Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Introduction
Robots are equipped with effectors.
Effectors
Assert a force on
the environment
Actuators
Communicates a
command to an effector
Types of Robots
1. Manipulators
Anchored to the workplace.
Common industrial robots.
2. Mobile Robots
Move using wheels, legs, etc.
Examples: delivering food in hospitals,
autonomous navigation, surveillance, etc.
Types of Robots
3. Hybrid (mobile with manipulators)
Examples: humanoid robot
(physical design mimics human torso)
Made by Honda Corp. in Japan.
Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Robot Hardware
Sensors:
a. Passive sensors.
True observers such as cameras.
b. Active sensors
Send energy into the environment,
like sonars.
Sensors
Examples of sensors:
• Tactile sensors (whiskers, bump panels)
• Global Positioning System
• Imaging sensors
• Odometry (distance travelled)
Effectors
Characterized by the degrees of freedom DF.
DF counts one for each independent direction
of movement.
6 degrees of freedom are required to place
an object at a particular orientation.
Other Types of Effectors
Unlike wheels, legs can handle tough terrains,
but they are slow on flat surfaces.
Devices vary from one leg to dozens of legs.
Robots can be
• dynamically stable
• dynamically unstable
Sources of Power
• The electric motor is the most popular source
• But you may also see:
• Pneumatic actuation using compressed
gas.
• Hydraulic actuation using pressurized
fluids.
Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Robotic Perception
Can be illustrated using a
Bayesian Belief Network.
It can be defined as a temporal inference
from sequences of actions and measurements.
Other Robotic Tasks
1. Localization
2. Mapping
3. Perception of
a. Temperature
b. Odors
c. Acoustic signals
Quantities can be estimated probabilistically.
Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Planning to Move
Types of motion:
a. Point-to-Point.
Deliver robot to target location.
b. Compliant motion.
Move while in contact to an obstacle
(robot pushing a box).
Configuration Space
Working Space:
Spatial coordinates.
Problem: not all coordinates are attainable
Configuration Space:
Represent robot joints.
With two joints we need two angles
(e.g., for shoulder and elbow).
Configuration Space
The space can be decomposed into
two subspaces:
a. Free space. Space of attainable
configurations.
b. Occupied Space. Space of unattainable
configurations.
Methods to Move
Cell Decomposition.
Decompose the free space into a number
of contiguous regions, called cells.
The problem is a discrete graph search problem.
Methods to Move
Cell Decomposition.
Disadvantages:
a. Limited to low-dimensional configurations.
b. Cells may be “mixed”.
(solution: make cells more granular).
c. Path may get too close to obstacles.
(solution: use a potential field).
Potential Field
A function defined over state space.
Value grows with distance to closest obstacle.
Tradeoff:
Minimize path length to goal while staying
away from obstacles.
Skeletonization
Reduce free space to a one-dimensional
representation.
Lower representation is called a skeleton.
Example is a Voronoi graph. (points equidistant
to two or more obstacles).
Steps:
-) Follow Voronoi graph until close to target
-) Leave graph and move to target.
Probabilistic Roadmap
Create random graph by creating a large
number of configurations.
Discard those that do not fall into free space.
Then join any two nodes by an arc if it is easy
to reach one node from the other.
Method is incomplete but scales better to high
dimensional configurations.
Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Dynamics and Control
Keeping a robot on track is not easy.
Use a controller to keep the robot on track.
Controllers that provide a force in negative
proportion to the observed error are known
as P controllers.
Dynamics and Control
Let y(t) be the reference path.
The control generated by the controller
has the form:
a(t) = K ( y(t) – x(t) )
K: gain parameter
Dynamics and Control
To achieve stability we use a PD controller
P – proportional
D – derivative
a(t) = K1 ( y(t) – x(t) ) + K2 d ( y(t) - x(t) ) / dt
K1: gain parameter
K2: differential component
Reactive Control
In some cases reflex-agents are more appropriate.
When a leg’s forward motion is blocked,
Simply retract it, lift it higher,
And try again.
Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Robotic Software
• Three layer architecture
• reactive layer
( low-level control)
• executive layer
(which reactive behavior to invoke?)
• deliberate layer
(planning)
Robotics
• Introduction
• Robot Hardware
• Robotic Perception
• Planning to Move
• Dynamics and Control
• Robotic Software
• Applications
Applications
• Industry and Agriculture
Assembly lines
Harvest, Mine
Excavate earth
• Transportation
Autonomous helicopters
Automatic wheelchairs
Transport food in hospitals
Applications
• Hazardous environments
Cleaning up nuclear waste
Collapse of World Trade Center
Transport bombs
• Exploration
Surface of Mars
Under the sea
Military activities
• Health Care (surgery)
• Personal Services
Applications
•Health Care
Surgery
• Personal Services
• Entertainment
Dog-like robots
• Human Augmentation
A Video
https://www.youtube.com/watch?v=6feEE716UEk