CT002-3-2 AI Methods: Robotics Control

CT002-3-2 AI Methods
Robotics Control
Learning Outcomes
• At the end of the session, you should be

able to:
– Describe what perception in the context of human and
robots.
– Understand various definitions related to
sensing/perception
– Understand variety of sensing techniques
– Understand challenges of sensing and perception in
robotics
CT002-3.5-2 AI-Methods
CE00371-1: Introduction to Software Development Problem Solving using programmed solutions
Robots
• Manufacturing robots
• Autonomous mobile robots
• Humanoid robots
Manufacturing robots
• Consists of a joint arm with a device or

end effector at the end of it such as a
gripper.
• Examples: welding, assembly and spray
painting
Autonomous mobile robots
• Remotely operated by human who can

see what the robot sees through its
camera and can directly control the
movements and actions of robot.
• Example: deliver food and medical
supplies to refugees in war.
:: QUESTION ::
Choose THREE(3) worlds from Webots,

discuss the list of actuators and sensors
with each its functionality.
Humanoid robots
• A complete robot which can interact with

their surroundings in much the same way
as human can, with head, eye, arm,
fingers and possibly even legs
• Example “Cog” developed at MIT
(Massachusetts Institute of Technology) by
Rodney Brooks and team.
Robot Architecture
Perception
The
Cognition Physical
World
Action
How Do You Design Smart Robots?
n icate
o mmu Sense Reason Act
C
Sense Reason Act
Environment
How Does a Robot “Sense”?
Can use vision…
Raw
data
Camera
“Perceived” info. (using Marker 1: Marker 2: Marker 3:

intelligent software ID = 1010110 ID = 0101010 ID = 0011000
d = 23” d = 42” d = 26”
algorithms on robot): q = 117o q = 86o q = 61o
a = 76o a = 0o a = 32o
How Does a Robot “Sense”?
Can use laser range scanner…
Raw data
(multiple scans)
Laser range
scanner
“Perceived” info. (using

intelligent software
algorithms on robot):
Example of Robots Using Laser
Range Scanners to
Cooperatively Follow the Leader
Parker, ORNL, 2002
And, there are lots of other
sensors…
• “Low-end” camera
• Infrared
• Sonar
• Microphone
• DGPS
• Compass
• Odometry (wheel encoders)
• Inclinometer (tilt)
• Tactile
• Chemical, radiation, wind, …
But it’s not so easy to sense…
What if camera breaks,

or the grass is too high,
or the lens gets dirty,
or the sensor gives bad data,
or 2 sensors tell you contradictory things,
or the lights get turned off,
or it starts raining,
or, … ?
How Does a Robot “Reason”?
“Hmm … I see that the blue

block is on the table. But I
want the blue block to be on
the red block. So, I’ll pick up
the blue block and move it.”
Current Situation Desired Situation
Other ways to “reason”…
• Constraint propagation
• Production (“expert”)
systems
• Decision networks
• Probabilistic reasoning
• Dynamic Bayesian networks
• Hidden Markov Models
• Genetic algorithms
• Neural networks
But it’s not so easy to reason…
What if blue block is glued to table,

or you accidentally drop your block,
or somebody keeps knocking down your blocks,
or the pathway is blocked,
or the building catches on fire,
or your Mom is calling,
or, … ?
How Does a Robot Convert
“Potential Fields” to Motion?
• “Go forward at 1 meter

per second”
– SetMotorState(1)
– SetSpeed(1000,0)
• “Steer hard to the right”
– SetSpeed(50,90)
• “Stop!”
– SetSpeed(0,0)
And, there are lots of other ways to
move…
• Legs
• Tracks
• Arms
• Wings…
Example of Robots Using Grippers to
Cooperate
Parker, ORNL, 1998
But, it’s not so easy to cooperate…
What if some of the robots fail,

or the wireless communication goes down,
or robots can’t find each other,
or one robot un-does what another robot just did,
or one robots thinks push while the other thinks
pull,
or one robot refuses to help another,
or, … ?
Why do we care about
Cooperating Autonomous Mobile
Robots?
• Teams of robots for search and rescue:
Robots at World Trade Center, September 2001
Teams of robots for hazardous
duty
Robots in Iraq and Afghanistan, current day
(development funded by DARPA, many contributors)
Teams of Robots for
Reconnaissance
and Homeland Security
DARPA’s vision
Parker, ORNL, 2002
Teams of Robots for Space
Exploration
And, Robots to Entertain You!
Veloso, CMU, 2003
“Old View” of Perception vs. “New
View”
• Traditional (“old view”) approach:

– Perception considered in isolation (i.e.,
disembodied)
– Perception “as king” (e.g., computer vision is

“the” problem)
– Universal reconstruction (i.e., 3D world

models)
“New View” of Perception
Perception without the context of action is meaningless.
• Action-oriented perception
– Perceptual processing tuned to meet motor activities’
needs
• Expectation-based perception
– Knowledge of world can constrain interpretation of what
is present in world
• Focus-of-attention methods
– Knowledge can constrain where things may appear in
the world
“New View” of Perception
• Active perception
– Agent can use motor control to enhance perceptual
processing via sensor positioning
• Perceptual classes
– Partition world into various categories of potential
interaction
“Open-loop” vs. “Closed-loop”
Control
• Closed-loop control system: Uses

sensory feedback from results of its output
to help compute subsequent controller
outputs
• Open-loop control system: Does not use

sensory feedback to evaluate the results
of its actions
Sensing/Perception Definitions
• Sensor: Device that measures some attribute of the world

• Transducer: Mechanism that transforms the energy
associated with what is being measured into another form of
energy
– Often used interchangeably with “sensor”
• Passive sensor: relies on environment to provide
medium/energy for observation (e.g., ambient light for
computer vision)
• Active sensor: puts out energy into the environment to
either change energy or enhance it (e.g., laser in a laser
range scanner)
• Active sensing: system for using an effector to dynamically
position a sensor for a “better look”
• “Active sensor” = “Active sensing”:
Behavioral Sensor Fusion
• Sensor suite: set of sensors for a particular robot

• Sensor fusion: any process that combines information
from multiple sensors into a single percept
• Multiple sensors used when:
– A particular sensor is too imprecise or noisy to give
reliable data
• Sensor reliability problems:
– False positive:
• Sensor leads robot to believe a percept is present
when it isn’t
– False negative:
• Sensor causes robot to miss a percept that is
actually present
Exercise
Describe the following perception in

robotics:
•Action-oriented perception
•Expectation-based perception
•Focus-of-attention methods
•Active perception
•Perceptual classes
Localization : Using
Kalman filter
Initial state
detects nothing:
Moves and
detects landmark:
Moves and
detects nothing:
Moves and
detects landmark:
The Kalman Filter
• Motion model is Gaussian…
• Sensor model is Gaussian…
• Each belief function is uniquely characterized
by its mean m and covariance matrix 
• Computing the posterior means computing a
new mean m and covariance  from old data
using actions and sensor readings
• What are the key limitations?
1) Unimodal distribution
2) Linear assumptions
Kalman Filter Components
(also known as: Way Too Many Variables…)
Linear discrete time dynamic system (motion model)

State Control input Process noise
xt 1  Ft xt  Bt ut  Gt wt
State transition Control input Noise input

function function function with covariance Q
Measurement equation (sensor model)

Sensor reading State Sensor noise with covariance R
zt 1  H t 1 xt 1  nt 1
Sensor function Note:Write these down!!!
The Kalman Filter…
Propagation (motion model):
xˆt 1/ t  Ft xˆt / t  Bt ut - State estimate is updated from system dynamics
T T
Pt 1/ t  Ft Pt / t Ft  Gt Qt Gt - Uncertainty estimate GROWS
Update (sensor model):

zˆt 1  H t 1 xˆt 1/ t - Compute expected value of sensor reading
rt 1  zt 1  zˆt 1 - Compute the difference between expected and “true”

T
St 1  H t 1 Pt 1/ t H t 1  Rt 1 - Compute covariance of sensor reading
T 1
K t 1  Pt 1/ t H t 1 St 1 - Compute the Kalman Gain (how much to correct est.)
xˆt 1/ t 1  xˆt 1/ t  K t 1rt 1 - Multiply residual times gain to correct state estimate
T 1
Pt 1/ t 1  Pt 1/ t  Pt 1/ t H t 1 S t 1 H t 1 Pt 1/ t - Uncertainty estimate SHRINKS
Kalman Filter Block Diagram
Some observations
• The larger the error, the smaller the effect on the final
state estimate
– If process uncertainty is larger, sensor updates will dominate
state estimate
– If sensor uncertainty is larger, process propagation will
dominate state estimate
• Improper estimates of the state and/or sensor
covariance may result in a rapidly diverging estimator
– As a rule of thumb, the residuals must always be bounded
within a ±3s region of uncertainty
– This measures the “health” of the filter
• Many propagation cycles can happen between
updates
Problems with the
Linear Model Assumption
• Many systems of interest are highly non-
linear, such as mobile robots
• In order to model such systems, a linear
process model must be generated out of the
non-linear system dynamics
• The Extended Kalman filter is a method by
which the state propagation equations and
the sensor models can be linearized about
the current state estimate
• Linearization will increase the state error
residual because it is not the best estimate
Q&A
Next Topic
Simulation and Modeling
Distinguish between simulation and modelling

What is :-
a. Simulator
b. Model
c. Real system
Process of simulation
Issues in simulation

CT002-3-2 AI Methods: Robotics Control

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CT002-3-2 AI Methods: Robotics Control

Uploaded by

Copyright:

Available Formats

CT002-3-2 AI Methods

• At the end of the session, you should be

• Consists of a joint arm with a device or

• Remotely operated by human who can

Choose THREE(3) worlds from Webots,

• A complete robot which can interact with

Sense Reason Act

“Perceived” info. (using Marker 1: Marker 2: Marker 3:

“Perceived” info. (using

Parker, ORNL, 2002

What if camera breaks,

“Hmm … I see that the blue

Current Situation Desired Situation

What if blue block is glued to table,

• “Go forward at 1 meter

Parker, ORNL, 1998

What if some of the robots fail,

Robots at World Trade Center, September 2001

Robots in Iraq and Afghanistan, current day

(development funded by DARPA, many contributors)

Parker, ORNL, 2002

Veloso, CMU, 2003

• Traditional (“old view”) approach:

– Perception “as king” (e.g., computer vision is

– Universal reconstruction (i.e., 3D world

• Closed-loop control system: Uses

• Open-loop control system: Does not use

• Sensor: Device that measures some attribute of the world

• Sensor suite: set of sensors for a particular robot

Describe the following perception in

Linear discrete time dynamic system (motion model)

State transition Control input Noise input

Measurement equation (sensor model)

Propagation (motion model):

xˆt 1/ t  Ft xˆt / t  Bt ut - State estimate is updated from system dynamics

Update (sensor model):

rt 1  zt 1  zˆt 1 - Compute the difference between expected and “true”

Simulation and Modeling

Distinguish between simulation and modelling

You might also like