Exposure to Controller Design Process:

By Means of Automatic Cruise Control Example

0
Automatic Cruise Control (ACC)

Automatic control system to maintain speed of an automobile

Can you construct a block diagram for information flow in an

ACC?

1
 Automatic Control System Fuel
Desired Throttle Actual Added, Road
Desired Vehicle
Valve Position Throttle Valve Grade
Speed
Throttle Position Vehicle
Control Speed
Valve Engine+
Algorithm
Control Gearbox
on ECU
System

Wheel
Estimated Vehicle Rotation
Speed Sensor

q Information (ie “arrows”) flow in a control system is often viewed as

an interconnection of info flowing through blocks
q“Block”: captures behavior/operation as an input-output map ie an
operator that transforms an “input” variable to an “output” variable

2
 ACC: Block Diagram Mass of fuel injected
Friction, loads Road grade, load
Desired Desired Actual
Vehicle Speed Throttle Throttle Vehicle
Throttle
Speed Position Position Engine+ Speed
Position
Control Gearbox
Control
Algo +Inertias
System Plant

Measured
Wheel speed Wheel
Speed
Sensor

Desired Actual Friction, loads

Desired Throttle
Position Motor Motor Actual
Position Voltage Voltage PMDC
Power Throttle
Control Motor+
Amplifier Position
Algorithm Inertia

Measured Throttle
Position
Potentiometer

3
The Problem and the Challenge

Feedback Controller Design Problem

Given model of P, design K such that closed-loop control
objectives are satisfied subject to constraints

n Typical control Objectives: Reference tracking, disturbance rejection,

noise rejection, robustness to model uncertainty
n Useful to remember: You are designing the behaviour of K

Why is it challenging?
n Conflicting objectives (eg reference tracking vs noise rejection)
n Signal, Model undertainty: Disturbances, sensor noise; Plant, actuator,
sensor behaviour not known precisely
n Behaviour of plant, actuator can display dynamics (past control inputs
affect the present)
n Actuator limitations (eg saturation |uac|<U)
4
ACC Controller Design Problem

ACC Controller Design Problem

Given model of P (map from actual throttle position, road grade etc to
vehicle speed & wheel speed), design K (map from desired vehicle speed &
measured wheel speed to desired throttle position) such that closed-loop
control objectives are satisfied (subject to limitation on that throttle position
actuator travel)

All usual challenges present here

n Conflicting objectives (eg reference tracking vs noise rejection)
n Signal, Model undertainty: Disturbances, sensor noise; Plant, actuator,
sensor behaviour not known precisely
n Behaviour of plant, actuator can display dynamics (past control inputs
affect the present)
n Actuator limitations (eg saturation |uac|<U)

5
Controller Design Process

n Construct a model of plant behaviour

n For the model of plant behaviour design controller behaviour

such that closed loop objectives are met
q Make simplifying but reasonable assumptions on plant, sensor,
actuator behaviour
q Controller choice depends on plant (& actuator, & sensor) behaviour
and desired performance objectives

n Note: Plant behaviour dynamic è Controller behaviour

can be dynamic (ie controllers outputs can depend on its
past inputs)

6
Control-Oriented Model of Plant
mf FL

α Te Tw Fw Fnet v
E G 1/Rw 1/M ∫

n Signals … α: actual throttle position; mf: mass of fuel injected; Te:

engine torque; Tw: torque at driven wheels; Fw: longitudinal driving force
due to engine+transmission; FL: all longitudinal loads; Fnet: net
longitudinal force;
n Systems/Operators related … E(engine): (α, mf)àTe; G (gearbox): Te
à Tw; Rw: radius of driven wheels; M: mass of car
n Major Assumptions
q Engine is treated as a static map (ie Te is assumed to change instantaneously based on on α, mf)
and a multiplication operator
q Wind drag ignored
q mf is treated as a disturbance (If air-fuel ratio is maintained precisely, α would determine mf as well
and mf would no longer be an external disturbance)

7
Control-Oriented Model of Plant
FL

α Te Tw Fw Fnet v
E G 1/Rw 1/M ∫

n With assumption of mf being determine by α (by an air-fuel ratio control

system), map from (α, FL) à v given by:

Mv(t ) = ( EG / Rw )a (t ) - FL (t )

8
 Cruise Control Algo Design Problem
n If the vehicle speed estimation system was assumed to be perfect and
perfect actuator, the design of the speed control algo boils down to
choosing α(t) such that the closed loop (ie map from (vdes, FL)àv
consisting of the control-oriented model of the plant in feedback with
speed control algo) satisfies some desired specifications

Loads (FL)

Desired Actual
Vehicle Speed Throttle Vehicle
(vdes) Position Speed (v)
Speed (α)
Control Mv(t ) = ( EG / Rw )a (t ) - FL (t )
Algo

9
ACC: Closed Loop Dynamics, P & PI Control
P Control PI Control
23
23 !"#$% '($#)*+,: .0̇ = 5 − 7"
!"#$% '($#)*+,: .0̇ = 5 − 7" 4!
4! 7889:#+; !G <=$%>=":
7889:#+; ! <=$%>=": '
5 = ?# 0$ % − 0 % + ?% ∫& 0$ I − 0 I 9I
5 = ?# (0$ % − 0 % )
<"=,89 B==C '($#)*+,
<"=,89 B==C '($#)*+,
D*%ℎ !G <=$%>=":
D*%ℎ ! <=$%>=":
'
23 23
.0̇ = ? (0 − 0) − 7" .0̇ = J? 0 − 0 + +?% L 0$ I − 0 I 9IK − 7"
4! # $ 4! # $ &
23 23 23
=> .0̇ + ?# 0 = ? 0 − 7" *8 .0̈ = ?# 0̇ $ − 0̇ + ?% (0$ − 0) − 7"̇
4! 4! # $ 4!
23 23 23 23
=> .0̈ + ?# 0̇ + ?% 0 = ?# 0̇ $ + ? 0 − 7"̇
4! 4! 4! 4! % $

Closed-loop (CL) Dynamics is influenced heavily by

choice of controller behaviour
10
ACC: P vs PI Control, Steady State Error
P Control CL Dynamics PI Control CL Dynamics

For constant vd , what happens For constant vd , what happens

to v if you wait long enough? to v if you wait long enough?

For constant FL, non zero

For constant FL , zero steady state error
steady state error

Different controller behaviours yield different

performances of CL dynamics
11
ACC: P vs PI Control, Transient Response
P Control CL Dynamics PI Control CL Dynamics

Suppose v was at a steady-state at Suppose v was at a steady-state at v= 30

v= 30 kmph. How will v behave when kmph. How will v behave when vd jumps from
vd jumps from say 30 kmph to 40 say 30 kmph to 40 kmph at say t = 5sec?
kmph at say t=5s? Assume road Assume road grade of 5deg as contributing
grade of 5deg as contributing to FL. to FL

12
What all did we do to ‘build’ our first cut ACC controllers?

n Construct block diagram for ACC system

n Build rudimentary understanding of the control objectives involved with the ACC
problem

n Use just enough knowledge about physics of plant to build control-oriented

plant model. Make simplifying assumptions of perfect sensor, actuator

n Seemingly arbitrarily chose two control structures (P Control, PI Control) as

choices for feedback controller

n Analyze behaviour of first and second order linear constant coefficient ODEs
towards deciding which structure of controller is better suited

n Simulate closed loop behaviour using a tool (that also allows you to relax some
simplifying assumptions) to check if your design controller behaviour satisfies
performance objectives

13
What we will (hopefully) learn to do for major chunk of course?

n Construct block diagrams

n Use just enough knowledge about physics of plant to build control-

oriented plant models of plant, actuators, sensors. Recognize that
plant/actuator models are often LCC ODEs

n Understand control objectives involved

n Based on understanding of the mathematics of LCC ODEs, design

control structures that accommodate plant/actuator dynamics such that
the ensuing closed loop would satisfy control objectives

n Simulate closed loop behaviour using a tool (that also allows you to
relax some simplifying assumptions) to check if your design controller
behaviour satisfies performance objectives

14