AE-3531

Introduction to Control Systems

Lu Gan
Aug 19, 2025
 Introduction

Lu Gan – [email protected]
 Lunar Lab @ GaTech
Our research focuses on robot vision, state estimation,
robot learning, autonomous navigation.
 Outline
• Course Logistics
• Introduction to Control
• Open-Loop Control & Feedback Control
• Control System Components
• Applications of Control
• Big Picture of the Class
 What is Control Engineering?

• Discipline concerned with modeling the dynamic

behavior of systems (analysis) and designing
systems to monitor and adjust the behavior of the
system to produce a desired response
• We need to
– Understand the system
– Model the system
– Monitor the system
– Influence the system
 Dynamic Systems

• A dynamic system is any system whose output

(or state) changes over time in response to its
input and initial conditions.
• It is governed by laws of physics (mechanical,
electrical, thermal, fluid, etc.).
• Its behavior is described mathematically by
differential equations (continuous-time) or difference
equations (discrete-time).
• E.g., mass–spring–damper, RLC circuit, room
heating system, population growth model
 Control Dynamic Systems

• Why we care about dynamic systems

• Most real-world systems evolve
• Performance matters: We want systems to respond
quickly, accurately, and stably despite disturbances
• Why the class is on controlling dynamic systems
• Static systems don’t need “control”
• Dynamics create challenges: Overshoot, oscillations,
delays, and instability all come from dynamics.
This course is about understanding, modeling, analyzing, and
designing controllers so that dynamic systems can behave the way
we want – stable, fast, accurate, robust.
Automotive Systems: Control for Fuel
Economy and Reduced Emissions
 Aerospace Applications

X-33 (Artist’s concept)

Columbia
Robotics
 What is Control Engineering?

The big question we will spend our term answering

Input Output
System

How do I to get what

change this I want here?
 Basic Control System Components

Plant = object to be controlled

u(t) y(t)
P for plant

•u(t) = controlled input

•y(t) = measured output
 Basic Control System Components

u(t) y(t)
P for plant
“other things
affecing the
d(t) n(t) x(0) output”

•u(t) = controlled input •d(t) = disturbances

•y(t) = measured output •n(t) = measurement noise
•x(0) = initial values
 Open-Loop Control Systems

Ex.

Soil Moisture
 Open-Loop Control Systems

• Open-loop control system: A system in which

the output has no affect on the control action.

• Pros & Cons?

• How to fix it?
• Feedback!
What is Feedback?
Example 1: Flyball Governor

James Watt
Other Examples of Feedback
 Feedback Control Systems

• Feedback control system/closed-loop control

system: A system that operates based on the error
which is the difference between the desired value
and the actual value.
Is controller a dynamic system

Proportional Control
 Is controller a dynamic system

Proportional–Integral–Derivative (PID) Control

 Is controller a dynamic system

• The plant is a dynamic system

• The controller is (often) a dynamic system
• Together, they form a new closed-loop dynamic
system whose behavior we analyze and design
 Control System Components
• Control Variable: The variable representing the quantity or condition that is measured
or controlled.
• Control Signal/Manipulated Variable: The quantity or condition that is varied by the
controller so as to affect the control variable.
• Plant: A piece of equipment or subsystem that performs a particular operation.
• Process: A progressively continuing operation marked by a series of changes that
succeed one another to achieve a particular result.
• System: A combination of components that act together and perform a certain
objective. These may or may not be physical, and may include more than one
plant/process.
• Disturbance: A disturbance is a signal that tends to adversely affect the output of a
system. These can be either internal or external.
• Feedback Control: Feedback control is an operation that, in presence of a disturbance,
tends to minimize the effect of the disturbance by reducing the difference between the
output and some reference value based on the size of the difference.

Control means measuring the value of the control variable and adjust the control
signal to achieve a target value of the control variable and limit its deviation.
 Control System Components
In control theory, a plant is the combination of process and actuator. An actuator is
a component of a machine that helps convert a supplied input signal into the
required form of mechanical energy.
 Control System Components
• Sensor measures the quantity that is subject to control.
• Actuator acts on the plant.
• Controller processes the sensor signals and drives the actuators.
• Control law is the rule for mapping sensor signals to actuator signals.

Feedback Control =
Sensing + Computation +
Actuation

Feedback Principles
– Robustness to
Uncertainty
– Design of Dynamics
Goals
• Stability: system maintains desired operating point (hold steady speed)
• Performance: system responds rapidly to changes (accelerate to 6 m/sec)
• Robustness: system tolerates perturbations in dynamics (mass, drag, etc)
Example 1: Temperature Control
Example 2: Speed Control
Applications of Control
 Plant Characteristics

• Linear vs. nonlinear

• Time varying vs. time invariant
• Several inputs and outputs (MIMO) vs. one input
and one output (SISO)
• Stochastic vs. deterministic (= not random)
• Distributed vs. centralized
Key Concepts for AE-3531
 Big Picture of the Course

Classical Control Modern Control

Approach: Approach:
•Transfer Functions (input–output relationships). •Based on State-Space Representation (internal state
•Analysis and design done in frequency domain (Bode, variables).
Nyquist) or via root locus. •Uses linear algebra and matrix methods.
Strengths: Strengths:
•Simple for Single-Input Single-Output •Naturally handles Multi-Input Multi-Output (MIMO)
(SISO) systems. systems.
•Intuitive graphical methods (root locus, Bode plots). •Incorporates initial conditions explicitly.
•Direct link to time-domain specs (overshoot, settling •Better suited for computer implementation and
time). simulation.
•Still widely used in industry for many practical systems •Enables advanced methods: optimal control (LQR),
(PID controllers in robotics, manufacturing, automotive, robust control, observers (Kalman filter).
etc.). Limitations:
Limitations: •More abstract, less intuitive for beginners.
•Hard to apply directly to multi-input multi-output •Requires system models in state form, which may not
(MIMO) systems. always be available.
•Doesn’t handle arbitrary initial conditions well. •In practice, many engineers still use classical tools for
•Less systematic for optimal/robust design. quick analysis.

Classical Control is the foundation, more practical, and more intuitive.

 Big Picture of the Course

Transfer Function State-Space Representation

•Input-output relation •Internal state variables

•Assume zero initial conditions •Explicitly includes initial state x(0).
•Only valid for linear time-invariant (LTI) systems. •Naturally supports multiple inputs and outputs.
•Hard to extend to multiple-input multiple-output •Modern control design: Essential for state feedback,
(MIMO) systems. observers, optimal control, robust control, etc.
•Primally frequency-domain tools in classical •Primarily time-domain tools in model control
control: Bode plots, Nyquist plots, and root locus

Both are mathematically equivalent for LTI systems — you can go from transfer
function to state-space and vice versa. The choice depends on whether you want to
focus on input-output behavior (TF) or internal dynamics (SS).
 Big Picture of the Course

Time-Domain Analysis Frequency-Domain Analysis

•Study the system’s response directly in •Study how the system responds to
the time variable, often to standard test sinusoidal inputs at different
inputs (step, impulse, ramp). frequencies.
•Control performance: how fast the •Control robustness and stability: how
system settles, how much overshoot it stable the system is when disturbances
has, steady-state error, etc. and uncertainties exist.

They are two perspectives to describe the same system. Together, they give a
complete picture of a system.
 Concluding Remarks

• AE3531 = Finite-dimensional, LTI, SISO,

deterministic plants.
• We will discuss all aspects of this problem as the
term develops! Feedback, even for LTI-SISO
plants, is a fascinating subject with tremendous
technological applications.
• Next lecture, we will briefly review dynamic
system modeling.
 References
[1] Kelly Griendling’s AE3531 Notes
[2] Jessy Grizzle’s EECS460 Notes
[3] Richard Murray’s CDS101 Notes
[4] Chance McColl’s AE3531 Notes