Robotics.ppt

Smart Home Technologies
Automation and Robotics

Motivation
 Intelligent Environments are aimed at improving
the inhabitants’ experience and task
performance
 Automate functions in the home
 Provide services to the inhabitants
 Decisions coming from the decision maker(s) in
the environment have to be executed.
 Decisions require actions to be performed on devices
 Decisions are frequently not elementary device
interactions but rather relatively complex commands
 Decisions define set points or results that have to be
achieved
 Decisions can require entire tasks to be performed

Automation and Robotics in
Intelligent Environments
 Control of the physical environment
 Automated blinds
 Thermostats and heating ducts
 Automatic doors
 Automatic room partitioning
 Personal service robots
 House cleaning
 Lawn mowing
 Assistance to the elderly and handicapped
 Office assistants
 Security services

Robots
 Robota (Czech) = A worker of forced labor
From Czech playwright Karel Capek's 1921 play “R.U.R”
(“Rossum's Universal Robots”)
 Japanese Industrial Robot Association (JIRA) :
“A device with degrees of freedom that can be
controlled.”
 Class 1 : Manual handling device
 Class 2 : Fixed sequence robot
 Class 3 : Variable sequence robot
 Class 4 : Playback robot
 Class 5 : Numerical control robot
 Class 6 : Intelligent robot

A Brief History of Robotics
 Mechanical Automata
 Ancient Greece & Egypt
 Water powered for ceremonies
 14th – 19th century Europe
 Clockwork driven for entertainment
 Motor driven Robots
 1928: First motor driven automata
 1961: Unimate
 First industrial robot
 1967: Shakey
 Autonomous mobile research robot
 1969: Stanford Arm
 Dextrous, electric motor driven robot arm
Maillardet’s Automaton
Unimate

Robots
 Robot Manipulators
 Mobile Robots

Robots
 Walking Robots
 Humanoid Robots

Autonomous Robots
 The control of autonomous robots involves a
number of subtasks
 Understanding and modeling of the mechanism
 Kinematics, Dynamics, and Odometry
 Reliable control of the actuators
 Closed-loop control
 Generation of task-specific motions
 Path planning
 Integration of sensors
 Selection and interfacing of various types of sensors
 Coping with noise and uncertainty
 Filtering of sensor noise and actuator uncertainty
 Creation of flexible control policies
 Control has to deal with new situations

Traditional Industrial Robots
 Traditional industrial robot control uses robot
arms and largely pre-computed motions
 Programming using “teach box”
 Repetitive tasks
 High speed
 Few sensing operations
 High precision movements
 Pre-planned trajectories and
task policies
 No interaction with humans

Problems
 Traditional programming techniques for
industrial robots lack key capabilities necessary
in intelligent environments
 Only limited on-line sensing
 No incorporation of uncertainty
 No interaction with humans
 Reliance on perfect task information
 Complete re-programming for new tasks

Requirements for Robots in
 Autonomy
 Robots have to be capable of achieving task
objectives without human input
 Robots have to be able to make and execute their
own decisions based on sensor information
 Intuitive Human-Robot Interfaces
 Use of robots in smart homes can not require
extensive user training
 Commands to robots should be natural for
inhabitants
 Adaptation
 Robots have to be able to adjust to changes in the
environment

Robots for Intelligent
Environments
 Service Robots
 Security guard
 Delivery
 Cleaning
 Mowing
 Assistance Robots
 Mobility
 Services for elderly and
People with disabilities

Autonomous Robot Control
 To control robots to perform tasks
autonomously a number of tasks have to be
addressed:
 Modeling of robot mechanisms
 Kinematics, Dynamics
 Robot sensor selection
 Active and passive proximity sensors
 Low-level control of actuators
 Closed-loop control
 Control architectures
 Traditional planning architectures
 Behavior-based control architectures
 Hybrid architectures

 Forward kinematics describes how the robots
joint angle configurations translate to locations
in the world
 Inverse kinematics computes the joint angle
configuration necessary to reach a particular
point in space.
 Jacobians calculate how the speed and
configuration of the actuators translate into
velocity of the robot
Modeling the Robot Mechanism
(x, y, z)
1
2
(x, y, )

 In mobile robots the same configuration in
terms of joint angles does not identify a unique
location
 To keep track of the robot it is necessary to
incrementally update the location (this process is
called odometry or dead reckoning)
 Example: A differential drive robot
Mobile Robot Odometry
(x, y, )
t
v
v
y
x
y
x
y
x
t
t
t






































 
R
L
R
L
y
R
L
x
d
r
r
v
r
v





















2
)
(
)
sin(
,
2
)
(
)
cos(
R
L

Actuator Control
 To get a particular robot actuator to a particular
location it is important to apply the correct
amount of force or torque to it.
 Requires knowledge of the dynamics of the robot
 Mass, inertia, friction
 For a simplistic mobile robot: F = m a + B v
 Frequently actuators are treated as if they were
independent (i.e. as if moving one joint would not
affect any of the other joints).
 The most common control approach is PD-control
(proportional, differential control)
 For the simplistic mobile robot moving in the x direction:
   
actual
desired
D
actual
desired
P v
v
K
x
x
K
F 




Robot Navigation
 Path planning addresses the task of computing
a trajectory for the robot such that it reaches
the desired goal without colliding with obstacles
 Optimal paths are hard to compute in particular for
robots that can not move in arbitrary directions (i.e.
nonholonomic robots)
 Shortest distance paths can be dangerous since they
always graze obstacles
 Paths for robot arms have to take into account the
entire robot (not only the endeffector)

Sensor-Driven Robot Control
 To accurately achieve a task in an intelligent
environment, a robot has to be able to react
dynamically to changes ion its surrounding
 Robots need sensors to perceive the environment
 Most robots use a set of different sensors
 Different sensors serve different purposes
 Information from sensors has to be integrated into
the control of the robot

Robot Sensors
 Internal sensors to measure the robot
configuration
 Encoders measure the rotation angle of a joint
 Limit switches detect when the joint has reached the
limit

Robot Sensors
 Proximity sensors are used to measure the distance or
location of objects in the environment. This can then be
used to determine the location of the robot.
 Infrared sensors determine the distance to an object by
measuring the amount of infrared light the object reflects back
to the robot
 Ultrasonic sensors (sonars) measure the time that an ultrasonic
signal takes until it returns to the robot
 Laser range finders determine distance by
measuring either the time it takes for a laser
beam to be reflected back to the robot or by
measuring where the laser hits the object

 Computer Vision provides robots with the
capability to passively observe the environment
 Stereo vision systems provide complete location
information using triangulation
 However, computer vision is very complex
 Correspondence problem makes stereo vision even more
difficult
Robot Sensors

Uncertainty in Robot Systems
Robot systems in intelligent environments have to
deal with sensor noise and uncertainty
 Sensor uncertainty
Sensor readings are imprecise and unreliable
 Non-observability
Various aspects of the environment can not be observed
The environment is initially unknown
 Action uncertainty
Actions can fail
Actions have nondeterministic outcomes

Probabilistic Robot Localization
Explicit reasoning about
Uncertainty using Bayes
filters:
Used for:
 Localization
 Mapping
 Model building
1
1
1
1 )
(
)
,
|
(
)
|
(
)
( 




 t
t
t
t
t
t
t
t dx
x
b
a
x
x
p
x
o
p
x
b 

Deliberative
Robot Control Architectures
 In a deliberative control architecture the robot
first plans a solution for the task by reasoning
about the outcome of its actions and then
executes it
 Control process goes through a sequence of sencing,
model update, and planning steps

Deliberative
Control Architectures
 Advantages
 Reasons about contingencies
 Computes solutions to the given task
 Goal-directed strategies
 Problems
 Solutions tend to be fragile in the presence of
uncertainty
 Requires frequent replanning
 Reacts relatively slowly to changes and unexpected
occurrences

Behavior-Based
 In a behavior-based control architecture the
robot’s actions are determined by a set of
parallel, reactive behaviors which map sensory
input and state to actions.

Behavior-Based
 Reactive, behavior-based control combines
relatively simple behaviors, each of which
achieves a particular subtask, to achieve the
overall task.
 Robot can react fast to changes
 System does not depend on complete knowledge of
the environment
 Emergent behavior (resulting from combining initial
behaviors) can make it difficult to predict exact
behavior
 Difficult to assure that the overall task is achieved

 Complex behavior can be achieved using very
simple control mechanisms
 Braitenberg vehicles: differential drive mobile robots
with two light sensors
 Complex external behavior does not necessarily require a
complex reasoning mechanism
Complex Behavior from Simple
Elements: Braitenberg Vehicles
+ +
“Coward” “Aggressive”
+ + - -
“Love” “Explore”
- -

Behavior-Based Architectures:
Subsumption Example
 Subsumption architecture is one of the earliest
behavior-based architectures
 Behaviors are arranged in a strict priority order
where higher priority behaviors subsume lower
priority ones as long as they are not inhibited.

Subsumption Example
 A variety of tasks can be robustly performed
from a small number of behavioral elements
© MIT AI Lab
http://www-robotics.usc.edu/~maja/robot-video.mpg

Reactive, Behavior-Based
Control Architectures
 Advantages
 Reacts fast to changes
 Does not rely on accurate models
 “The world is its own best model”
 No need for replanning
 Problems
 Difficult to anticipate what effect combinations of
behaviors will have
 Difficult to construct strategies that will achieve
complex, novel tasks
 Requires redesign of control system for new tasks

Hybrid Control Architectures
 Hybrid architectures combine
reactive control with abstract
task planning
 Abstract task planning layer
 Deliberative decisions
 Plans goal directed policies
 Reactive behavior layer
 Provides reactive actions
 Handles sensors and actuators

Hybrid Control Policies
Task Plan:
Behavioral
Strategy:

Example Task:
Changing a Light Bulb

Hybrid Control Architectures
 Advantages
 Permits goal-based strategies
 Ensures fast reactions to unexpected changes
 Reduces complexity of planning
 Problems
 Choice of behaviors limits range of possible tasks
 Behavior interactions have to be well modeled to be
able to form plans

Traditional Human-Robot
Interface: Teleoperation
Remote Teleoperation: Direct
operation of the robot by the
user
 User uses a 3-D joystick or an
exoskeleton to drive the robot
 Simple to install
 Removes user from dangerous areas
 Problems:
 Requires insight into the mechanism
 Can be exhaustive
 Easily leads to operation errors

Human-Robot Interaction in
 Personal service robot
 Controlled and used by untrained users
 Intuitive, easy to use interface
 Interface has to “filter” user input
 Eliminate dangerous instructions
 Find closest possible action
 Receive only intermittent commands
 Robot requires autonomous capabilities
 User commands can be at various levels of complexity
 Control system merges instructions and autonomous
operation
 Interact with a variety of humans
 Humans have to feel “comfortable” around robots
 Robots have to communicate intentions in a natural way

Example: Minerva the Tour
Guide Robot (CMU/Bonn)
© CMU Robotics Institute
http://www.cs.cmu.edu/~thrun/movies/minerva.mpg

Intuitive Robot Interfaces:
Command Input
 Graphical programming interfaces
 Users construct policies form elemental blocks
 Problems:
 Requires substantial understanding of the robot
 Deictic (pointing) interfaces
 Humans point at desired targets in the world or
 Target specification on a computer screen
 Problems:
 How to interpret human gestures ?
 Voice recognition
 Humans instruct the robot verbally
 Problems:
 Speech recognition is very difficult
 Robot actions corresponding to words has to be defined

Intuitive Robot Interfaces:
Robot-Human Interaction
 He robot has to be able to communicate its
intentions to the human
 Output has to be easy to understand by humans
 Robot has to be able to encode its intention
 Interface has to keep human’s attention without
annoying her
 Robot communication devices:
 Easy to understand computer screens
 Speech synthesis
 Robot “gestures”

Human-Robot Interfaces
 Existing technologies
 Simple voice recognition and speech synthesis
 Gesture recognition systems
 On-screen, text-based interaction
 Research challenges
 How to convey robot intentions ?
 How to infer user intent from visual observation (how
can a robot imitate a human) ?
 How to keep the attention of a human on the robot ?
 How to integrate human input with autonomous
operation ?

Integration of Commands and
Autonomous Operation
 Adjustable Autonomy
 The robot can operate at
varying levels of autonomy
 Operational modes:
 Autonomous operation
 User operation / teleoperation
 Behavioral programming
 Following user instructions
 Imitation
 Types of user commands:
 Continuous, low-level
instructions (teleoperation)
 Goal specifications
 Task demonstrations
Example System

"Social" Robot Interactions
 To make robots acceptable to average users
they should appear and behave “natural”
 "Attentional" Robots
 Robot focuses on the user or the task
 Attention forms the first step to imitation
 "Emotional" Robots
 Robot exhibits “emotional” responses
 Robot follows human social norms for behavior
 Better acceptance by the user (users are more forgiving)
 Human-machine interaction appears more “natural”
 Robot can influence how the human reacts

"Social" Robot Example: Kismet
© MIT AI Lab
http://www.ai.mit.edu/projects/cog/Video/kismet/kismet_face_30fps.mpg

"Social" Robot Interactions
 Advantages:
 Robots that look human and that show “emotions”
can make interactions more “natural”
 Humans tend to focus more attention on people than on
objects
 Humans tend to be more forgiving when a mistake is
made if it looks “human”
 Robots showing “emotions” can modify the way in
which humans interact with them
 Problems:
 How can robots determine the right emotion ?
 How can “emotions” be expressed by a robot ?

Human-Robot Interfaces for
 Robot Interfaces have to be easy to use
 Robots have to be controllable by untrained users
 Robots have to be able to interact not only with their
owner but also with other people
 Robot interfaces have to be usable at the
human’s discretion
 Human-robot interaction occurs on an irregular basis
 Frequently the robot has to operate autonomously
 Whenever user input is provided the robot has to react to it
 Interfaces have to be designed human-centric
 The role of the robot is it to make the human’s life
easier and more comfortable (it is not just a tech toy)

 Intelligent Environments are non-stationary and
change frequently, requiring robots to adapt
 Adaptation to changes in the environment
 Learning to address changes in inhabitant preferences
 Robots in intelligent environments can frequently
not be pre-programmed
 The environment is unknown
 The list of tasks that the robot should perform might
not be known beforehand
 No proliferation of robots in the home
 Different users have different preferences
Adaptation and Learning for
Robots in Smart Homes

Adaptation and Learning
In Autonomous Robots
 Learning to interpret sensor information
 Recognizing objects in the environment is difficult
 Sensors provide prohibitively large amounts of data
 Programming of all required objects is generally not
possible
 Learning new strategies and tasks
 New tasks have to be learned on-line in the home
 Different inhabitants require new strategies even for
existing tasks
 Adaptation of existing control policies
 User preferences can change dynamically
 Changes in the environment have to be reflected

Learning Approaches for
Robot Systems
 Supervised learning by teaching
 Robots can learn from direct feedback from the
user that indicates the correct strategy
 The robot learns the exact strategy provided by the user
 Learning from demonstration (Imitation)
 Robots learn by observing a human or a robot
perform the required task
 The robot has to be able to “understand” what it observes
and map it onto its own capabilities
 Learning by exploration
 Robots can learn autonomously by trying different
actions and observing their results
 The robot learns a strategy that optimizes reward

Learning Sensory Patterns
Chair
 Learning to Identify Objects
 How can a particular object be
recognized ?
 Programming recognition strategies is
difficult because we do not fully
understand how we perform recognition
 Learning techniques permit the robot
system to form its own recognition
strategy
 Supervised learning can be used by
giving the robot a set of pictures and
the corresponding classification
 Neural networks
 Decision trees
:
:
:
:

Learning Task Strategies by
Experimentation
 Autonomous robots have to be able to learn
new tasks even without input from the user
 Learning to perform a task in order to optimize the
reward the robot obtains (Reinforcement Learning)
 Reward has to be provided either by the user or the
environment
 Intermittent user feedback
 Generic rewards indicating unsafe or inconvenient actions or
occurrences
 The robot has to explore its actions to determine what
their effects are
 Actions change the state of the environment
 Actions achieve different amounts of reward
 During learning the robot has to maintain a level of safety

Example: Reinforcement
Learning in a Hybrid Architecture
 Policy Acquisition Layer
 Learning tasks without
supervision
 Abstract Plan Layer
 Learning a system model
 Basic state space compression
 Reactive Behavior Layer
 Initial competence and
reactivity

Example Task:
Learning to Walk

Scaling Up: Learning Complex
Tasks from Simpler Tasks
 Complex tasks are hard to learn since they
involve long sequences of actions that have to
be correct in order for reward to be obtained
 Complex tasks can be learned as shorter
sequences of simpler tasks
 Control strategies that are expressed in terms of
subgoals are more compact and simpler
 Fewer conditions have to be considered if simpler
tasks are already solved
 New tasks can be learned faster
 Hierarchical Reinforcement Learning
 Learning with abstract actions
 Acquisition of abstract task knowledge

Conclusions
 Robots are an important component in Intelligent
Environments
 Automate devices
 Provide physical services
 Robot Systems in these environments need particular
capabilities
 Autonomous control systems
 Simple and natural human-robot interface
 Adaptive and learning capabilities
 Robots have to maintain safety during operation
 While a number of techniques to address these
requirements exist, no functional, satisfactory solutions
have yet been developed
 Only very simple robots for single tasks in intelligent
environments exist

Robotics.ppt

More Related Content

Similar to Robotics.ppt (20)

More from UmaDeviAnanth (6)

Recently uploaded (20)

Robotics.ppt