Sie sind auf Seite 1von 38

Embedded Microcontroller Solutions for

Real-Time Control of Autonomous Mobile


Robots
A Ph.D. Dissertation
presented by Ioan Susnea, in February 2010
Under the supervision of Prof. Dr. Adrian
Filipescu
UNIVERSITY DUNAREA DE JOS OF GALATI,
ROMANIA
The Main Thesis of this Research
A distributed network of embedded
microcontrollers can solve the most
important problems of the
navigation control of autonomous
robots.
Justification
Such solutions may lead to significant
cost reduction for robotic systems and
can contribute to the development of
commercial Personal Robotic
Assistants for the elderly and disabled.
There is an acute social need for
solutions in this field.
Difficulties
Commonly used microcontrollers have
severe constraints in what concerns
computation power and data storage
capacity.
The tasks associated with robot control
are very complex.
Research Strategies
In order to compensate the limited
computational power of microcontrollers, a
distributed multi-processor structure has
been proposed.
Appropriate control algorithms have been
selected
New control algorithms, compatible with
microcontrollers have been developed.
Main Problems Considered in
this Research
Robot navigation problems from the
perspective of control theory
General Assumptions
KINEMATIC CONTROL
DECISION
D
A
T
A

A
C
Q
U
I
S
I
T
I
O
N
SENSORS
ACTUATORS
COMPUTING &
COMMUNCATION
EQUIPMENT
WIRELESS
LINK
AUTONOMOUS VEHICLE SMART ENVIRONMENT
DYNAMIC CONTROL
The following structure of the control system is proposed:
Each block is implemented with a distinct embedded module
Where Am I?
The Problem of Localization in Robotic
Systems
Objective of the research:
Design of an active ultrasonic beacon able to:
- Allow the measurement of the distance
between the robot and the beacon,
-Allow the measurement of the azimuth of the
beacon relative to the robot,
- Allow the identification of the beacons.
The principle of measuring the distance
between the robot and the beacon
t
t
t
t
T
r
a
n
s
m
i
t
t
e
r
R
e
c
e
i
v
e
r
Radio signal
Radio signal
Ultrasonic burst
Ultrasonic
burst
t
p
The distance is computed by measuring the propagation
time of the ultrasonic signal t
P .
The principle of measuring the azimuth
of the beacon relative to the robot
L
R
d2
d1

t
1
t
2
a
The azimuth of the sound source is determined by
measuring the interaural time delay ITD
) arcsin(
a
t c
=
The Proposed Solution
The structure of the active beacons
Structure of the Receiver Located on the
Robot
A.M.
DEMOD.
DTMF
DECODER
BURST
DETECTOR
MCU
OMNIDIRECTIONAL
MICROPHONE
RADIO
MODEM
A.M.
DEMOD.
DTMF
DECODER
OMNIDIRECTIONAL
MICROPHONE LEFT
OMNIDIRECTIONAL
MICROPHONE RIGHT
B.P. FILTER
B.P. FILTER
SQUARE
WAVE
SHAPER
SQUARE
WAVE
SHAPER
PRECEDENCE
DETECTOR
UP/DOWN
COUNTER
MCU
REF. CLOCK
BEACON ID
STROBE
Where Am I Going?
The Problem of Decision in Robotic Systems
SENSORS
Internal representation
of the environment
DECISION
GOALS
ACTIONS
An automated system is a decision and action
system.
Objective of the Research
Development of a programming language for
programming the decision in robotic systems
flexible enough to describe most of the situations
related to the navigation of an autonomous robot
in a controlled environment, and simple enough to
be accessible to users with little or no technical
skills.
We called this language RDPL - Robot Decision
Programming Language.
The Proposed Solution
The interaction between the robot and
the environment can be described by
sentences of the following type:
IF((condition
1
)&&(condition
2
)&&...&&(con
dition
N
)) THEN MOVE_TO GOAL
k
IF((condition
1
)&&(condition
2
)&&...&&(con
dition
N
)) THEN Activate_output
k
where (condition
j
) is a logic
function of the input and output of the
system.
RDPL - Robot Decision
Programming Language
The proposed language can be classified among
robot programming languages developed as
extensions of existing programming languages. It is
derived from the standard IEC61131-3, intended for
programming the PLCs.
RDPL is a compiled language, based on a reduced
set of primitives, with the following general syntax:
FNAME dest, op1, op2, op3, op4
Distinctive Features of RDPL
Inputs and outputs are distributed, being associated
with sensors and actuators located (mainly) in the
smart environment.
A limited set of inputs is associated with a set of
voice commands, such as STOP, MOVE, LEFT,
RIGHT. In this approach, the recognition of a voice
command generates a transition of the associated
digital input.
A special category of outputs is represented by the
motion commands:
MOVE TRIGGER, GOAL_X,GOAL_Y, GOAL_HEADING, EDGE
Structure of the Distributed Input-Output
System
RADIO MODEM
Input
Output
Module(s)
Voice commands
Sensors
Actuators
PROTOCOL
CONVERTER
RS-485 BUS
R
S
-
2
3
2
The Implementation of RDPL
SCAN INPUT
FETCH
EVALUATE
FUNCTION
INCREMENT PC
UPDATE
OUTPUT
RESET PC
LAST
FUNCTION?
NO
YES
PC - PROGRAM COUNTER
How Should I Get There?
Controlling the motion of the robot to
the goal
It is assumed that the robots move in a fully
known environment. Moreover, this
environment is manipulated by placing
special sensors and actuators intended for
the interaction with the robots.
Therefore, this research does not cover the
problems of map building and path
planning. It is limited to finding solutions
for defining and following paths.
Following a path defined as a trail of
Virtual Pheromones
Any attempt to model the natural pheromones
should address the following problems:
- Diffusion - is a process that creates spatial
gradients of the intensity of pheromones,
detectable by the sensors.
- Evaporation decreases the intensity of the
pheromone source with the time.
- Aggregation comes from the superposition of
the effects of multiple pheromone sources.
The model of artificial
pheromones
] 1 , 0 [ :
+
R p

\
|

=
0
1
) (
0

x
p
x p
Difusion
|

\
|
=

t
p t p 1 ) (
0
Evaporation
|

\
|

|

\
|
=

t d
p P
N
k
k
k
R
1 1
1
0
Aggregation
Additional assumptions:
S
1
S
2
S
3
S
4
S
n
d
1
d
2
d
n
p
1
p
2
p
n

P
Directive
sensitivity

=
|

\
|
=
N
k
k
k
R
d
p P
1
0
1

Time invariance
Differential sensing with two antennas
Defining the path
a
(x ,y )
R R
The map that embeds the information on the spatial
distribution of pheromones is stored in the memory of
a pheromone server, located in the environment.
Computing the pheromone intensity
for a given posture of the robot
2
0
2
0
) ( ) ( y y x x di
i i
+ =
|

\
|
=

i
i
d
p p 1
0
|
|

\
|

=
i
i
i
d
x x
0
arcsin
|
|

\
|
|

\
|
+ |

\
|
=

= =
2
1
2
1
sin cos
N
i
i i
N
i
i i R
p p P
Path following - The bio-mimetic
approach
If P
L
and P
R
are the pheromone intensities detected by two antennas L
(Left) and R (Right), the reference speeds for the drive wheels are:


=
| |
|) | 1 (
R L
R L
L
P P K
P P K
v

=
|) | 1 (
| |
R L
R L
R
P P K
P P K
v
for P >P
L R
for P <=P
L R
for P >=P
L R
for P <P
L R

PHEROMONE
SOURCES
LEFT ANTENNA
RIGHT ANTENNA
V
R
V
L
P
L
P
R
+1
-
1
+1
-
1
p
(
d
)
i
i
INPUT
LAYER
HIDDEN
LAYER
OUTPUT
LAYER
The equivalent neural network
Path Following - The Fuzzy
Approach
) ( ) ( ) ( t P t P t e
R L
=
) ( ) ( ) ( '
1
=
k k k
t e t e t e
The FLC simultaneously generates references
for v
L
and v
R
, the speeds of the differential drive
wheels.
Experiments with real robots
http://www.youtube.com/watch?v=0LJQmJS5ZdY
http://www.youtube.com/watch?v=GHBOP8HfJcI
http://www.youtube.com/watch?v=xUdYQMwD83s
http://www.youtube.com/watch?v=6UqSDp0GEzU
Avoiding Obstacles
Objective of the research
Development of a reactive obstacle
avoidance algorithm compatible with low
cost sensors and with the limitations of
microcontrollers.
The Bubble Rebound
Algorithm
A
B
Current
heading

-90
o
+90
o
0
o

sonar_readings

1
2
3
4
-1
-2
-3
-4
Defining the sensitivity bubble
and polar representation of the
information provided by sensors.
The Bubble Rebound
Algorithm
START
ADJUST HEADING
TO GOAL
OBSTACLE?
OBSTACLE?
MOVE STRAIGHT
TO GOAL
GOAL
REACHED?
STOP
COMPUTE
NEW HEADING
ADJUST
MOTION
GOAL
VISIBLE?
KEEP MOVING
YES
YES
YES
YES
NO
NO
NO
NO
O1
G
S
H
V
Computing the Rebound Angle
N

=
0
(


=
2
,
2
0
N N
i
i
i

=
=
=
2
2
2
2
N
N
i
i
N
N
i
i i
R
D
D

0
- angular pace of the distribution of sensors on
the circumference
Di - Distance reported by sensor i
Experiments with real robots
Start
Goal
Pioneer avoiding an obstacle
Published Works Related to this
Dissertation
[1] Susnea I. Mitescu M. Microcontrollers in practice ISBN:
3540253017, Springer Verlag, 2005
[2] Susnea I. Vasiliu G. Filipescu A. Radaschin A., Virtual
Pheromones for Real-Time Control of Autonomous Mobile
Robots, in Studies of Informatics and Control, Vol 18, issue 3,
2009.
[3] Susnea I . Vasiliu G, Filipescu A, Coman G., On the
Implementation of a Robotic Assistant for the Elderly. A Novel
Approach, 7
th
WSEAS Int. Conf. on Computational, Iintelligence
Man-machine Systems and Cybernatics (CIMMACS '08), Cairo,
Egipt, December 29-31, 2008, pp.215-220
Published Works Related to this
Dissertation
[4] Susnea I, Filipescu A, Vasiliu G., Filipescu S., Path
Following, Real-time, Embedded Fuzzy Control of a Mobile
Platform Wheeled Mobile Robot, IEEE International Conference
on Automation and Logistics, ICAL 2008, 1-3 Sep., Qingdao,
China, IEEE ICAL 2008 CD Proceedings, pp.91-96
[5] Susnea I, Vasiliu G, Filipescu A, Real-Time, Embedded
Fuzzy Control of the Pioneer3-DX Robot for Path Following, The
12
th
WSEAS International Conference on SYSTEMS, Heraklion,
Greece, July 22-24, 2008, pp.334-338
Published Works Related to this
Dissertation
[6] Susnea I . Vasiliu G, Filipescu A, RFID Digital Pheromones
for Generating Stigmergic Behaviour to Autonomous Mobile
Robots, 4th WSEAS Int. Conf. on Dynamic Systems and Control
(Control '08), CORFU ISLAND, GREECE, October 26-28, 2008,
pp.20-24
[7] Susnea I., Vasiliu G., Filipescu A., Coman G., Radaschin A.,
Real-Time Control of Autonomous Mobile Robots Using Virtual
Pheromones, The 7th ASCC Conference, Hong Kong 2009
[8] Susnea I., Vasiliu G., Filipescu A., Serbencu A., Radaschin
A., Virtual Pheromones to Control Mobile Robots. A Neural
Network Approach, The IEEE International Conference on
Automation and Logistics, ICAL 2009, Shenyang, China.
Published Works Related to this
Dissertation
[9] Filipescu A, Susnea I, Stancu Al, Stamatescu G, Path
following, real-time, embedded fuzzy control of a mobile
platform pioneer 3-dx, 8th WSEAS International Conference on
Systems Theory and Scientific Computation (ISTASC08),
Rhodes (Rodos) Island, Greece, August 20-22, 2008, pp.334-335
[10] Filipescu A., Susnea I., Filipescu A., Stamatescu G.
Distributed System of Mobile Platform Obstacle Avoidance and
Control As Robotic Assistant for Disabled and Elderly, To be
presented at the 7th IEEE International Conference on Control &
Automation ICCA09, Christchurch, New Zealand, Dec. 2009
[11] Filipescu A., Susnea I., Filipescu S., Stamatescu G.,
Wheeled Mobile Robot Control Using Virtual Pheromones and
Neural Networks, To be presented at the 7th IEEE International
Conference on Control & Automation ICCA09, Christchurch,
New Zealand, Dec. 2009
Conclusions
Some of the major fields of robot navigation:
localization, decision, path following, and
obstacle avoidance, have been addressed from
an interdisciplinary perspective.
The proposed solutions have practical
applications and may contribute to the evolution
of knowledge in the respective fields.
Some new research directions have been
suggested.

Das könnte Ihnen auch gefallen