Sim.I.

am: A Robot Simulator

Coursera: Control of Mobile Robots
Jean-Pierre de la Croix and Matthew Hale
Last Updated: March 9, 2016

Contents
1 Introduction 2
1.1 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 MATLAB Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Mobile Robot Simulator 3

2.1 IR Range Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Differential Wheel Drive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Wheel Encoders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 Simulator 6

4 Programming Assignments 7
4.1 Week 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2 Week 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1
1 Introduction
This manual is going to be your resource for using the simulator with the programming assignments
featured in the Coursera course, Control of Mobile Robots (and included at the end of this manual). It
will be updated from time to time whenever new features are added to the simulator or any corrections
to the course material are made.

1.1 Installation
Download simiam-coursera-week-X.zip (where X is the corresponding week number for the assignment)
from the Optional Programming Assignment page under Week X. Make sure to download a new copy of
the simulator before you start a new week’s programming assignment, or whenever an announcement is
made that a new version is available. It is important to stay up-to-date, since new versions may contain
important bug fixes or features required for the programming assignment.
Unzip the .zip file to any directory.

1.2 Requirements
You will need a reasonably modern computer to run the robot simulator. While the simulator will run
on hardware older than a Pentium 4, it will probably be a very slow experience. You will also need a
copy of MATLAB.
Thanks to support from MathWorks, a license for MATLAB and all required toolboxes are available
to all students beginning in Week 2 and lasting until the end of the course. You can easily put off
the optional Week 1 MATLAB assignment until Week 2 without falling behind. Check the
Installing MATLAB page accessible from the Week 2 page for a link that will let you install MATLAB
on your computer.

1.3 MATLAB Help

Any questions about using MATLAB for the programming assignments should be posted to the course
discussion forums.

2
 (a) Simulated QuickBot (b) Actual QuickBot

Figure 1: The simulated QuickBot and the mobile robot it is based upon.

2 Mobile Robot Simulator

The programming exercises use a simulated version of a robot called the QuickBot. The simulated
QuickBot is equipped with five infrared (IR) range sensors, of which three are located in the front and
two are located on its sides. The simulated QuickBot has a two-wheel differential drive system (two
wheels, two motors) with a wheel encoder for each wheel. You can build the QuickBot yourself by
following the hardware videos located under the Course Resources tab. These hardware videos are
not part of the course and their content is therefore unsupported by course staff. We
provide them as a courtesy to learners in the course who may be interested in them.
Figure 1 shows the simulated and actual QuickBot mobile robot. The robot simulator recreates the
QuickBot as faithfully as possible. For example, the range, output, and field of view of the simulated
IR range sensors match the specifications in the datasheet for the actual Sharp GP2D120XJ00F infrared
proximity sensors on the QuickBot.

2.1 IR Range Sensors

In this section we cover some of the details pertaining to the five simulated IR sensors onboard the
simulated QuickBot. The orientations (relative to the body of the QuickBot, as shown in Figure 1a) of
IR sensors 1 through 5 are 90◦ , 45◦ , 0◦ , −45◦ , and −90◦ , respectively. IR range sensors are effective in
the range 0.04 m to 0.3 m only. However, the IR sensors return raw values in the range of [0.4, 2.75]V
instead of the measured distances. Figure 2a demonstrates the function that maps these sensors values to
distances. To complicate matters slightly, the BeagleBone Black onboard the physical QuickBot digitizes
the analog output voltage using a voltage divider and a 12-bit, 1.8V analog-to-digital converter (ADC).
To faithfully recreate the QuickBot in simulation, we simulate the effect of this digitization. Figure 2b
is a look-up table to demonstrate the relationship between the ADC output, the analog voltage from the
IR proximity sensor, and the approximate distance that corresponds to this voltage.
Any controller can access the IR array through the robot object that is passed into its execute
function. For example,
ir_distances = robot.get_ir_distances();
for i=1:numel(robot.ir_array)
fprintf(’IR #%d has a value of %d’, i, robot.ir_array(i).get_range());
fprintf(’or %0.3f meters.\n’, ir_distances(i));
end
It is assumed that the function get ir distances properly converts from the ADC output to an

3
 3

Distance (m) Voltage (V) ADC Out

0.04 2.750 917
2.5
0.05 2.350 783
0.06 2.050 683
2 0.07 1.750 583
0.08 1.550 517
Voltage (V)

0.09 1.400 467

1.5
0.10 1.275 425
0.12 1.075 358
1 0.14 0.925 308
0.16 0.805 268
0.18 0.725 242
0.5
0.20 0.650 217
0.25 0.500 167
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0.30 0.400 133
Distance (m)

(a) Analog voltage output when an object is be- (b) A look-up table for interpolating a distance (m)
tween 0.04m and 0.3m in the IR proximity sensor’s from the analog (and digital) output voltages.
field of view.

Figure 2: A graph and a table illustrating the relationship between the distance of an object within the
field of view of an infrared proximity sensor and the analog (and digital) ouptut voltage of the sensor.

analog output voltage, and then from the analog output voltage to a distance in meters. The conversion
from ADC output to analog output voltage is simply
 
1000 · Vanalog
VADC = .
3
NOTE: For Week 2, the simulator uses a different voltage divider on the ADC; therefore,
VADC = Vanalog · 1000/2. This has been fixed in subsequent weeks!
Converting from the the analog output voltage to a distance is a little bit more complicated, because
a) the relationships between analog output voltage and distance is not linear, and b) the look-up table
provides a coarse sample of points on the curve in Figure 2a. MATLAB has a polyfit function to fit
a curve to the values in the look-up table, and a polyval function to interpolate a point on that fitted
curve. The combination of the these two functions can be use to approximate a distance based on the
analog output voltage, and this will be done as part of the programming assignment in Week 2.

2.2 Differential Wheel Drive

Since the simulated QuickBot has a differential wheel drive (i.e., is not a unicyle), it has to be controlled
by specifying the angular velocities of the right and left wheel (vr , vl ), instead of the linear and angular
velocities of a unicycle (v, ω). These velocities are computed by a transformation from (v, ω) to (vr , v` ).
Recall that the dynamics of the unicycle are defined as

ẋ = v cos(θ)
ẏ = v sin(θ) (1)
θ̇ = ω.

4
The dynamics of the differential drive are defined as
R
ẋ = (vr + v` )cos(θ)
2
R
ẏ = (vr + v` )sin(θ) (2)
2
R
θ̇ = (vr − v` ),
L
where R is the radius of the wheels and L is the distance between the wheels.
The speed of the simulated QuickBot can be set in the following way assuming that the uni to diff
function has been implemented, which transforms (v, ω) to (vr , v` ):
v = 0.15; % m/s
w = pi/4; % rad/s
% Transform from v,w to v_r,v_l and set the speed of the robot
[vel_r, vel_l] = obj.robot.dynamics.uni_to_diff(robot,v,w);
obj.robot.set_speeds(vel_r, vel_l);
The maximum angular wheel velocity for the physical QuickBot is approximately 80 RPM or 8.37
rad/s and this value is reflected in the simulator. It is therefore important to note that if the simulated
QuickBot is controlled to move at maximum linear velocity, it is not possible to achieve any angular
velocity, because the angular velocity of the wheel will have been maximized. Therefore, there exists a
tradeoff between the linear and angular velocity of the QuickBot: the faster the robot should turn, the
slower it has to move forward.

2.3 Wheel Encoders

Each of the wheels is outfitted with a wheel encoder that increments or decrements a tick counter
depending on whether the wheel is moving forward or backwards, respectively. Wheel encoders may be
used to infer the relative pose of the simulated robot. This inference is called odometry. The relevant
information needed for odometry is the radius of the wheel (32.5mm), the distance between the wheels
(99.25mm), and the number of ticks per revolution of the wheel (16 ticks/rev). For example,

R = robot.wheel_radius; % radius of the wheel

L = robot.wheel_base_length; % distance between the wheels
tpr = robot.encoders(1).ticks_per_rev; % ticks per revolution for the right wheel

fprintf(’The right wheel has a tick count of %d\n’, robot.encoders(1).state);

fprintf(’The left wheel has a tick count of %d\n’, robot.encoders(2).state);

For more information about odometry, see the Week 2 programming assignment.

5
3 Simulator
Start the simulator with the launch command in MATLAB from the command window. It is important
that this command is executed inside the unzipped folder (but not inside any of its subdirectories).

Figure 3: launch starts the simulator.

Figure 3 is a screenshot of the graphical user interface (GUI) of the simulator. The GUI can be
controlled by the bottom row of buttons. The first button is the Home button and returns you to the
home screen. The second button is the Rewind button and resets the simulation. The third button is the
Play button, which can be used to play and pause the simulation. The set of Zoom buttons or the mouse
scroll wheel allows you to zoom in and out to get a better view of the simulation. Clicking, holding, and
moving the mouse allows you to pan around the environment. You can click on a robot to follow it as it
moves through the environment.

6
4 Programming Assignments
The following sections serve as a tutorial for getting through the simulator portions of the programming
exercises. Places where you need to either edit or add code are marked off by a set of comments. For
example,

%% START CODE BLOCK %%

[edit or add code here]
%% END CODE BLOCK %%

To start the simulator with the launch command from the command window, it is important that
this command is executed inside the unzipped folder (but not inside any of its subdirectories).

4.1 Week 1
This week’s exercises will help you learn about MATLAB and robot simulator. If you do not already
have access to MATLAB, a student license will be made available to registered participants
of this course starting in Week 2. You can easily put off the optional Week 1 MATLAB
assignment until Week 2 without falling behind.

1. Install the GUI Layout Toolbox using the instructions provided in the Preparing for Programming
Assignments page under Week 1.
2. Familiarize yourself with the simulator by reading this manual and downloading the robot simulator
posted on the Optional Programming Assignment 1: Instructions section under Week 1.
3. Start the simulator by running the launch command from the command window. If the GUI
Layout Toolbox has been installed correctly (instructions for this can be found in the Preparing
for Programming Assignments page under Week 1 ), the home screen should appear. Pressing the
Play button will then show a robot which is stationary (it will move once we design controllers for
it next week!).

7
4.2 Week 2
If you are installing MATLAB this week, follow the instructions on the page titled In-
stalling MATLAB under Week 2 first. Then return to Week 1’s assignment (above) before
completing this week.
Start by downloading the robot simulator for this week from the Week 2 programming assignment.
Before you can design and test controllers in the simulator, you will need to implement three components
of the simulator:

1. Implement the transformation from unicycle dynamics to differential drive dynamics, i.e. convert
from (v, ω) to the right and left angular wheel speeds (vr , vl ).
In the simulator, (v, ω) corresponds to the variables v and w, while (vr , vl ) correspond to the
variables vel r and vel l. The function used by the controllers to convert from unicycle dynamics
to differential drive dynamics is located in +simiam/+robot/+dynamics/DifferentialDrive.m.
The function is named uni to diff, and inside of this function you will need to define vel r (vr )
and vel l (vl ) in terms of v, w, R, and L. R is the radius of a wheel, and L is the distance separating
the two wheels. Make sure to refer to Section 2.2 on “Differential Wheel Drive” for the dynamics.
2. Implement odometry for the robot, such that as the robot moves around, its pose (x, y, θ) is esti-
mated based on how far each of the wheels have turned. Assume that the robot starts at (0,0,0).
The tutorial located at www.orcboard.org/wiki/images/1/1c/OdometryTutorial.pdf covers how
odometry is computed. The general idea behind odometry is to use wheel encoders to measure the
distance the wheels have turned over a small period of time, and use this information to approximate
the change in pose of the robot.
The pose of the robot is composed of its position (x, y) and its orientation θ on a 2 dimensional
plane (note: the video lecture may refer to robot’s orientation as φ). The currently estimated
pose is stored in the variable state estimate, which bundles x (x), y (y), and theta (θ). The
robot updates the estimate of its pose by calling the update odometry function, which is located
in +simiam/+controller/+quickbot/QBSupervisor.m. This function is called every dt seconds,
where dt is 0.033s (or a little more if the simulation is running slower).

% Get wheel encoder ticks from the robot

right_ticks = obj.robot.encoders(1).ticks;
left_ticks = obj.robot.encoders(2).ticks;

% Recall the wheel encoder ticks from the last estimate

prev_right_ticks = obj.prev_ticks.right;
prev_left_ticks = obj.prev_ticks.left;

% Previous estimate
[x, y, theta] = obj.state_estimate.unpack();

% Compute odometry here

R = obj.robot.wheel_radius;
L = obj.robot.wheel_base_length;
m_per_tick = (2*pi*R)/obj.robot.encoders(1).ticks_per_rev;

The above code is already provided so that you have all of the information needed to estimate the
change in pose of the robot. right ticks and left ticks are the accumulated wheel encoder ticks
of the right and left wheel. prev right ticks and prev left ticks are the wheel encoder ticks
of the right and left wheel saved during the last call to update odometry. R is the radius of each
wheel, and L is the distance separating the two wheels. m per tick is a constant that tells you
how many meters a wheel covers with each tick of the wheel encoder. So, if you were to multiply

8
 m per tick by (right ticks-prev right ticks), you would get the distance travelled by the right
wheel since the last estimate.
Once you have computed the change in (x, y, θ) (let us denote the changes as x dt, y dt, and
theta dt) , you need to update the estimate of the pose:

theta_new = theta + theta_dt;

x_new = x + x_dt;
y_new = y + y_dt;

3. Read the ”IR Range Sensors” section in the manual and take note of the table in Figure 2b, which
maps distances (in meters) to raw IR values. Implement code that converts raw IR values to
distances (in meters).
To retrieve the distances (in meters) measured by the IR proximity sensor, you will need to imple-
ment a conversion from the raw IR values to distances in the get ir distances function located
in +simiam/+robot/Quickbot.m.

function ir_distances = get_ir_distances(obj)

ir_array_values = obj.ir_array.get_range();
ir_voltages = ir_array_values;
coeff = [];
ir_distances = polyval(coeff, ir_voltages);
end

The variable ir array values is an array of the IR raw values. Divide this array by 500 to compute
the ir voltages array. The coeff should be the coefficients returned by

coeff = polyfit(ir_voltages_from_table, ir_distances_from_table, 5);

where the first input argument is an array of IR voltages from the table in Figure 2b and the second
argument is an array of the corresponding distances from the table in Figure 2b. The third argument
specifies that we will use a fifth-order polynomial to fit to the data. Instead of running this fit every
time, execute the polyfit once in the MATLAB command line, and enter them manually on the
third line, i.e. coeff = [ ... ];. If the coefficients are properly computed, then the last line
will use polyval to convert from IR voltages to distances using a fifth-order polynomial using the
coefficients in coeff.

How to test it all

To test your code, the simulator is set to run a single P-regulator that will steer the robot to a particular
angle (denoted θd or, in code, theta d). This P-regulator is implemented in +simiam/+controller/
GoToAngle.m. If you want to change the linear velocity of the robot, or the angle to which it steers, edit
the following two lines in +simiam/+controller/+quickbot/QBSupervisor.m
obj.theta_d = pi/4;
obj.v = 0.1; %m/s
1. To test the transformation from unicycle to differential drive, first set obj.theta d=0. The robot
should drive straight forward. Now, set obj.theta d to positive or negative π4 . If positive, the robot
should start off by turning to its left, if negative it should start off by turning to its right. Note:
If you haven’t implemented odometry yet, the robot will just keep on turning in that direction.
2. To test the odometry, first make sure that the transformation from unicycle to differential drive
works correctly. If so, set obj.theta d to some value, for example π4 , and the robot’s P-regulator
should steer the robot to that angle. You may also want to uncomment the fprintf statement in
the update odometry function to print out the current estimate position to see if it make sense.
Remember, the robot starts at (x, y, θ) = (0, 0, 0).

9
3. To test the IR raw to distances conversion, edit +simiam/+controller/GoToAngle.m and uncom-
ment the following section:

% for i=1:numel(ir_distances)
% fprintf(’IR %d: %0.3fm\n’, i, ir_distances(i));
% end

This for loop will print out the IR distances. If there are no obstacles (for example, walls) around
the robot, these values should be close (if not equal to) 0.3m. Once the robot gets within range of
a wall, these values should decrease for some of the IR sensors (depending on which ones can sense
the obstacle). Note: The robot will eventually collide with the wall, because we have not designed
an obstacle avoidance controller yet!

10