Control of

Mobile Robots
Dr. Magnus Egerstedt
Professor
School of Electrical and
Computer Engineering

Module 2
Mobile Robots

How make mobile robots move in effective, safe,

predictable, and collaborative ways using modern
control theory?

School of Electrical and Computer Engineering

Lecture 2.1 Driving Robots Around?

What does it take to drive a robot from point A to point B?

?
sensors

controller
robot model

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.1.1

Divide and Conquer

The world is dynamic and fundamentally unknown
The controller must be able to respond to environmental
conditions
Instead of building one complicated controller divide and
conquer: Behaviors
Go-to-goal
Avoid-obstacles
Follow-wall
Track-target

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.1.2

Behaviors

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.1.3

Behaviors

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.1.4

Lecture 2.2 Differential Drive Robots

In order to control mobile robots, we need models
Differential drive wheeled robots a very common type

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.2.1

Model 1.0
vr

(x, y)
L

x =

y =

R
2 (vr

+ v ) cos

R
2 (vr

+ v ) sin

R
L (vr

v )

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.2.2

The Unicycle Model

But it is not very natural to think in terms of wheel velocities
Go directly for translational and angular velocities
Inputs:

(x, y)

Dynamics:

x = v cos
y = v sin

=

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.2.3

Model 2.0

x = v cos
y = v sin

=

Design for this model!

x =

y =

R
2v
v = (vr + v )
= vr + v
2
R
R
L
= (vr v )
= vr v
L
R

R
2 (vr

+ v ) cos

R
2 (vr

+ v ) sin

R
L (vr

v )

2v + L
vr =
2R
2v L
v =
2R

Implement this model!

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.2.4

Lecture 2.3 Odometry

The state of the robot is (x, y, )
How do we obtain this state information?
Two possibilities:
External sensors
Internal sensors
Orientation: Compass,
Position: Accelerometers, Gyroscopes,
Wheel Encoders

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.3.1

Wheel Encoders
Wheel encoders give the distance moved by each wheel
Assume the wheels are following an arc (short time scale)
Dr
Dc

D + Dr
Dc =
2

D!

x = x + Dc cos()
y = y + Dc sin()
Dr D

=+
L
Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.3.2

Wheel Encoders
But how do we know how far each wheel has moved?
Dr
Dc
D!

For both wheels:

Assume each wheel has N

ticks per revolution
Most wheel encoders give the
total tick count since the
beginning

tick = tick tick

tick
D = 2R
N

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.3.3

A Major Disclaimer

DRIFT!!!

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.3.4

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.3.5

Lecture 2.4 Sensors

Robots need to know what the world around them looks like
The standard sensor suite includes a skirt of range sensors:
IR, Ultra-Sound, LIDAR,
Other standard external sensors include
Vision
Tactile
GPS

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.4.1

Range-Sensor Skirts
We will mainly deal with range-sensors

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.4.2

The Disk Abstraction

Instead of worrying about the resolution of the sensors, assume
we know the distance and direction to all obstacles around us
(that are close enough)

(d2 , 2 )

Angles are relative to

the robots heading

(d1 , 1 )

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.4.3

The Disk Abstraction

Instead of worrying about the resolution of the sensors, assume
we know the distance and direction to all obstacles around us
(that are close enough)

(d2 , 2 )

If we know our own pose

(position and orientation):

(d1 , 1 )

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

x1 = x + d1 cos(1 + )
y1 = y + d1 sin(1 + )

2.4.3

Example: Rendezvous

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.4.4

Example: Rendezvous

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.4.4

Lecture 2.5 Behavior-Based Robotics

The world is fundamentally
unknown and changing
Does not make sense to over-plan
Key idea: Develop a library of
useful controllers (=behaviors)
Switch among controllers in
response to environmental
changes

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.5.1

Building a Behavior v.1

Assume we have a differential-drive, wheeled mobile robot
driving at a constant speed

x = v0 cos
y = v0 sin

=

Want to drive in a desired heading

=???

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.5.2

Building a Behavior v.1

We have a reference, a model, a control input, and a tracking
error:

r = d , e = d , =

We not use PID?

= KP e + KI

ed + KD e

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.5.3

Dealing with Angles

This typically will not work since we are dealing with angles:

d = 0, = 100 e = 100
Solution: Ensure that e [, ]
Standard trick is to use atan2!

e = atan2(sin(e), cos(e)) [, ]

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.5.4

Example: Navigation
Problem: Go to a goal location without bumping in to
obstacles:

?
At a minimum, we need two behaviors: Go-to-goal and Avoidobstacles!

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.5.5

Lecture 2.6 Go-To-Goal

How drive a robot to a goal location?

x = v0 cos
y = v0 sin

=
(xg , yg )

d = arctan
(x, y)

e = d , = PID(e)

yg y
xg x

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.6.1

Attempt 1
= K(d )

ANGLES!!!

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.6.2

Attempt 2
= K(d )

GAIN TOO LOW!!!

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.6.3

Attempt 3
= Kbig (d )

JUST RIGHT!!!

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.6.4

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.6.5

Lecture 2.7 The GRITS Robot Simulator

A MATLAB-based
simulator
Simulates Khepera III
differential drive mobile
robot(-s) with IR rangesensors and wheel-encoders
in cluttered environments
Available at
http://gritslab.gatech.edu/projects/robot-simulator/
as stand-alone executable and MATLAB package
Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.7.1

Lecture 2.8 Obstacle-Avoidance

How avoid driving into obstacles?
We can use the same idea by defining a desired heading

(xo , yo )
(x, y)

There is no obvious
correct answer

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.8.1

Where to Go?
Choice depends on direction to goal.
Not pure but blended

2 = obst
2

3 = goal
Pure go-to-goal

4 = F (obst , goal )
1 = obst +

Blended

This is pure avoidance

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.8.2

Arbitration Mechanisms
This example illustrate two fundamentally different
arbitration mechanisms
Winner takes all = Hard switches
Blending = Combined behaviors
Both approaches have merit in different situations
Performance?
Analysis?
We will see how to design systematic behaviors and
arbitration mechanisms

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.8.3

Module 2: Mobile Robots

Module 2: Mobile robots
Module 3: Enough chit-chat! Lets start over in a more
systematic and formal manner!
Linear Systems

Magnus Egerstedt, Control of Mobile Robots, Georgia Ins<tute of Technology

2.8.4