformation
Non Convex Optimization Problem
Control Problem: ApproximatedMatching (Linear setting)
-
7/27/2019 Canudas de Wit Innsbruck
56/73
P / 56
Problem 2: Coupled unilateral passivity
Convex Optimization Problem
Knownpassive
operator
Apparent
Admittance
Control Problem: ApproximatedMatching (Robustness)
-
7/27/2019 Canudas de Wit Innsbruck
57/73
P / 57
Problem 3: Coupled Bilateral passivity
Convex Multi-criteria Optimization Problem
is a sector bounded nonlinearity,
G
K
uy
v
HE
F
-
7/27/2019 Canudas de Wit Innsbruck
58/73
Outline
-
7/27/2019 Canudas de Wit Innsbruck
59/73
1. IntroductionChallenges
Product Evolution
Breakthroughs
2. ExamplesClutch SynchronizationSteer-by-Wire
Adaptive Cruise Control
3. Conclusions
ACC: 3-Levels of Complexity
-
7/27/2019 Canudas de Wit Innsbruck
60/73
P / 60
Front/Rear longitudinal
inter-vehicular control
+ Multi-lane control
+ infrastructureexchange of information
& collaborative control.
Some Data
-
7/27/2019 Canudas de Wit Innsbruck
61/73
P / 61
The French National Observatory of Road Security has reported 90220car accidents during the year 2003.
28% correspond to accidents resulting in rear-end collisions:
Causes of accidents:
Misestimated vehicular inter-distance
Badly adapted or insufficient brakingApproaching speed too high
Erroneous estimation of the road
conditions
Statistics:
55 % Drivers do not respect the minimum 2 sec safety time
25% Drivers do not respect the minimum 50 m safety distance
Example of safety distance :Constant Time-Headway rule
-
7/27/2019 Canudas de Wit Innsbruck
62/73
P / 62
Constantand similarbraking capacity for both vehicles
Does not account for any comfort criteria.
Relative braking
distanceReaction time
distance
Threshold
distance
Newtonian motion equation
Stationary solution
Typical Safety & Comfort Specs
-
7/27/2019 Canudas de Wit Innsbruck
63/73
P / 63
Safety:Constant time headway (Chien and Ioannou, 1992),
Variable time headway (Yanakiev and Kanellakopoulos, 1995),
Potential forces (Gerdes et. al., 2001).
Comfort:
Drivers behavior model (Persson et. al., 1999),
Human perception-based model (Fancher, et. al. 2001),
Design of braking and jerks profiles (Yi & Chung, 2001).
New Specs Combining Safety & Comfort are needed
Inter-distance Vehicle Policy
-
7/27/2019 Canudas de Wit Innsbruck
64/73
P / 64
Desired Properties:
Avoid collisions (safety)
Limit Jerk (comfort)
Respect Braking Capabilities
Leader
Free zone Collision zone
Follower
Safe zone
Inter-distance Vehicle Policycont.
-
7/27/2019 Canudas de Wit Innsbruck
65/73
P / 65
Leader
Free zone Collision zone
Follower
Safe zone
Given:
Maximum velocity,
deceleration & Jerk:
Minimum inter-distance
Design goals:
Define the Safe zone
Made the orange zone
invariant
Respecting the velocity,
deceleration & jerk
constraints
Inspiration from Compliant Contact Models
-
7/27/2019 Canudas de Wit Innsbruck
66/73
P / 66
Genesis:
Hertz 1881
Marhefka & Orin, 1996
Nonlinear damping
Integral curves of
Problem Formulation
-
7/27/2019 Canudas de Wit Innsbruck
67/73
P / 67
Family of Integral
curves as a function of
the paramercc* = 0.025
c* = 0.025
c* = 0.025
Stop & Go Scenario
It li D i
-
7/27/2019 Canudas de Wit Innsbruck
68/73
P / 68
Italian Driver
Model
Comparison with the constanttime-headway policy
-
7/27/2019 Canudas de Wit Innsbruck
69/73
P / 69
Advantages:
Less conservative safe distance
Better matches the drivers average inter-vehicular distance at low
velocities.
Control Scheme: Model Tracking
-
7/27/2019 Canudas de Wit Innsbruck
70/73
P / 70
Inter-distance
Dynamics
d
wReference Model
d^
udr
Inter-distanceController
d
Warning Mode
+
+
PARAMETERS:
safety (anti-collision) comfort (jerks)
vehicle
characteristics
road conditions
traffic conditions, etc.Sensors
Experimental Results:Hard stop (ARCOS-Program)
-
7/27/2019 Canudas de Wit Innsbruck
71/73
P / 71
Jerks
Thanks to :Axel Von-arnim (LCPC) & Cyril Royre (LIVIC)
Test Track : Satory (France).
Vehicle : LOLA (LIVIC-LCPC-INRETS)
Inter-distance
Summarizing
Nonlinear dampers:
-
7/27/2019 Canudas de Wit Innsbruck
72/73
P / 72
Nonlinear dampers:More flexibility (integral curve shapes)
Continuous acceleration & Jerks (comfort assessment)
On-line adaptability:Road & traffic conditions
Model is not a true exosystem:
Model is driven by leader vehicle acceleration,Provides bounded solutions (integral curves)
Warning mode:
Use the model to provide driver assistance only
Switching Controllers are welcomed:Switching between different sensors.
-
7/27/2019 Canudas de Wit Innsbruck
73/73